From stolcke@ICSI.Berkeley.EDU Tue Jul 5 15:29:39 1994 Return-Path: Received: from cli.com by jingles.cli.com.cliftp (4.1/SMI-4.1) id AA02813; Tue, 5 Jul 94 15:29:39 CDT Received: from icsia.ICSI.Berkeley.EDU by cli.com (4.1/SMI-4.1) id AA21383; Tue, 5 Jul 94 14:24:55 CDT Received: from tiramisu (stolcke@tiramisu.ICSI.Berkeley.EDU [128.32.201.72]) by icsia.ICSI.Berkeley.EDU (8.6.8/HUB+V8$Revision: 1.21 $) with ESMTP id MAA01102; Tue, 5 Jul 1994 12:10:17 -0700 Received: from localhost (stolcke@localhost) by tiramisu (8.6.8/1.8) with ESMTP id MAA05980; Tue, 5 Jul 1994 12:10:14 -0700 Message-Id: <199407051910.MAA05980@tiramisu> X-Authentication-Warning: tiramisu: stolcke owned process doing -bs To: schelter@posso.ibp.fr (William Schelter) Cc: gcl@cli.com, tkunze@vader.kgw.TU-Berlin.DE, poirot@aio.jsc.nasa.gov, cwitty@csl.sri.com, kcng@iet.com, bm1822@adasv2.sbc.com Subject: Re: pcl (clos) under gcl 1.0 In-Reply-To: Your message of Tue, 05 Jul 1994 18:59:00 +0200. <9407051659.AA14629@posso.ibp.fr> Date: Tue, 05 Jul 1994 12:10:13 PDT From: Andreas Stolcke In message <9407051659.AA14629@posso.ibp.fr>you wrote: > I have put a complete version of PCL (the xerox common lisp object system) > adapted for GNU common lisp (GCL version 1.0) on > great! a small problem seems to remain: defclass.lisp makes use of the CltL2 macro `declaim' which is not defined in GCL. It's trivial to fix it to use the old `proclaim' instead. I was successful in build pcl-gcl for sun4 and testing it on some 20k lines of CLOS code from my thesis. I wasn't as lucky with the sgi4d version. I fails to compile gcl-low.c because cc complains about the line use_fast_links(); generated at line 1877. Any ideas? s --Andreas *** defclass.lisp.orig Mon Dec 21 13:24:38 1992 --- defclass.lisp Tue Jul 5 10:39:11 1994 *************** *** 178,189 **** (if defstruct-p '(load eval) *defclass-times*) `(progn ,@(mapcar #'(lambda (x) ! `(declaim (ftype (function (t) t) ,x))) #+cmu *readers* #-cmu nil) ,@(mapcar #'(lambda (x) #-setf (when (consp x) (setq x (get-setf-function-name (cadr x)))) ! `(declaim (ftype (function (t t) t) ,x))) #+cmu *writers* #-cmu nil) (let ,(mapcar #'cdr *initfunctions*) (load-defclass ',name --- 178,189 ---- (if defstruct-p '(load eval) *defclass-times*) `(progn ,@(mapcar #'(lambda (x) ! `(proclaim '(ftype (function (t) t) ,x))) #+cmu *readers* #-cmu nil) ,@(mapcar #'(lambda (x) #-setf (when (consp x) (setq x (get-setf-function-name (cadr x)))) ! `(proclaim '(ftype (function (t t) t) ,x))) #+cmu *writers* #-cmu nil) (let ,(mapcar #'cdr *initfunctions*) (load-defclass ',name From chrisp@max.tiac.net Fri Jul 15 07:26:13 1994 Return-Path: Received: from cli.com by jingles.cli.com.cliftp (4.1/SMI-4.1) id AA03840; Fri, 15 Jul 94 07:26:13 CDT Received: from max.tiac.net by cli.com (4.1/SMI-4.1) id AA13439; Fri, 15 Jul 94 06:26:51 CDT Received: (from chrisp@localhost) by max.tiac.net (8.6.8/8.6.6.Beta9) id HAA13751; Fri, 15 Jul 1994 07:23:42 -0400 Date: Fri, 15 Jul 1994 07:23:42 -0400 From: Chris Patti { Feoh } Message-Id: <199407151123.HAA13751@max.tiac.net> To: wfs@math.utexas.edu Cc: gcl@cli.com Subject: Here are the diffs to gcl.el that made it work for me.. Hi. I had to bash gcl.el a bit to get it to work. Here is the diff: 83c83,84 < (require 'shell) --- > (require 'sshell) > (load "inf-lisp") 97c98,100 < (t (inferior-lisp-mode)))) --- > (t > (inferior-lisp-mode) > (setq inferior-lisp-buffer (get-buffer "*gcl-lisp*"))))) 347d349 < 'run' was failing on make-shell because you had required 'shell instead of 'sshell. The bigger problem was that you were trying to use inferior-lisp-mode which doesn't come pre-loaded . it's only loaded after inferior-lisp is called which autoloads the rest of the package.. The way I got around this is to load inf-lisp.el which contains the appropriate code. The only other problem was that once we'd successfully entered inferior-lisp- mode emacs didn't know which lisp to send to, so I set the inferior-lisp-buffer variable to the buffer named *gcl-lisp* which gets created by make-shell. That's about it. Am I missing something blatantly obvious? Did I do all this for naught because I forgot some subtle trick in the installation procedure? -Chris P.S. It's so short I'm going to include the full gcl.el here since I'm not sure I did the right thing with diff. From rcthakur@map.iitd.ernet.in Tue Oct 18 04:51:55 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02281; Tue, 18 Oct 94 04:51:55 CDT Received: from sangam.ncst.ernet.in by cli.com (4.1/SMI-4.1) id AA19734; Tue, 18 Oct 94 03:41:41 CDT Received: (from root@localhost) by sangam.ncst.ernet.in (8.6.8.1/8.6.6) with UUCP id OAA22204; Tue, 18 Oct 1994 14:11:10 +0500 Received: from map.iitd.ernet.in (netearth) by henna.iitd.ernet.in (4.1/SMI-4.1-MHS-7.0 ) id AA16123; Tue, 18 Oct 94 14:10:49 IST X-Organisation: Indian Institute of Technology, New Delhi. Received: by map.iitd.ernet.in; id AA01059; Tue, 18 Oct 1994 14:10:48 +0530 Received: by map; Tue, 18 Oct 1994 14:10:48 +0530 id AA01059; Tue, 18 Oct 1994 14:10:48 +0530 Date: Tue, 18 Oct 1994 14:10:47 +0530 (IST) From: "Mr. R.C. Thakur SSO-II" Subject: To: gcl-request@cli.com Cc: gcl@cli.com Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII subscribe mailing list From @festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk Tue Oct 18 13:05:13 1994 Return-Path: <@festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk> Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02318; Tue, 18 Oct 94 13:05:13 CDT Received: from sun2.nsfnet-relay.ac.uk by cli.com (4.1/SMI-4.1) id AA20943; Tue, 18 Oct 94 11:46:04 CDT Via: uk.ac.edinburgh.festival; Tue, 18 Oct 1994 16:43:48 +0100 Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa03962; 18 Oct 94 16:44 BST Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Tue, 18 Oct 94 16:42:48 BST Date: Tue, 18 Oct 94 16:42:47 BST Message-Id: <22533.9410181542@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: null-terminated strings? To: gcl@cli.com Are strings in GCL null-terminated? It would make it much easier to pass them to C functions if they were. -- jeff From wfs@math.utexas.edu Tue Oct 18 16:53:20 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02360; Tue, 18 Oct 94 16:53:20 CDT Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA21644; Tue, 18 Oct 94 15:21:00 CDT Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA25421; Tue, 18 Oct 94 15:18:00 -0500 Date: Tue, 18 Oct 94 15:17:53 -0500 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Tue, 18 Oct 94 15:17:53 -0500 Message-Id: <9410182017.AA06001@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA06001; Tue, 18 Oct 94 15:17:53 -0500 To: gcl@cli.com Cc: jeff@aiai.edinburgh.ac.uk In-Reply-To: Jeff Dalton's message of Tue, 18 Oct 94 16:42:47 BST <22533.9410181542@subnode.aiai.ed.ac.uk> Subject: Re: null-terminated strings? Reply-To: wfs@math.utexas.edu They are not null terminated. but the function object_to_string(x) takes a lisp string and returns a C string which is null terminated. In either the case the string is with fill pointer and fill pointer < length or in the case length mod 4 is not 0, the result will not require a copy. Otherwise it is copied. The functions like defentry automatically invoke this. Another thing to worry about is that if your string is relocatable, then it will move on gc, so be careful if you let C hang onto it past a garbage collection. From attardi@DI.UniPi.IT Wed Oct 19 05:54:06 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02423; Wed, 19 Oct 94 05:54:06 CDT Received: from apollo.di.unipi.it by cli.com (4.1/SMI-4.1) id AA23119; Wed, 19 Oct 94 04:47:31 CDT Received: from omega.di.unipi.it by apollo.di.unipi.it with SMTP (1.37.109.4/16.2) id AA08170; Wed, 19 Oct 94 10:46:24 +0100 Organization: Dipartimento di Informatica - Universita' di Pisa - Italy Received: by omega (5.0/SMI-4.1) id AA01232; Wed, 19 Oct 1994 10:45:45 --100 Date: Wed, 19 Oct 1994 10:45:45 --100 From: attardi@DI.UniPi.IT Message-Id: <9410190945.AA01232@omega> To: jeff@aiai.edinburgh.ac.uk Cc: gcl@cli.com In-Reply-To: <22533.9410181542@subnode.aiai.ed.ac.uk> (message from Jeff Dalton on Tue, 18 Oct 94 16:42:47 BST) Subject: Re: null-terminated strings? Content-Length: 568 >From <@festival.ed.ac.uk:jeff@aiai.edinburgh.ac.uk> Tue Oct 18 18:55:26 1994 Date: Tue, 18 Oct 94 16:42:47 BST From: Jeff Dalton Are strings in GCL null-terminated? It would make it much easier to pass them to C functions if they were. ECoLisp has null terminated strings, stored in non relocatable pages to avoid the problems mentioned by Schelter. Many changes were required to the original KCL code to support this feature. You may look at the ECoLisp code: ftp.di.unipi.it:/pub/lang/lisp/ecl-0.15.tar.gz -- Beppe From wfs@math.utexas.edu Wed Oct 19 13:57:54 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02633; Wed, 19 Oct 94 13:57:54 CDT Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA24249; Wed, 19 Oct 94 12:39:38 CDT Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA18540; Wed, 19 Oct 94 12:36:18 -0500 Date: Wed, 19 Oct 94 12:36:06 -0500 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Wed, 19 Oct 94 12:36:06 -0500 Message-Id: <9410191736.AA07089@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA07089; Wed, 19 Oct 94 12:36:06 -0500 To: gcl@cli.com Cc: jeff@aiai.edinburgh.ac.uk, attardi@DI.UniPi.IT In-Reply-To: attardi@DI.UniPi.IT's message of Wed, 19 Oct 1994 10:45:45 --100 <9410190945.AA01232@omega> Subject: Re: null-terminated strings? Reply-To: wfs@math.utexas.edu has null terminated strings, stored in non relocatable pages to avoid the problems mentioned by Schelter. Perhaps I was not sufficiently precise. If you are passing information frequently to C via a string, to avoid any consing you may reuse a string: (setq s (make-array (+ n 1) :element-type 'string-char :static t :fill-pointer 0)) if n is the longest string you are interested in. A convenient way to copy into a string with fill pointer is to use format.. (format s "~a" x) From hxl@epic.com Thu Oct 20 19:22:32 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03386; Thu, 20 Oct 94 19:22:32 CDT Received: from epic.com ([192.58.180.22]) by cli.com (4.1/SMI-4.1) id AA28579; Thu, 20 Oct 94 18:02:24 CDT Received: from weber.epic ([192.58.180.44]) by epic.com (4.1/JMA.3) id AA13483; Thu, 20 Oct 94 16:06:49 PDT Received: by weber.epic (4.1/SMI-4.1) id AA17430; Thu, 20 Oct 94 16:02:19 PDT Date: Thu, 20 Oct 94 16:02:19 PDT From: hxl@epic.com (Charlie Xiaoli Huang) Message-Id: <9410202302.AA17430@weber.epic> To: gcl@cli.com Subject: ftp.cli.com not responding? Reply-To: hxl@epic.com Hi, I'm having trouble with ftp.cli.com, which is supposedly the archive site for gcl related materials. It doesn't seem to allow anonymous ftp. Am I doing something wrong? hxl From boyer@cli.com Fri Oct 21 13:25:22 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03796; Fri, 21 Oct 94 13:25:22 CDT Received: from rita (rita.cli.com) by cli.com (4.1/SMI-4.1) id AA00759; Fri, 21 Oct 94 12:17:13 CDT From: boyer@cli.com (Robert S. Boyer) Received: by rita (4.1) id AA11522; Fri, 21 Oct 94 12:17:13 CDT Date: Fri, 21 Oct 94 12:17:13 CDT Message-Id: <9410211717.AA11522@rita> To: nqthm-users@cli, gcl@cli Subject: anonymous ftp access to ftp.cli.com is impossible for some In the past, some software such as GCL and Nqthm has been advertised as being available through anonymous ftp from ftp.cli.com. Yesterday I learned, however, that anonymous ftp to ftp.cli.com does not work from some Internet sites, namely from any site whose IP number cannot be resolved into a hostname. Until yesterday, people encountering such difficulties simply had their attempted connections broken, without explanation, immediately after supplying the "anonymous" login name. Now at least they should receive a message that suggests that they try to initiate the ftp connection from some machine whose hostname is published in some way that ftp.cli.com can detect. I regret that this restriction exists. It is certainly not something I favor. I apologize to those who have been disconnected without explanation. Bob From hxl@epic.com Mon Oct 24 17:33:02 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04495; Mon, 24 Oct 94 17:33:02 CDT Received: from epic.com ([192.58.180.22]) by cli.com (4.1/SMI-4.1) id AA06756; Mon, 24 Oct 94 15:43:44 CDT Received: from healy.epic ([192.58.180.73]) by epic.com (4.1/JMA.3) id AA29314; Mon, 24 Oct 94 13:49:19 PDT Received: by healy.epic (4.1/SMI-4.1) id AA29489; Mon, 24 Oct 94 13:43:55 PDT Date: Mon, 24 Oct 94 13:43:55 PDT From: hxl@epic.com (Charlie Xiaoli Huang) Message-Id: <9410242043.AA29489@healy.epic> To: gcl@cli.com Cc: wfs@math.utexas.edu Subject: compiler optimization --- what am i doing wrong? Reply-To: hxl@epic.com Hi, I'm trying to assess the efficiency of compiled code in gcl (1.1 beta.) I've got the following piece of code to do the hanoi tower problem: (defun hanoi (n) (labels ((move-them (n from to helper) (if (> n 1) (progn (move-them (- n 1) from helper to) ; (format t "~a --> ~a~%" from to) (move-them (- n 1) helper to from))))) (move-them n 'L 'C 'R))) When I compile and run this on a Sparc 10-50, the run times I got are: (hanoi 20) -> 3.7 sec. (hanoi 25) -> 151.56 sec. (hanoi 30) -> (killed after a hour) Whereas an almost identical program in scheme, when compiled and run in Scheme->C from DEC, yields: (hanoi 20) -> 0.3 sec. (hanoi 25) -> 13.50 sec. (hanoi 30) -> 447.20 sec. (By the way, I also coded it in C, which run about 3 times faster than Scheme->C, and about 20-30 times that of gcl.) I looked at the C code generated. Here's the guts of the c code from gcl: /* local function MOVE-THEM */ static L2(base0) register object *base0; { register object *base=vs_base; register object *sup=base+VM2; VC2 vs_check; {object V2; object V3; object V4; object V5; V2=(base[0]); V3=(base[1]); V4=(base[2]); V5=(base[3]); vs_top=sup; TTL:; if(!(number_compare((V2),small_fixnum(1))>0)){ goto T6;} base[4]= number_minus((V2),small_fixnum(1)); base[5]= (V3); base[6]= (V5); base[7]= (V4); vs_top=(vs_base=base+4)+4; L2(base0); vs_top=sup; V2= number_minus((V2),small_fixnum(1)); {object V6; V6= (V5); V5= (V3); V3= (V6);} goto TTL; return; T6:; T6:; base[4]= Cnil; vs_top=(vs_base=base+4)+1; return; } } [ Admittedly I could have added type information and other minor tuning. But my scheme code didn't have any of that anyway.] As much as I can tell, this isn't significantly different from what Scheme-> produces: TSCP hanoi_m2004( n2006, f2007, t2008, h2009 ) TSCP n2006, f2007, t2008, h2009; { TSCP X2, X1; L2042: if ( LTE( _S2CINT( n2006 ), _S2CINT( _TSCP( 4 ) ) ) ) goto L2043; X1 = _TSCP( DIFFERENCE( _S2CINT( n2006 ), _S2CINT( _TSCP( 4 ) ) ) ); hanoi_m2004( X1, f2007, h2009, t2008 ); X1 = _TSCP( DIFFERENCE( _S2CINT( n2006 ), _S2CINT( _TSCP( 4 ) ) ) ); X2 = h2009; h2009 = f2007; f2007 = X2; n2006 = X1; GOBACK( L2042 ); L2043: return( FALSEVALUE ); } Yet the run time difference is close to 10-fold. I wonder if anyone has looked at the cause of such differences. Since the generated code is almost the same, it would be a pity for gcl's run time to be so much slower than Scheme->C's. Is it caused by the fact that all parameters in gcl's functions are passed in arrays and therefore indirectly accessed, whereas in Scheme->C's code they are scalars? Regards, hxl From wfs@math.utexas.edu Mon Oct 24 18:20:58 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04540; Mon, 24 Oct 94 18:20:58 CDT Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA06990; Mon, 24 Oct 94 16:59:02 CDT Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA17779; Mon, 24 Oct 94 16:55:34 -0500 Date: Mon, 24 Oct 94 16:55:08 -0500 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Mon, 24 Oct 94 16:55:08 -0500 Message-Id: <9410242155.AA12954@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA12954; Mon, 24 Oct 94 16:55:08 -0500 To: hxl@epic.com, gcl@cli.com Cc: In-Reply-To: Charlie Xiaoli Huang's message of Mon, 24 Oct 94 13:43:55 PDT <9410242043.AA29489@healy.epic> Subject: Re: compiler optimization --- what am i doing wrong? Reply-To: wfs@math.utexas.edu Unfortunately labels constructs are currently slow in gcl [as you noted the argument passing is slow], and will be improved. However when I code this as a regular function, and add the information that n will be a fixnum, you can see that the gcl times are almost half the times you reported for scheme. I suspect the only difference between what you see in this and what hand written C would get, is that we do the recursive calls thru a pointer, to allow things like tracing and redefines to take place. ; hanoi N gcl scheme->c gcl (sun c) ; 20 .27 .39 ; 25 7.8 13.5 7.4 ; 30 251 442 I am using your figures for scheme-c Note that if I set the compiler to not do the link calls [ie no tracing possible], then the time is 6.4 for hanoi 25 using sun c compiler [free version]. Inserting file ~/foo.lisp ---Begin File ~/foo.lisp--- (proclaim '(ftype (function (fixnum t t t) t) hanoi-move)) (defun hanoi-move (n from to helper) (declare (fixnum n)) (if (> n 1) (progn (hanoi-move (the fixnum (- n 1)) from helper to) ; (format t "~a --> ~a~%" from to) (hanoi-move (the fixnum (- n 1)) helper to from)))) ;When I compile and run this on a Sparc 10 the run times I got are: ; ; hanoi N gcl scheme->c ; 20 .27 .39 ; 25 7.8 13.5 ; 30 251 442 ---End File ~/foo.lisp--- From attardi@DI.UniPi.IT Tue Oct 25 04:29:12 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04741; Tue, 25 Oct 94 04:29:12 CDT Received: from apollo.di.unipi.it by cli.com (4.1/SMI-4.1) id AA08146; Tue, 25 Oct 94 03:31:20 CDT Received: from omega.di.unipi.it by apollo.di.unipi.it with SMTP (1.37.109.4/16.2) id AA18759; Tue, 25 Oct 94 09:30:25 +0100 Organization: Dipartimento di Informatica - Universita' di Pisa - Italy Received: by omega (5.0/SMI-4.1) id AA04465; Tue, 25 Oct 1994 09:29:58 --100 Date: Tue, 25 Oct 1994 09:29:58 --100 From: attardi@DI.UniPi.IT Message-Id: <9410250829.AA04465@omega> To: hxl@epic.com Cc: gcl@cli.com In-Reply-To: <9410242043.AA29489@healy.epic> (hxl@epic.com) Subject: Re: compiler optimization --- what am i doing wrong? Content-Length: 1180 I have tried your example with ECoLisp, a Lisp->C compiler derived from (A)KCL which uses standard C calling conventions. This is the compiled code: /* function definition for HANOI */ static L1(int narg, object V1) { VT3 VLEX3 CLSR3 TTL: RETURN(LC2(4,(V1),VV[0],VV[1],VV[2]) /* MOVE-THEM */); } ;;; Emitting code for MOVE-THEM. /* local function MOVE-THEM */ static LC2(int narg, object V1, object V2, object V3, object V4) { VT4 VLEX4 CLSR4 {int V5; V5= fix(V1); TTL: if(!((V5)>(1))){ goto L2;} LC2(4,MAKE_FIXNUM((V5)-(1)),(V2),(V4),(V3))/* MOVE-THEM */; V5= (V5)-(1); {object V6; V6= (V4); V4= (V2); V2= (V6);} goto TTL; L2: VALUES(0) = Cnil; RETURN(1); } } by adding (declare (fixnum n)) in move-them. Even if this code could still be improved, the times are similar to those of Scheme->C: (hanoi 20) -> 0.233 sec. (hanoi 25) -> 13.717 sec. (hanoi 30) -> 486.817 sec. -- Beppe PS ECoLisp is available for ftp from ftp.di.unipi.it:/pub/lang/lisp. From loosemore-sandra@CS.YALE.EDU Wed Dec 7 13:01:09 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02781; Wed, 7 Dec 94 13:01:09 CST Received: from NEBULA.SYSTEMSZ.CS.YALE.EDU (SYSTEMSZ-GW.CS.YALE.EDU) by cli.com (4.1/SMI-4.1) id AA06057; Wed, 7 Dec 94 11:52:31 CST Received: from FUNCTOR.SYSTEMSZ.CS.YALE.EDU by NEBULA.SYSTEMSZ.CS.YALE.EDU via SMTP; Wed, 7 Dec 1994 12:54:22 -0500 Received: by FUNCTOR.SYSTEMSZ.CS.YALE.EDU (Sendmail-5.67b/res.client.cf-3.7) id AA03228; Wed, 7 Dec 1994 12:54:07 -0500 Date: Wed, 7 Dec 1994 12:54:07 -0500 From: loosemore-sandra@CS.YALE.EDU (Sandra Loosemore) Message-Id: <199412071754.AA03228@FUNCTOR.SYSTEMSZ.CS.YALE.EDU> To: gcl@cli.com Subject: problems building gcl-1.1 We're having troubles building gcl-1.1 here (on a sparc running SunOS 4.1.3). It's croaking with a zillion error messages trying to compile c/sfasl.c (and rel_sun4.c, which it includes). As far as I can figure out, it looks like this file isn't even intended to be compiled by itself (it says it's supposed to be included in unixfasl.c) and is missing some critical #includes. Any clue as to what's going on here? Is the problem in the makefile or the sun4-specific code files? -Sandra From smith@io.litton.com Fri Dec 9 14:20:40 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03352; Fri, 9 Dec 94 14:20:40 CST Received: from lcsdns1.litton.com by cli.com (4.1/SMI-4.1) id AA12827; Fri, 9 Dec 94 13:13:25 CST Received: from io.dsd.litton.com by lcsdns1.litton.com (4.1/SMI-4.1) id AA09032; Fri, 9 Dec 94 11:10:56 PST Received: by io.dsd.litton.com (Concentrix-2800 3.0/SMI-4.1) id AA12190; Fri, 9 Dec 94 11:05:36 PST Date: Fri, 9 Dec 94 11:05:36 PST From: Bill Smith Message-Id: <9412091905.AA12190@io.dsd.litton.com> To: gcl@cli.com Subject: Solaris 2.3 port We're trying to build gcl 1.1 on a Solaris 2.3 system using the Sparcworks C compiler, but we're running into lots of difficulty. Has anyone done this? Any advice? We get lots of compiler warnings (which we're ignoring for the time being), but finally get some errors with data types we can't resolve and which are making us suspicious that we haven't built our configuration files correctly. Thanks Bill Smith smith@io.dsd.litton.com (header address probably incorrect) From Chandan.Haldar@blr.sni.de Wed Dec 14 06:07:36 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04593; Wed, 14 Dec 94 06:07:36 CST Received: from mail.sni.de by cli.com (4.1/SMI-4.1) id AA23211; Wed, 14 Dec 94 05:02:17 CST Received: from blrsn01.blr.sni.de by mail.sni.de with SMTP (PP) id <17748-0@mail.sni.de>; Wed, 14 Dec 1994 11:59:30 +0100 Received: by blr.sni.de (4.1/SMI-4.1) id AA19894; Wed, 14 Dec 94 13:00:18+050 Date: Wed, 14 Dec 94 13:00:18+050 From: Chandan.Haldar@blr.sni.de (Dr. Chandan Haldar) Message-Id: <9412140800.AA19894@blr.sni.de> To: gcl@cli.com Subject: Japanese ... Apology for non-GCL query, but this is the most likely place I know where I could ask this! Does anybody know of a public domain Japanese-English dictionary (either Kana-English or Kanji-English) and/or a Japanese grammar spec in some machine-readable form? Where? Thanks. -------------------------------------------------------------------------- Dr. Chandan Haldar -------------------------------------------------------------------------- Siemens Information Systems Limited 29 Infantry Road, Bangalore 560 001, India. Phone: +91 80 559 1113 Fax: +91 80 559 1214 Internet: Chandan.Haldar@blr.sni.de -------------------------------------------------------------------------- From leverich@rand.org Fri Dec 16 22:29:01 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05066; Fri, 16 Dec 94 22:29:01 CST Received: from rand.org by cli.com (4.1/SMI-4.1) id AA01765; Fri, 16 Dec 94 21:38:00 CST Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.9/8.6.9) with ESMTP id TAA28867 for ; Fri, 16 Dec 1994 19:37:34 -0800 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.9/8.6.9) with ESMTP id TAA07114 for ; Fri, 16 Dec 1994 19:37:40 -0800 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.9/8.6.9) with SMTP id TAA17006 for ; Fri, 16 Dec 1994 19:37:39 -0800 Message-Id: <199412170337.TAA17006@atlantis.rand.org> To: gcl@cli.com Reply-To: Brian_Leverich@rand.org Subject: SGC on Linux Date: Fri, 16 Dec 94 19:37:37 PST From: Brian Leverich Before the RAND mafia reinvents the wheel, does anyone have the SGC working under Linux? The more recent kernels should have the mprotect functionality, I think, but GCL (at least GCL 1.0) didn't seem to like what the OS served up. BTW, with a little help from Scott Musman at MITRE, we've now got CLX, GARNET, and GNU ROSS all running together under Linux. Makes for a nice environment. We'll be putting the patched versions of everything up for ftp after a little cleaning up. Cheers, B. ----------- Dr. Brian Leverich Information Systems Scientist, The RAND Corporation leverich@rand.org From windley@lal.cs.byu.edu Mon Dec 19 13:31:36 1994 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05367; Mon, 19 Dec 94 13:31:36 CST Received: from leopard.cs.byu.edu by cli.com (4.1/SMI-4.1) id AA05481; Mon, 19 Dec 94 11:47:01 CST Received: from jaguar.cs.byu.edu by leopard.cs.byu.edu with SMTP (1.38.193.4/16.2) id AA26630; Mon, 19 Dec 1994 10:51:45 -0700 Received: from localhost by jaguar.cs.byu.edu (1.38.193.4/CS-Client) id AA13985; Mon, 19 Dec 1994 10:49:08 -0700 Message-Id: <9412191749.AA13985@jaguar.cs.byu.edu> To: gcl@cli.com Subject: intermitent error: The function NIL is undefined. Date: Mon, 19 Dec 1994 10:49:07 -0700 From: "Phil Windley" I am using GCL 1.0 on an HP7xx to run the HOL theorem proving system. I've been having a problem which after some debuggin appears to be something in LISP, not a problem with HOL. I get the following message: Error: The function NIL is undefined. Fast links are on: do (use-fast-links nil) for debugging Error signalled by FUN%6115%20. Backtrace: > funcall > lambda > cons > FUN%6115%20 evaluation failed lisp error This happens on some applications *some of the time*. In other words, sometimes the expression will work and ither times this message is generated. I cannot figure out a pattern as to why it works sometimes. The expressions are fairly large, but certainly not the largest I do. These same expressions worked without failure under AKCL 1.615. Does anyone know what could be causing this? --phil-- __________________________________________________________________________ Phillip J. Windley, Asst. Professor | windley@cs.byu.edu Laboratory for Applied Logic | Dept. of Computer Science, TMCB 3370 | Brigham Young University | Phone: 801.378.3722 Provo UT 84602-6576 | Fax: 801.378.7775 ------------------------------------------------------------------------ If you use WWW, I can be found here orhere. From jmacd@uclink.berkeley.edu Sun Jan 1 06:21:12 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA07382; Sun, 1 Jan 95 06:21:12 CST Received: from uclink3.berkeley.edu by cli.com (4.1/SMI-4.1) id AA29957; Sun, 1 Jan 95 05:32:46 CST Received: from uclink.berkeley.edu by uclink3.berkeley.edu (8.6.8/1.33(web)-OV2) id DAA07095; Sun, 1 Jan 1995 03:32:45 -0800 Received: by uclink.berkeley.edu (8.6.9/1.33(web)-OV4) id DAA24759; Sun, 1 Jan 1995 03:32:44 -0800 Date: Sun, 1 Jan 1995 03:32:44 -0800 From: jmacd@uclink.berkeley.edu (Joshua Peck Macdonald) Message-Id: <199501011132.DAA24759@uclink.berkeley.edu> To: gcl@cli.com Subject: xakcl question I'm not sure if this is an appropriate place to post a question regarding xakcl which I"ve (well sort of I guess) compiled for FreeBSD. It builds and loads and saves everything properly, but I can't seem to get anything done and I'm not sure why because I try to initialize a display with (setq *default-display* (XOpenDisplay (get-c-string ""))) as it suggests but get-c-string causes an error. It seems to be defined like this: (defCfun "object get_c_string(s) object s;" 0 " return(s->st.st_self);" ) (defentry get_c_string_2 (object) (object get_c_string)) (defun get-c-string (string) (get_c_string_2 (concatenate 'string string "^@"))) but it signals a fatal error in PRIN1. I'm interested in trying out thisinterface to X through gcl and would appreciate any help. I'm sorry if this is a stupid/inappropriate/misplaced question. thanks in advance, josh From leverich@rand.org Sun Jan 1 17:29:31 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA07455; Sun, 1 Jan 95 17:29:31 CST Received: from rand.org by cli.com (4.1/SMI-4.1) id AA00440; Sun, 1 Jan 95 16:30:17 CST Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.9/8.6.9) with ESMTP id OAA16604; Sun, 1 Jan 1995 14:30:04 -0800 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.9/8.6.9) with ESMTP id OAA04274; Sun, 1 Jan 1995 14:30:03 -0800 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.9/8.6.9) with SMTP id OAA08869; Sun, 1 Jan 1995 14:30:01 -0800 Message-Id: <199501012230.OAA08869@atlantis.rand.org> To: jmacd@uclink.berkeley.edu (Joshua Peck Macdonald) Cc: gcl@cli.com, Brian_Leverich@rand.org Reply-To: Brian_Leverich@rand.org Subject: Re: xakcl question In-Reply-To: Your message of Sun, 01 Jan 95 03:32:44 PST. <199501011132.DAA24759@uclink.berkeley.edu> Date: Sun, 01 Jan 95 14:29:59 PST From: Brian Leverich We use Linux and CLX rather than FreeBSD and xakcl. To get that stuff to run, we needed the following patch to the header file for the machine type: /* Linux version by Scott Musman (MITRE) and Mike Harm (USC) */ #define LISTEN_FOR_INPUT(fp) \ if((c=0,ioctl((fp)->_fileno, FIONREAD, &c),c<=0) \ && ((fp)->_IO_read_ptr >= (fp)->_IO_read_end)) \ return 0 The other thing we had to do was hack CLX to handle the endianness of the Intel architecture. Sorry if these are orthogonal to your problem - I'm still recovering from last night's partying and I only have 3 neurons confirmed firing. -B -- Your message was: (from "jmacd@uclink.berkeley.edu") I'm not sure if this is an appropriate place to post a question regarding xakcl which I"ve (well sort of I guess) compiled for FreeBSD. It builds and loads and saves everything properly, but I can't seem to get anything done and I'm not sure why because I try to initialize a display with (setq *default-display* (XOpenDisplay (get-c-string ""))) as it suggests but get-c-string causes an error. From hxl@epic.com Tue Jan 3 19:56:58 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA07850; Tue, 3 Jan 95 19:56:58 CST Received: from epic.com (gate.epic.com) by cli.com (4.1/SMI-4.1) id AA03927; Tue, 3 Jan 95 19:10:54 CST Received: from weber.epic (weber.epic.com) by epic.com (4.1/JMA.3) id AA21412; Tue, 3 Jan 95 17:18:04 PST Received: by weber.epic (4.1/SMI-4.1) id AA13538; Tue, 3 Jan 95 17:11:53 PST Date: Tue, 3 Jan 95 17:11:53 PST From: hxl@epic.com (Charlie Xiaoli Huang) Message-Id: <9501040111.AA13538@weber.epic> To: gcl@cli.com Subject: gprof doesn't work? (sun4, gcc 2.4.5, gcl 1.1) Reply-To: hxl@epic.com I can't seem to compile a profile-able version of gcl1.1. This is what I did: % ./add-defs sun4 % make ---> saved_gcl and everything was fine! % mkdir go % rm -f o/*.o % (cd go; ln ../o/makefile .) % make kcp ---> got as far as raw_gcl Here's the error message and a stack trace from the end of "make kcp": ../unixport/raw_gcl ../unixport/ < foo GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o Initializing ../lsp/evalmacros.o Initializing ../lsp/top.o Initializing ../lsp/module.o loading ../lsp/autoload.lsp >#<"COMPILER" package> COMPILER>#<"SYSTEM" package> SYSTEM>*COMMAND-ARGS* SYSTEM>#<"USER" package> >#<"LISP" package> LISP>#<"SLOOP" package> SLOOP>6 SLOOP>#<"USER" package> >Initializing ../lsp/predlib.o Error: Someone tried to RETURN-FROM a TAGBODY. Error signalled by TAGBODY. Broken at BREAK-LEVEL. Type :H for Help. SYSTEM>>Bye. Thanks, hxl From hxl@epic.com Tue Jan 3 19:57:48 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA07854; Tue, 3 Jan 95 19:57:48 CST Received: from epic.com (gate.epic.com) by cli.com (4.1/SMI-4.1) id AA03920; Tue, 3 Jan 95 19:06:18 CST Received: from weber.epic (weber.epic.com) by epic.com (4.1/JMA.3) id AA21399; Tue, 3 Jan 95 17:13:27 PST Received: by weber.epic (4.1/SMI-4.1) id AA13509; Tue, 3 Jan 95 17:07:17 PST Date: Tue, 3 Jan 95 17:07:17 PST From: hxl@epic.com (Charlie Xiaoli Huang) Message-Id: <9501040107.AA13509@weber.epic> To: gcl@cli.com Subject: FFI questions Reply-To: hxl@epic.com Hello, I'm trying to do some work with the FFI provided by gcl. I read the kcl report and the defCfun/defentry/clines stuff are farily clear. But there're a few things that are done but are "officially documented", which I'd like to seek some clarifications of, so that I can be sure my code will continue to work when I upgrade the version of my gcl: First, when I look at some of the code in the lsp and misc directory, I see access to things like (from misc/rusage.lsp): (clines "static mygetrusage(x,y) int x; object y;{return getrusage(x,y->str.str_self);}") which accesses the internals of an object, and assumes it can be implicitly used as a struct pointer. Is this supported and documented? If not, what is the proper way to achieve the same or similar effect? Second, the chapter on FFI in the kcl report includes things like: ('nil a) and ((cons a b) c) that can be embedded in a defCfun, but didn't explain the full syntax and semantics of them. Anywhere else I can look them up? In general, what are the C functions (or lisp-ish helper functions like the ones above) that one can use in defCfun to talk to the lisp side? I see there're things like make_simple_string and CMPmakefix_num. But would like to see a formal list and definitions if possible. Please email response to hxl@epic.com. Thanks, hxl From reading@io.litton.com Wed Jan 4 17:42:27 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA08190; Wed, 4 Jan 95 17:42:27 CST Received: from lcsdns1.litton.com by cli.com (4.1/SMI-4.1) id AA07178; Wed, 4 Jan 95 16:47:05 CST Received: from io.dsd.litton.com by lcsdns1.litton.com (4.1/SMI-4.1) id AA01076; Wed, 4 Jan 95 14:44:23 PST Received: by io.dsd.litton.com (Concentrix-2800 3.0/SMI-4.1) id AA05683; Wed, 4 Jan 95 14:47:17 PST Date: Wed, 4 Jan 95 14:47:17 PST From: John D. Reading Message-Id: <9501042247.AA05683@io.dsd.litton.com> To: gcl@cli.com Subject: Problems building PCL Sorry to bother you all, but I am not sure where else to ask. I have downloaded pcl-gcl-1.0 and tried to build it (I also have gcc version 2.6.3 and gcl version 1.1), and things proceeded fine right up until it core dumped. The last few lines of output follow: Compiling VECTOR... Compiling vector.lisp. End of Pass 1. ;; Note: Tail-recursive call of SYMBOL-OR-CONS-LESSP was replaced by iteration. ;; Note: Tail-recursive call of SYMBOL-OR-CONS-LESSP was replaced by iteration. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling vector.o. Loading binary of VECTOR... Compiling SLOTS-BOOT... Compiling slots-boot.lisp. ; (DEFMACRO ACCESSOR-SLOT-BOUNDP ...) is being compiled. ;; Warning: The variable SYM is not used. Segmentation Fault - core dumped *** Error code 139 make: Fatal error: Command failed for target `compile' (on host callisto) I really don't have the faintest idea of what to do. Can anybody give me any help? Thanks, John -- John D. Reading | Where there is no solution, there is no problem. reading@io.dsd.litton.com | -- John G. Price III (818) 597-5152 | From horrocks@cs.man.ac.uk Fri Jan 13 15:11:59 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA09998; Fri, 13 Jan 95 15:11:59 CST Received: from m1.cs.man.ac.uk by cli.com (4.1/SMI-4.1) id AA28740; Fri, 13 Jan 95 13:55:58 CST Received: from r9.cs.man.ac.uk by m1.cs.man.ac.uk (4.1/SMI-4.1:AL5l) id AA21218; Fri, 13 Jan 95 00:55:33 GMT From: Ian Horrocks Message-Id: <9501130055.AA06816@r9.cs.man.ac.uk> Subject: macroexpand/evaluate To: gcl@cli.com Date: Fri, 13 Jan 1995 00:55:31 +0000 (GMT) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 361 I am having problems with the GCL compiler due to the way it evaluates macro expansions at compile time. I read in the documentation that this behavior can be altered using one of the system flags but it didn't say which one - can anyone assist? Talking of documentation is fuller documentation available from any source? Ian Horrocks horrocks@cs.man.ac.uk From hxl@epic.com Tue Jan 17 19:28:41 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA10673; Tue, 17 Jan 95 19:28:41 CST Received: from epic.com (gate.epic.com) by cli.com (4.1/SMI-4.1) id AA03839; Tue, 17 Jan 95 18:03:24 CST Received: from weber.epic (weber.epic.com) by epic.com (4.1/JMA.3) id AA20565; Tue, 17 Jan 95 16:10:47 PST Received: by weber.epic (4.1/SMI-4.1) id AA09898; Tue, 17 Jan 95 16:04:26 PST Date: Tue, 17 Jan 95 16:04:26 PST From: hxl@epic.com (Charlie Xiaoli Huang) Message-Id: <9501180004.AA09898@weber.epic> To: gcl@cli.com, wfs@math.utexas.edu Subject: gcl: is there a better way to do this? Reply-To: hxl@epic.com I need to interface an application with GCL. One of the API's provided by the application is something like int foo(int argc, char **argv) (the API was originally written with TCL in mind...) How do I accomplish an FFI in GCL with such an interface? This is what I'm doing now, but I'd like a more elegant solution: ----- (clines "static int argc; char *argv[512];" "static void reset_args() { argc = 0;}" "static int give_me_argv() { return (int)argv;}" "static int give_me_argc() { return argc;}" "static void add_args(s) char *s; " " { argv[argc++] = s; argv[argc] = 0;}") (defentry reset-args () (void "reset_args")) (defentry add-args (string) (void "add_args")) (defentry give-me-argv () (int "give_me_argv")) (defentry give-me-argc () (int "give_me_argc")) (defentry foo1 (int int) (int "foo")) (defun foo (arg &rest others) (reset-args) (add-args arg) (dolist (x others) (add-args x)) (foo1 (give-me-argc) (give-me-argv))) ---- Thanks in advance, hxl From anders.vinjar@notam.uio.no Sat Jan 21 13:19:06 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11386; Sat, 21 Jan 95 13:19:06 CST Received: from pat.uio.no by cli.com (4.1/SMI-4.1) id AA14214; Sat, 21 Jan 95 12:20:57 CST Received: from ulrik.uio.no by pat.uio.no with local-SMTP (PP) id <19220-0@pat.uio.no>; Sat, 21 Jan 1995 18:00:52 +0100 Received: by alban.notam.uio.no ; Sat, 21 Jan 95 17:59:28 +0100 Date: Sat, 21 Jan 95 17:59:28 +0100 From: anders.vinjar@notam.uio.no Message-Id: <9501211659.AA02885@alban.notam.uio.no> To: gcl@cli.com Subject: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? Hello. I get som strange results while trying to build GCL-1.1 on the above system. Can anyone tell me what the following error message means? ....... (cut away) ../unixport/raw_gcl ../unixport/ < foo GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o This file was dumped with FASD version ~a not ~a. <------!! Lisp initialization failed. make command rm -f xbin/gcl ; echo "#!/bin/sh" > xbin/gcl; \ echo "/hf/notam/u1/andersvi/lisp/gcl-1.1/unixport/saved_gcl -dir /hf/notam/u1/andersvi/lisp/gcl-1.1/unixport/ \$@ " \ >> xbin/gcl chmod 755 xbin/gcl (cd cmpnew ; make collectfn.o) ../unixport/saved_gcl ../unixport/ -compile collectfn ../unixport/saved_gcl: Not found *** Error code 1 (bu21) make: fatal error. *** Error code 1 (bu21) make: fatal error. my-prompt> To get any way on the Indy i had to change the following in "c/save_sgi.c": /* extern char *sbrk(); changed to the following line:*/ extern void *sbrk(); I also had to change a definition of _iob() to __iob() (with two underscores) in "c/save-sgi4.c". Anders Vinjar From cheimets@blynken.harvard.edu Sun Jan 22 09:35:47 1995 Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11485; Sun, 22 Jan 95 09:35:47 CST Received: from cfa.harvard.edu by cli.com (4.1/SMI-4.1) id AA15114; Sun, 22 Jan 95 08:39:56 CST Return-Path: Received: by cfa.harvard.edu; Sun, 22 Jan 95 09:39:50 -0500 Received: by blynken.harvard.edu (931110.SGI/920502.SGI) for @cfa.harvard.edu:gcl@cli.com id AA02035; Sun, 22 Jan 95 09:40:25 -0500 Date: Sun, 22 Jan 95 09:40:25 -0500 From: cheimets@blynken.harvard.edu (Peter Cheimets) Message-Id: <9501221440.AA02035@blynken.harvard.edu> To: anders.vinjar@notam.uio.no Cc: gcl@cli.com In-Reply-To: anders.vinjar@notam.uio.no's message of Sat, 21 Jan 95 17:59:28 +0100 <9501211659.AA02885@alban.notam.uio.no> Subject: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? I had to make the exact same changes, and got the exact same results. If you get any further please let me know. When I got a hold of Schelter this is the reply I got: >The point is that the format of the .o files may well have changed. >Try to determine if when you come to beginning of the read_fasd >that you are indeed positioned at what was the end of the .o file. >[put some printf("(%d)",ftell(fp)); >or some such in..] I am not a C kinda guy, so I could do nothing with this. Peter Cheimets Smithsonian Astrophysical Observatory ,__o- Smithsonian Institution _-\_<, 60 Garden St. (*)/'(*)- Cambridge, MA 01238 cheimets@cfa.harvard.edu (617)495-7384 X134 (voice) (617)495-7098 (FAX) From donhan@engin.umich.edu Mon Jan 23 13:59:00 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11704; Mon, 23 Jan 95 13:59:00 CST Received: from maize.engin.umich.edu by cli.com (4.1/SMI-4.1) id AA16936; Mon, 23 Jan 95 12:53:44 CST Received: (donhan@localhost) by maize.engin.umich.edu (8.6.8/8.6.4) id NAA08290; Mon, 23 Jan 1995 13:53:36 -0500 Date: Mon, 23 Jan 1995 13:53:36 -0500 From: Dongsuk Han Message-Id: <199501231853.NAA08290@maize.engin.umich.edu> To: anders.vinjar@notam.uio.no Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? Cc: gcl@cli.com Hi, I stopped building GCL-1.1 on IRIX 5.3 exactly at the same place. Please let me know if you hear from anyone on the net. ps) Are we getting enough attentions yet? ================================================================== Dongsuk "Don" Han Phone: (313) 764-2410 Transportation Research Institute Fax: (313) 936-1068 University of Michigan E-mail: donhan@umich.edu 2901 Baxter Road Ann Arbor, MI 48109-2150 From pw@snoopy.mv.com Mon Jan 23 15:58:50 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11728; Mon, 23 Jan 95 15:58:50 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA17319; Mon, 23 Jan 95 14:40:02 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id PAA12941; Mon, 23 Jan 1995 15:39:20 -0500 Date: Mon, 23 Jan 1995 15:39:20 -0500 From: "Paul F. Werkowski" Message-Id: <199501232039.PAA12941@snoopy.mv.com> To: donhan@engin.umich.edu Cc: anders.vinjar@notam.uio.no, gcl@cli.com In-Reply-To: <199501231853.NAA08290@maize.engin.umich.edu> (message from Dongsuk Han on Mon, 23 Jan 1995 13:53:36 -0500) Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? >>>>> "Dongsuk" == Dongsuk Han writes: Dongsuk> Hi, I stopped building GCL-1.1 on IRIX 5.3 exactly at the Dongsuk> same place. Please let me know if you hear from anyone Dongsuk> on the net. Dongsuk> ps) Are we getting enough attentions yet? Some of us learned a lot more that we wanted to about the innards of unix and how they vary with different vendors implementations by wanting to run AKCL/GCL on some new box. At one time I also was faced with total ignorance of "FASD" and had to do a lot of debugging in totally foreign territory at a time when I just wanted to deal in Lisp. If I had an INDY box and wanted to run GCL I would have several options, a) Appeal to Internet readers to see if someone had already blazed that trail. (You've sort of done that already) b) Try and debug the problem myself - and send any changes to the maintainer for inclusion in the next version. c) Wait for the next release and hope someone did one of the above. d) Buy a commercial Lisp implementation. e) Buy hardware and an OS that is known to be compatible with GCL. "Anything for free has no guarantee" Paul From donhan@engin.umich.edu Mon Jan 23 17:33:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11774; Mon, 23 Jan 95 17:33:56 CST Received: from maize.engin.umich.edu by cli.com (4.1/SMI-4.1) id AA17641; Mon, 23 Jan 95 16:19:53 CST Received: (donhan@localhost) by maize.engin.umich.edu (8.6.8/8.6.4) id RAA12223; Mon, 23 Jan 1995 17:19:32 -0500 Date: Mon, 23 Jan 1995 17:19:32 -0500 From: Dongsuk Han Message-Id: <199501232219.RAA12223@maize.engin.umich.edu> To: pw@snoopy.mv.com Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? Cc: anders.vinjar@notam.uio.no, gcl@cli.com >> Some of us learned a lot more that we wanted to about the >> innards of unix and how they vary with different vendors >> implementations by wanting to run AKCL/GCL on some new box. >> At one time I also was faced with total ignorance of "FASD" >> and had to do a lot of debugging in totally foreign territory >> at a time when I just wanted to deal in Lisp. If I had an >> INDY box and wanted to run GCL I would have several options, >> >> a) Appeal to Internet readers to see if someone had already >> blazed that trail. (You've sort of done that already) I sent a message to gcl mailing group - to see if anyone had gone through this and solved the problem and - to let the maintainer know that not only one or two are interested in porting GCL on SGI machines. I was succesful last time when I posted similar message about GCL 1.0 on IRIX 4.0.5. A few of kind netters helped me by sending me E-mails. >> b) Try and debug the problem myself - and send any changes to >> the maintainer for inclusion in the next version. I spent several hours until I hit the error shown on anders.vinjar's message. I admit that I have not gone into a real "debugging" yet but I am pretty sure that I've reached the limit of my (LISP + C) knowledge. I spent more than a week debugging version 1.0 on IRIX 4.0.5 without any clue on how things are working. I haven't checked if version 1.1 include the fixes for IRIX 4.0.5 yet. If not, I will send the patches to the maintainer. >> c) Wait for the next release and hope someone did one of the above. To make the hope bigger, I am telling the people who know more than I do about GCL (and who love GCL more than I do) that GCL is not easy to port on SGI machines. I've seen many messages like mine from the archive of this mailing list but porting AKCL/KCL/GCL on SGI has not been very successful for a long time. >> d) Buy a commercial Lisp implementation. Do you have any suggestion? >> e) Buy hardware and an OS that is known to be compatible with GCL. I compiled AKCL/GCL on Solaris, SUN OS, HP UX and IRIX 4.0.5. Now I have a reason to make GCL working on SGI. If I remember correctly, SGI is in the machine list that GCL is "supposed" to run on. I am not asking a question of porting GCL on Apollo DN3XX. >> "Anything for free has no guarantee" Thank you for your tips, and I am sorry if I bothered you with message that you did not want to read. >> Paul -- Don -- PS) I am truely sorry for sending rather personal message to this group. I just wish Paul did not CC to the whole group :( From pw@snoopy.mv.com Mon Jan 23 17:54:12 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11779; Mon, 23 Jan 95 17:54:12 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA17705; Mon, 23 Jan 95 16:44:39 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id RAA16274; Mon, 23 Jan 1995 17:44:16 -0500 Date: Mon, 23 Jan 1995 17:44:16 -0500 From: "Paul F. Werkowski" Message-Id: <199501232244.RAA16274@snoopy.mv.com> To: gcl@cli.com In-Reply-To: <199501232219.RAA12223@maize.engin.umich.edu> (message from Dongsuk Han on Mon, 23 Jan 1995 17:19:32 -0500) Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? >>>>> "Dongsuk" == Dongsuk Han writes: Dongsuk> I spent several hours until I hit the error shown on Dongsuk> anders.vinjar's message. I admit that I have not gone Dongsuk> into a real "debugging" yet but I am pretty sure that Dongsuk> I've reached the limit of my (LISP + C) knowledge. I Dongsuk> spent more than a week debugging version 1.0 on IRIX Dongsuk> 4.0.5 without any clue on how things are working. A week or two, or maybe a month of Sundays is about right for this one. >>> d) Buy a commercial Lisp implementation. Dongsuk> Do you have any suggestion? Nope, since my personal budget prohibits it and my employer can't afford it, I find GCL to be superb! I just have to be prepared for new debugging challenges when OS things change. I use FreeBSD (freebieSD). The price is right and the value per buck is infinitely high. >>> e) Buy hardware and an OS that is known to be compatible with >>> GCL. >>> "Anything for free has no guarantee" Dongsuk> Thank you for your tips, and I am sorry if I bothered you Dongsuk> with message that you did not want to read. Not bothered at all, and I apologize if I offended. I was responding to the apparent complaint of not enough attention. Dongsuk> PS) I am truely sorry for sending rather personal message Dongsuk> to this group. I just wish Paul did not CC to the whole Dongsuk> group :( Again, sorry to offend. Paul From attardi@DI.UniPi.IT Tue Jan 24 05:34:05 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11854; Tue, 24 Jan 95 05:34:05 CST Received: from apollo.di.unipi.it by cli.com (4.1/SMI-4.1) id AA18762; Tue, 24 Jan 95 04:15:10 CST Received: from omega.di.unipi.it by apollo.di.unipi.it with SMTP (1.37.109.4/16.2) id AA07563; Tue, 24 Jan 95 11:09:23 +0100 Organization: Dipartimento di Informatica - Universita' di Pisa - Italy Received: by omega (5.0/SMI-4.1) id AA00620; Tue, 24 Jan 1995 11:09:06 --100 Date: Tue, 24 Jan 1995 11:09:06 --100 From: attardi@DI.UniPi.IT Message-Id: <9501241009.AA00620@omega> To: donhan@engin.umich.edu Cc: pw@snoopy.mv.com, anders.vinjar@notam.uio.no, gcl@cli.com In-Reply-To: <199501232219.RAA12223@maize.engin.umich.edu> (message from Dongsuk Han on Mon, 23 Jan 1995 17:19:32 -0500) Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? Content-Length: 487 I am making available ECoLisp version 0.17, which runs on SGI with IRIX 4.0.5H, for anonymous ftp from: - ftp.icsi.berkeley.edu [128.32.201.7], directory /pub/ai/ecl - ftp.di.unipi.it [131.114.4.36], directory /pub/lang/lisp ECoLisp is a derivative of KCL which incorporates many of the changes in AKCL. To install it, after unpacking it, just run: configure This will create a directory sgi, then cd sgi make To see configuration options, issue configure --help. -- Beppe From anders.vinjar@notam.uio.no Tue Jan 24 11:17:32 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11893; Tue, 24 Jan 95 11:17:32 CST Received: from pat.uio.no ([129.240.2.50]) by cli.com (4.1/SMI-4.1) id AA19095; Tue, 24 Jan 95 08:51:28 CST Received: from ulrik.uio.no by pat.uio.no with local-SMTP (PP) id <03078-0@pat.uio.no>; Tue, 24 Jan 1995 15:42:59 +0100 Received: by alban.notam.uio.no ; Tue, 24 Jan 95 15:41:26 +0100 Date: Tue, 24 Jan 95 15:41:26 +0100 From: anders.vinjar@notam.uio.no Message-Id: <9501241441.AA15734@alban.notam.uio.no> To: cheimets@blynken.harvard.edu Cc: anders.vinjar@notam.uio.no, gcl@cli.com In-Reply-To: <9501221440.AA02035@blynken.harvard.edu> (cheimets@blynken.harvard.edu) Subject: Re: SGI-Indy (IRIX 5.2) and GCL-1.1: What is "FASD"? I had to make the exact same changes, and got the exact same results. If you get any further please let me know. I will. Thanks for a decent reply. -anders From eblood@winky.reno.nv.us Wed Jan 25 00:44:05 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA11993; Wed, 25 Jan 95 00:44:05 CST Received: from heather.greatbasin.com by cli.com (4.1/SMI-4.1) id AA21767; Tue, 24 Jan 95 23:48:28 CST Received: from winky.reno.nv.us (winky.reno.nv.us [140.174.194.250]) by heather.greatbasin.com (8.6.9/8.6.5) with SMTP id VAA09583; Tue, 24 Jan 1995 21:32:05 -0800 Message-Id: <199501250532.VAA09583@heather.greatbasin.com> X-Authentication-Warning: heather.greatbasin.com: Host winky.reno.nv.us didn't use HELO protocol To: gcl@cli.com Cc: eblood@winky.reno.nv.us Subject: Porting to FreeBSD 2.0 Date: Mon, 23 Jan 1995 14:07:41 -0800 From: Eric Blood I'm getting an unreferenced symbol when compiling for 386-bsd target. It's looking for _sigstack. I tracked down the module to main.o, and then checked main.c. I'm not quite sure what the sigstack code is supposed to be doing. The documentation on my side says that sigstack has been changed to sigaltstack. So, I tried to hack (hack being the keyword) sigstack into a sigaltstack. It compiles a little bit farther but fails with: Script started on Mon Jan 23 14:04:52 1995 $ make (cd bin; make all) (cd mp ; make all) make all1 "MPFILES=./mpi-386.o ./libmport.a" rm -f o/cmpinclude.h ; ln h/cmpinclude.h o (cd o; make all) (cd lsp; make all) (cd cmpnew; make all) (cd unixport; make saved_gcl) `saved_gcl' is up to date. make command rm -f xbin/gcl ; echo "#!/bin/sh" > xbin/gcl; echo "/usr/users/eblood/gcl-1.1/unixport/saved_gcl -dir /usr/users/eblood/gcl-1.1/unixport/ \$@ " >> xbin/gcl chmod 755 xbin/gcl (cd cmpnew ; make collectfn.o) ../unixport/saved_gcl ../unixport/ -compile collectfn ̆€P: not found ../unixport/saved_gcl: 10: Syntax error: ")" unexpected *** Error code 2 Stop. *** Error code 1 Stop. Tell me what I should do next to help port. Thanks. EVB From eblood@cs.unr.edu Wed Jan 25 02:45:21 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12001; Wed, 25 Jan 95 02:45:21 CST Received: from echo.cs.unr.edu by cli.com (4.1/SMI-4.1) id AA21954; Wed, 25 Jan 95 01:45:41 CST Received: from localhost (localhost [127.0.0.1]) by echo.cs.unr.edu (8.6.9/8.6.9) with SMTP id WAA29270 for gcl@cli.com; Tue, 24 Jan 1995 22:54:52 -0800 From: Eric V Blood Message-Id: <199501250654.WAA29270@echo.cs.unr.edu> X-Authentication-Warning: echo.cs.unr.edu: Host localhost didn't use HELO protocol To: gcl@cli.com Subject: Porting to FreeBSD 2.0 Date: Tue, 24 Jan 95 22:54:46 -0800 X-Mts: smtp I'm getting an unreferenced symbol when compiling for 386-bsd target. It's looking for _sigstack. I tracked down the module to main.o, and then checked main.c. I'm not quite sure what the sigstack code is supposed to be doing. The documentation on my side says that sigstack has been changed to sigaltstack. So, I tried to hack (hack being the keyword) sigstack into a sigaltstack. It compiles a little bit farther but fails with: Script started on Mon Jan 23 14:04:52 1995 $ make (cd bin; make all) (cd mp ; make all) make all1 "MPFILES=./mpi-386.o ./libmport.a" rm -f o/cmpinclude.h ; ln h/cmpinclude.h o (cd o; make all) (cd lsp; make all) (cd cmpnew; make all) (cd unixport; make saved_gcl) `saved_gcl' is up to date. make command rm -f xbin/gcl ; echo "#!/bin/sh" > xbin/gcl; echo "/usr/users/eblood/gcl-1.1/unixport/saved_gcl -dir /usr/users/eblood/gcl-1.1/unixport/ \$@ " >> xbin/gcl chmod 755 xbin/gcl (cd cmpnew ; make collectfn.o) ../unixport/saved_gcl ../unixport/ -compile collectfn ̆€P: not found ../unixport/saved_gcl: 10: Syntax error: ")" unexpected *** Error code 2 Stop. *** Error code 1 Stop. Tell me what I should do next to help port. Thanks. EVB Eric V. Blood eblood@cs.unr.edu From nils@guru.ims.uni-stuttgart.de Wed Jan 25 09:12:13 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12047; Wed, 25 Jan 95 09:12:13 CST Received: from guru.ims.uni-stuttgart.de by cli.com (4.1/SMI-4.1) id AA22424; Wed, 25 Jan 95 08:06:34 CST Received: by guru.ims.uni-stuttgart.de (Smail3.1.28.1 #16) id m0rX8Q9-0003UwC; Wed, 25 Jan 95 15:10 MET Message-Id: Date: Wed, 25 Jan 95 15:10 MET From: nils@guru.ims.uni-stuttgart.de (Cornelis van der Laan) To: eblood@cs.unr.edu, gcl@cli.com Subject: Re: Porting to FreeBSD 2.0 You have successfully built a saved_kcl which isn't executable (wrong header or so...). Here is how to dump it correctly as a static demand paged image: Use the appended h/386-bsd.h, get the unexec.c code from the latest emacs distribution (I think since 19.24 the code works), install it as c/unexfreebsd.c and recompile the whole stuff. Note though, that under FreeBSD 1.1.5 the gcl-1.x dists don't work for me, they crash with bus errors when using pcl and/or clx. akcl-1.624 works without problems, including the sgc code (ca. 25% faster garbage collection). Feel free to contact me for more info and, if you succeed in building on FreeBSD 2.0, would you please tell me how you did it? I'm still to upgrade but I'm waiting for the 2.1 cd... Nils h/386-bsd.h: = = = = = = = = = = = = = = = = = = = = = = = = #include "bsd.h" #define ADDITIONAL_FEATURES \ ADD_FEATURE("386BSD");\ ADD_FEATURE("CLX-LITTLE-ENDIAN"); #define I386 /* ?? this is apparently not used anywhere */ #define IEEEFLOAT #define HAVE_YP_UNBIND #define HAVE_XDR /* put xdr_float into libc first... */ #define USE_ATT_TIME /* begin listen for input */ #undef LISTEN_FOR_INPUT /* default in bsd.h is loser in 386bsd */ #if 1 /* Required for CLX to work correctly */ #if defined IN_FILE #include #include #include #include #endif #define LISTEN_FOR_INPUT(fp) \ {\ int fd = (fp)->_file;\ static struct timeval tv = {0,0};\ static fd_set rmask; FD_ZERO(&rmask); FD_SET(fd,&rmask);\ select(fd+1,&rmask,NULL,NULL,&tv);\ return (FD_ISSET(fd,&rmask));\ } #endif /* end listen for input */ /* we dont need to worry about zeroing fp->_base , to prevent */ #define FCLOSE_SETBUF_OK #undef LD_COMMAND #define LD_COMMAND(command,main,start,input,ldarg,output) \ sprintf(command, "ld -N -x -A %s -T %x %s %s -o %s", \ main,start,input,ldarg,output) #define DATA_BEGIN (char *)N_DATADDR(header) #define A_TEXT_OFFSET(x) (sizeof (struct exec)) #define A_TEXT_SEEK(hdr) (N_TXTOFF(hdr) + A_TEXT_OFFSET(hdr)) #define TEXT_START 0 #define DATA_START 0 #define UNIXSAVE "unexfreebsd.c" /* from emacs distribution */ #define RELOC_FILE "rel_sun3.c" /* for SFASL - enabled in bsd.h */ #define LITTLE_ENDIAN /* also in */ #define HZ 100 /* in FreeBSD 1.1.5 - check for 2.0 */ /* this for GC */ /* #define PAGEWIDTH 12 /* i386 sees 4096 byte pages */ /* try out the gnu malloc */ #define GNU_MALLOC /* works if PAGEWIDTH==11 */ #define GNUMALLOC #define INSTALL_SEGMENTATION_CATCHER \ (void) signal(SIGSEGV,segmentation_catcher); \ (void) signal(SIGBUS,segmentation_catcher) /* Begin for cmpinclude */ /* yes we have alloca */ #define HAVE_ALLOCA #define WANT_SGC #ifdef WANT_SGC /* begin defines for SGC */ /* SGC is a performance winner for large applications as it doesn't run the entire image through the pager during collection. SGC requires the 'mprotect' function. Need Jeffrey Hsu's kernel patch for signal handlers. Should be in FreeBSD versions later than 1.0.2. Also need to add #include before #include in c/sgc.c Also - the above handler conflicts with use of '(un)catch-bad-signals' You may want to modify unixint.c to account for SGC use. */ #define SGC #define SIGPROTV SIGBUS #endif /* end of SGC mods */ /* _setjmp and _longjmp exist on bsd and are more efficient and handle the C stack which is all we need. [I think!] */ #define setjmp _setjmp #define longjmp _longjmp = = = = = = = = = = = = = = = = = = = = = = = = From pw@snoopy.mv.com Wed Jan 25 15:52:26 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12131; Wed, 25 Jan 95 15:52:26 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA23863; Wed, 25 Jan 95 14:45:21 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id PAA00499; Wed, 25 Jan 1995 15:44:58 -0500 Date: Wed, 25 Jan 1995 15:44:58 -0500 From: "Paul F. Werkowski" Message-Id: <199501252044.PAA00499@snoopy.mv.com> To: gcl@cli.com In-Reply-To: <199501250654.WAA29270@echo.cs.unr.edu> (message from Eric V Blood on Tue, 24 Jan 95 22:54:46 -0800) Subject: Re: Porting to FreeBSD 2.0 Here is the FreeBSD.h/ that work for me. You also need unexfreebsd.c from the emacs kit. ----------- FreeBSD.h ----------- /* * FreeBSD.h for akcl-1-624 akcl-1-625 ... * * Hacked Jan/94 by Werkowski (pw@snoopy.mv.com)for FreeBSD 2.0R * */ #include "bsd.h" #undef LD_COMMAND #define LD_COMMAND(command,main,start,input,ldarg,output) \ sprintf(command, "ld -dc -Bstatic -N -x -A %s -T %x %s %s -o %s", \ main,start,input,ldarg,output) #define ADDITIONAL_FEATURES \ ADD_FEATURE("386BSD");\ ADD_FEATURE("FreeBSD");\ ADD_FEATURE("CLX-LITTLE-ENDIAN"); #define IEEEFLOAT #undef HAVE_XDR #define USE_ATT_TIME /* begin listen for input */ #undef LISTEN_FOR_INPUT /* default in bsd.h is loser in 386bsd */ #if 1 /* Required for CLX to work correctly */ #if defined IN_FILE #include #include #include #include #endif #define LISTEN_FOR_INPUT(fp) \ {\ int fd = (fp)->_file;\ static struct timeval tv = {0,0};\ static fd_set rmask; FD_ZERO(&rmask); FD_SET(fd,&rmask);\ select(fd+1,&rmask,NULL,NULL,&tv);\ return (FD_ISSET(fd,&rmask));\ } #endif /* end listen for input */ #ifdef IN_GBC #include #endif /* we dont need to worry about zeroing fp->_base , to prevent */ #define FCLOSE_SETBUF_OK #define DATA_BEGIN (char *)N_DATADDR(header); /* * need to ensure following is appended to any imported unex? file *#ifdef UNIXSAVE *#include "save.c" *#endif */ /* /pfw: this from freebsd 1.1 */ #define UNIXSAVE "unexfreebsd.c" #define RELOC_FILE "rel_sun3.c" /* for SFASL - enabled in bsd.h */ /* this for GC */ #define PAGEWIDTH 12 /* i386 sees 4096 byte pages */ #undef HAVE_SIGVEC #undef HAVE_SIGACTION #define SETUP_SIG_STACK \ { static struct sigaltstack estack ; \ estack.ss_size = SIGSTKSZ ; \ estack.ss_flags = 0; \ if (sigaltstack(&estack, 0) < 0) \ perror("sigaltstack");} \ #define INSTALL_SEGMENTATION_CATCHER \ (void) signal(SIGSEGV,segmentation_catcher); \ (void) signal(SIGBUS,segmentation_catcher) #define WANT_SGC #ifdef WANT_SGC /* begin defines for SGC */ #define SGC #define SIGPROTV SIGBUS #endif /* end of SGC mods */ /* Begin for cmpinclude */ #ifndef __GNUC__ #define HAVE_ALLOCA #include #endif /* If can mprotect pages and so selective gc will work */ #define SGC /* End for cmpinclude */ From pw@snoopy.mv.com Wed Jan 25 17:34:59 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12155; Wed, 25 Jan 95 17:34:59 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA24329; Wed, 25 Jan 95 16:36:49 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id RAA00379; Wed, 25 Jan 1995 17:36:26 -0500 Date: Wed, 25 Jan 1995 17:36:26 -0500 From: "Paul F. Werkowski" Message-Id: <199501252236.RAA00379@snoopy.mv.com> To: gcl@cli.com Subject: Re: GCL on FreeBSD 2.0 There is one critical thing for the subject combo. The 'libc/vprintf.c' code found in the 2.0 kit is broken such that GCL will not properly compile certain files such as found in PCL. Vprintf.c has been fixed in the FreeBSD-current distribution and you should pull the new version from FreeBSD.cdrom.com:/pub/FreeBSD/src/lib/libc/stdio/vfprintf.c. If you need this file and can't get it, please contact me directly and I will send you a patch. Paul From starnet!bass!lakin@apple.com Wed Jan 25 18:03:55 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12160; Wed, 25 Jan 95 18:03:55 CST Received: from apple.com by cli.com (4.1/SMI-4.1) id AA24471; Wed, 25 Jan 95 17:14:07 CST Received: by apple.com (5.61/8-Oct-1993-eef) id AA16265; Wed, 25 Jan 95 14:50:04 -0800 Received: from bass by StarConn.com with uucp (Smail3.1.28.1 #1) id m0rXF1O-00024QC; Wed, 25 Jan 95 13:12 PST Received: by pgc.com (4.1/SMI-4.1) id AA03227; Wed, 25 Jan 95 13:21:00 PST Date: Wed, 25 Jan 95 13:21:00 PST From: starnet!bass!lakin@apple.com (Fred Lakin) Message-Id: <9501252121.AA03227@pgc.com> To: gcl@cli.com Subject: speed of CLX under GCL Reply-To: lakin@pgc.com About 4 years ago I was using AKCL and CLX for a large graphics program. The time required to draw stuff on the screen [text and lines, not pixmaps] was quite slow. A friend convinced me to do a test with one of the commerical lisps, and without change to the code I saw a speedup of 10X to 20X. Now I would like to try GCL/AKCL again. Has CLX speed been improved? [I believe some of the slowdown was attributed to ACKL's boxing of small integers, but I could be wrong]. I also note in the GNU mailing of June 94 that "There is an Xlib interface via C." Is this different than just using the foreign function interface to C code which uses the Xlib? I do like the convenience of CLX but not the peformance penalties. Thanks in advance to anyone with experience or opinions. regards, -f Fred Lakin :: lakin@pgc.com :: (415) 424-1086 The Performing Graphics Company :: 512 Military Way, Palo Alto, CA 94306 USA http://www.internex.net/pgc/webdog.html From attardi@DI.UniPi.IT Thu Jan 26 03:37:33 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12199; Thu, 26 Jan 95 03:37:33 CST Received: from apollo.di.unipi.it by cli.com (4.1/SMI-4.1) id AA25576; Thu, 26 Jan 95 02:25:50 CST Received: from omega.di.unipi.it by apollo.di.unipi.it with SMTP (1.37.109.4/16.2) id AA02062; Thu, 26 Jan 95 09:23:24 +0100 Organization: Dipartimento di Informatica - Universita' di Pisa - Italy Received: by omega (5.0/SMI-4.1) id AA03208; Thu, 26 Jan 1995 09:23:20 --100 Date: Thu, 26 Jan 1995 09:23:20 --100 From: attardi@DI.UniPi.IT Message-Id: <9501260823.AA03208@omega> To: lakin@pgc.com Cc: gcl@cli.com In-Reply-To: <9501252121.AA03227@pgc.com> (starnet!bass!lakin@apple.com) Subject: Re: speed of CLX under GCL Content-Length: 415 I have built a version of CLX which runs over 20X faster than in AKCL. To do this I had to create a new C interface mechanism (definline), and to make several changes to the original CLX code. All of this is part of my own variant of KCL, called ECoLisp, which you can ftp from: - ftp.icsi.berkeley.edu [128.32.201.7], directory /pub/ai/ecl - ftp.di.unipi.it [131.114.4.36], directory /pub/lang/lisp -- Beppe From siochil@csgrad.cs.vt.edu Thu Jan 26 11:29:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12249; Thu, 26 Jan 95 11:29:56 CST Received: from csgrad.cs.vt.edu by cli.com (4.1/SMI-4.1) id AA26423; Thu, 26 Jan 95 10:21:07 CST Received: by csgrad.cs.vt.edu; (5.65/1.1.8.2/19Sep94-1023AM) id AA12989; Thu, 26 Jan 1995 11:21:01 -0500 Message-Id: <9501261621.AA12989@csgrad.cs.vt.edu> Subject: UNSUBSCRIBE To: gcl@cli.com Date: Thu, 26 Jan 1995 11:21:00 -0500 (EST) From: "Lucia Siochi" X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 16 UNSUBSCRIBE GCL From rpeters2@d.umn.edu Thu Jan 26 11:35:06 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12255; Thu, 26 Jan 95 11:35:06 CST Received: from ub.d.umn.edu by cli.com (4.1/SMI-4.1) id AA26278; Thu, 26 Jan 95 09:50:12 CST Received: (from rpeters2@localhost) by ub.d.umn.edu (8.6.9/8.6.9) id JAA15321 for gcl@cli.com; Thu, 26 Jan 1995 09:50:04 -0600 From: "Randy J. Peterson" Message-Id: <199501261550.JAA15321@ub.d.umn.edu> Subject: UNSUBSCRIBE To: gcl@cli.com Date: Thu, 26 Jan 1995 09:50:03 -0600 (CST) X-Mailer: ELM [version 2.4 PL22] Content-Type: text Content-Length: 16 UNSUBSCRIBE GCL From harmon@xn.ll.mit.edu Thu Jan 26 14:13:12 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12270; Thu, 26 Jan 95 14:13:12 CST Received: from xn.ll.mit.edu by cli.com (4.1/SMI-4.1) id AA26907; Thu, 26 Jan 95 12:59:20 CST Received: from saturn.mi.ll.mit.edu by xn.ll.mit.edu id AA01803g; Thu, 26 Jan 95 14:00:40 EST Received: by saturn.mi.ll.mit.edu; Thu, 26 Jan 95 13:57:34 EST Date: Thu, 26 Jan 95 13:57:34 EST From: harmon@xn.ll.mit.edu Message-Id: <9501261857.AA06133@saturn.mi.ll.mit.edu> To: gcl@cli.com Subject: please unsubscribe me UNSUBSCRIBE GCL From @festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk Fri Jan 27 19:59:28 1995 Return-Path: <@festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk> Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12653; Fri, 27 Jan 95 19:59:28 CST Received: from sun2.nsfnet-relay.ac.uk by cli.com (4.1/SMI-4.1) id AA01577; Fri, 27 Jan 95 19:04:55 CST Via: uk.ac.edinburgh.festival; Sat, 28 Jan 1995 01:04:51 +0000 Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa03958; 28 Jan 95 1:04 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Sat, 28 Jan 95 01:04:16 GMT Date: Sat, 28 Jan 95 01:04:14 GMT Message-Id: <6578.9501280104@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: compiler bug w/ catch & throw To: gcl@cli.com There seems to be a bug, and it looks like it's been around for a long time. The simplest example I've come up with so far is (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) (f) should return 10. Below are two transcripts, one using AKCL 1.615 and the other using GCL 1.1. In each case, I show that F works fine interpreted. But when a compiled F is called, there's an error: The tag A is undefined. Similar problems occur when using block and return-from instead of catch and throw, although this particular F works in GCL 1.1 if rewritten to use block and return-from. ---------------------------------------------------------------------- spottisvax$ akcl AKCL (Austin Kyoto Common Lisp) Version(1.615) Sun Oct 10 18:04:32 BST 1993 Contains Enhancements by W. Schelter >(defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) F >(f) 10 >(compile 'f) Compiling gazonk0.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling gazonk0.lsp. Loading gazonk0.o start address -T 1d8e18 Finished loading gazonk0.o # >(f) Error: The tag A is undefined. Fast links are on: do (use-fast-links nil) for debugging Error signalled by F. Broken at F. Type :H for Help. >>Bye. ---------------------------------------------------------------------- spottisvax$ gcl GCL (GNU Common Lisp) Version(1.1) Mon Nov 14 00:41:18 GMT 1994 Licensed under GNU Public Library License Contains Enhancements by W. Schelter >(defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) F >(f) 10 >(compile 'f) Compiling gazonk0.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling gazonk0.lsp. Loading gazonk0.o start address -T 21e234 Finished loading gazonk0.o # >(f) Error: The tag A is undefined. Fast links are on: do (use-fast-links nil) for debugging Error signalled by F. Broken at F. Type :H for Help. >>Bye. spottisvax$ ---------------------------------------------------------------------- -- jeff From yuasa@katura.tutics.tut.ac.jp Sat Jan 28 22:25:31 1995 Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12765; Sat, 28 Jan 95 22:25:31 CST Received: from higw.tut.ac.jp by cli.com (4.1/SMI-4.1) id AA02863; Sat, 28 Jan 95 21:17:21 CST Received: from tutuser.tut.ac.jp by higw.tut.ac.jp (5.65+1.6W/6.4J.6) id AA03467; Sun, 29 Jan 95 12:20:26 JST Received: from katura.tutics.tut.ac.jp by tutuser.tut.ac.jp (5.65+1.6W/6.4J.6) id AA10869; Sun, 29 Jan 95 12:10:43 +0900 Received: by katura.tutics.tut.ac.jp (5.65+1.6W/6.4JAIN-1.0) id AA03278; Sun, 29 Jan 95 12:14:04 JST Date: Sun, 29 Jan 95 12:14:04 JST From: yuasa@katura.tutics.tut.ac.jp (Taiichi Yuasa) Return-Path: Message-Id: <9501290314.AA03278@katura.tutics.tut.ac.jp> To: jeff@aiai.edinburgh.ac.uk Subject: RE: compiler bug w/ catch & throw Cc: gcl@cli.com >> There seems to be a bug, and it looks like it's been around for >> a long time. The simplest example I've come up with so far is >> >> (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) >> >> (f) should return 10. In the original KCL, both interpreted and compiled f return 10. -- Taiichi From @festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk Sun Jan 29 12:21:37 1995 Return-Path: <@festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk> Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12828; Sun, 29 Jan 95 12:21:37 CST Received: from sun2.nsfnet-relay.ac.uk by cli.com (4.1/SMI-4.1) id AA03486; Sun, 29 Jan 95 11:32:32 CST Via: uk.ac.edinburgh.festival; Sun, 29 Jan 1995 17:31:13 +0000 Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa16733; 29 Jan 95 17:31 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Sun, 29 Jan 95 17:30:48 GMT Date: Sun, 29 Jan 95 17:30:48 GMT Message-Id: <7152.9501291730@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: RE: compiler bug w/ catch & throw To: Taiichi Yuasa In-Reply-To: Taiichi Yuasa's message of Sun, 29 Jan 95 12:14:04 JST Cc: gcl@cli.com > >> There seems to be a bug, and it looks like it's been around for > >> a long time. The simplest example I've come up with so far is > >> > >> (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) > >> > >> (f) should return 10. > > In the original KCL, both interpreted and compiled f return 10. I've just checked in AKCL 1.505, and it it works there too. -- jeff From nils@guru.ims.uni-stuttgart.de Sun Jan 29 13:06:22 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12833; Sun, 29 Jan 95 13:06:22 CST Received: from guru.ims.uni-stuttgart.de by cli.com (4.1/SMI-4.1) id AA03510; Sun, 29 Jan 95 12:21:50 CST Received: by guru.ims.uni-stuttgart.de (Smail3.1.28.1 #16) id m0rYCwT-0003UwC; Sat, 28 Jan 95 14:11 MET Message-Id: Date: Sat, 28 Jan 95 14:11 MET From: nils@guru.ims.uni-stuttgart.de (Cornelis van der Laan) To: gcl@cli.com, jeff@aiai.edinburgh.ac.uk Subject: Re: compiler bug w/ catch & throw I can't verify this here, running a (slightly patched) AKCL 1.624 as well as GCL 1.0 on FreeBSD 1.1.5: Starting /usr/local/bin/akcl ... AKCL (Austin Kyoto Common Lisp) Version(1.624) Sat Jan 21 20:29:04 MET 1995 Contains Enhancements by W. Schelter OPTIMIZE levels: Safety=1 (No runtime error checking), Space=1, Speed=3 Using conditions package CLCS 2/1/90 Loading /home/nils/.akclrc.lisp Using MK defsystem version v2.5 08-JAN-92 Finished loading /home/nils/.akclrc.lisp > >(defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) F >(f) 10 >(compile 'f) ;; Compiling gazonk0.lisp. ;; End of Pass 1. ;; End of Pass 2. OPTIMIZE levels: Safety=1 (No runtime error checking), Space=1, Speed=3 ;; Finished compiling gazonk0.lisp. Loading gazonk0.o Finished loading gazonk0.o # >(f) 10 Compiling with optimization levels of (safety 0) (space 0) (speed 3) results in the very same behaviour. Ditto GCL 1.0. Something must be wrong with your vax port... Nils > GCL (GNU Common Lisp) Version(1.1) Mon Nov 14 00:41:18 GMT 1994 > Licensed under GNU Public Library License > Contains Enhancements by W. Schelter > >(defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) > F > >(f) > 10 > >(compile 'f) > Compiling gazonk0.lsp. > End of Pass 1. > End of Pass 2. > OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, > Speed=3 > Finished compiling gazonk0.lsp. > Loading gazonk0.o > start address -T 21e234 Finished loading gazonk0.o > # > >(f) > Error: The tag A is undefined. > Fast links are on: do (use-fast-links nil) for debugging > Error signalled by F. > Broken at F. Type :H for Help. > >>Bye. From @festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk Sun Jan 29 13:27:47 1995 Return-Path: <@festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk> Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12837; Sun, 29 Jan 95 13:27:47 CST Received: from sun2.nsfnet-relay.ac.uk by cli.com (4.1/SMI-4.1) id AA03518; Sun, 29 Jan 95 12:39:10 CST Via: uk.ac.edinburgh.festival; Sun, 29 Jan 1995 18:39:04 +0000 Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa22052; 29 Jan 95 18:38 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Sun, 29 Jan 95 18:38:39 GMT Date: Sun, 29 Jan 95 18:38:38 GMT Message-Id: <7207.9501291838@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: Re: compiler bug w/ catch & throw To: Cornelis van der Laan , gcl@cli.com In-Reply-To: Cornelis van der Laan's message of Sat, 28 Jan 95 14:11 MET > I can't verify this here, running a (slightly patched) AKCL 1.624 > as well as GCL 1.0 on FreeBSD 1.1.5: Interesting... The cases that didn't work for me were on a 386 running FreeBSD 2.0-RELEASE #3, but it's possible the AKCL and the GCl were compiled using some different OS (the AKCl almost certainly was). The case that did work for me (AKCL 1.505) was on a Sun (though it was also a system built before some OS upgrades). Anyway, this raises several possibilities... -- jeff From reggie@miles.phys.washington.edu Sun Jan 29 15:32:09 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12847; Sun, 29 Jan 95 15:32:09 CST Received: from miles.phys.washington.edu (pacts44.phys.washington.edu) by cli.com (4.1/SMI-4.1) id AA03641; Sun, 29 Jan 95 14:43:13 CST Received: from miles.phys.washington.edu (localhost [127.0.0.1]) by miles.phys.washington.edu (8.6.9/8.6.9) with ESMTP id MAA03633; Sun, 29 Jan 1995 12:43:16 -0800 Message-Id: <199501292043.MAA03633@miles.phys.washington.edu> To: Jeff Dalton Cc: Taiichi Yuasa , gcl@cli.com Subject: Re: compiler bug w/ catch & throw In-Reply-To: Your message of "Sun, 29 Jan 1995 17:30:48 GMT." <7152.9501291730@subnode.aiai.ed.ac.uk> Date: Sun, 29 Jan 1995 12:43:12 -0800 From: "Reginald S. Perry" >"Jeff" == Jeff Dalton writes: > >> There seems to be a bug, and it looks like it's been around for > >> a long time. The simplest example I've come up with so far is > >> > >> (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) > >> > >> (f) should return 10. > > In the original KCL, both interpreted and compiled f return 10. > I've just checked in AKCL 1.505, and it it works there too. Here is the result from gcl-1.0 on Linux. miles$ gcl GCL (GNU Common Lisp) Version(1.0) Wed Aug 24 01:38:16 PDT 1994 Licensed under GNU Public Library License Contains Enhancements by W. Schelter >(defun f () (catch 'a (funcall #'(lambda () (throw 'a 10))))) F >(f) 10 >(compile (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10)))))) Compiling gazonk0.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling gazonk0.lsp. Loading gazonk0.o start address -T 1e0000 Finished loading gazonk0.o # >(f) 10 >Bye. miles$ gcl GCL (GNU Common Lisp) Version(1.0) Wed Aug 24 01:38:16 PDT 1994 Licensed under GNU Public Library License Contains Enhancements by W. Schelter >(compile (defun f () (catch 'a (funcall #'(lambda () (throw 'a 10)))))) Compiling gazonk0.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling gazonk0.lsp. Loading gazonk0.o start address -T 1e0000 Finished loading gazonk0.o # >(f) 10 >Bye. ------------------- Reginald S. Perry -- University of Washington, Department of Physics FM-15 Seattle, Washington 98195 reggie@phys.washington.edu From Chandan.Haldar@blr.sni.de Sun Jan 29 23:50:47 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA12891; Sun, 29 Jan 95 23:50:47 CST Received: from mail.sni.de by cli.com (4.1/SMI-4.1) id AA04401; Sun, 29 Jan 95 22:41:36 CST Received: from blrsn01.blr.sni.de by mail.sni.de with SMTP (PP) id <25439-0@mail.sni.de>; Mon, 30 Jan 1995 05:40:55 +0100 Received: from blrast03.blr.sni.de by blr.sni.de (4.1/SMI-4.1) id AA10147; Mon, 30 Jan 95 10:12:29+050 Date: Mon, 30 Jan 95 10:12:29+050 From: Chandan.Haldar@blr.sni.de (Dr. Chandan Haldar) Message-Id: <9501300512.AA10147@blr.sni.de> Received: by blrast03.blr.sni.de (5.0/SMI-SVR4) id AA00369; Mon, 30 Jan 95 10:24:21 GMT To: jeff@aiai.edinburgh.ac.uk Cc: yuasa@katura.tutics.tut.ac.jp, gcl@cli.com In-Reply-To: <7152.9501291730@subnode.aiai.ed.ac.uk> (message from Jeff Dalton on Sun, 29 Jan 95 17:30:48 GMT) Subject: RE: compiler bug w/ catch & throw Content-Length: 565 > I've just checked in AKCL 1.505, and it it works there too. Also in AKCL 1.600 and in GCL 1.1 (on Sparc 10 Solaris 1 with cc). -------------------------------------------------------------------------- Dr. Chandan Haldar -------------------------------------------------------------------------- Siemens Information Systems Limited 29 Infantry Road, Bangalore 560 001, India. Phone: +91 80 551 1012 Fax: +91 80 551 1214 Internet: Chandan.Haldar@blr.sni.de -------------------------------------------------------------------------- From readingj@cerfnet.com Mon Jan 30 14:28:26 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13186; Mon, 30 Jan 95 14:28:26 CST Received: from nic.cerf.net by cli.com (4.1/SMI-4.1) id AA06203; Mon, 30 Jan 95 13:23:36 CST Received: (from readingj@localhost) by nic.cerf.net (8.6.9/8.6.9) id LAA02737 for gcl@cli.com; Mon, 30 Jan 1995 11:23:31 -0800 Date: Mon, 30 Jan 1995 11:23:31 -0800 From: "John D. Reading" Message-Id: <199501301923.LAA02737@nic.cerf.net> To: gcl@cli.com Subject: Help building PCL?? Sorry to bug you, but has anyone out there built PCL under GCL 1.1 on Solaris 2.3? I am getting a core dump while building PCL, and I am not progressing very well on tracking down the source of the error. Can anyone help me? Thanks, John Reading readingj@cerf.net reading@io.dsd.litton.com From wfs@math.utexas.edu Mon Jan 30 15:09:11 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13206; Mon, 30 Jan 95 15:09:11 CST Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA06353; Mon, 30 Jan 95 13:57:33 CST Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA21073; Mon, 30 Jan 95 13:54:25 -0600 Date: Mon, 30 Jan 95 13:54:20 -0600 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Mon, 30 Jan 95 13:54:20 -0600 Message-Id: <9501301954.AA00894@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA00894; Mon, 30 Jan 95 13:54:20 -0600 To: gcl@cli.com In-Reply-To: "John D. Reading"'s message of Mon, 30 Jan 1995 11:23:31 -0800 <199501301923.LAA02737@nic.cerf.net> Subject: Re: Help building PCL?? Reply-To: wfs@math.utexas.edu PCL: On math.utexas.edu:pub/gcl/pcl-gcl-1.1.tgz should contain correct code for pcl and gcl-1.1 This was contributed by Richard Harris who has done a good deal of work pcl. -------------------- Xlib: There are two ways of having Xlib 1) CLX lisp code with X distribution. It is desirable to use some speedups for the arithmetic operations rascal.ics.utexas.edu:7 -rw-r--r-- 1 root 6494 Oct 22 1991 /usr2/ftp/pub/clx-5-opts.tar.Z contains such speedups. 2) Use something like -rw-r--r-- 1 wfs 106547 Oct 3 19:24 /u/ftp/pub/gcl/xgcl-2.tgz which contains bindings for Xlib functions in lisp, but does a link with the C xlibrary. I am also about to release a TCL/TK interface for GCL which will provide a third (and I think for most purposes) preferable way of interacting with the window system. -------------------- Catch: It appears from various evidence that the 'catch' bug reported by dalton, was probably due to a changed include file on his system or a changed C compiler flag, since the base gcl system had been compiled (eg changing size of jmp_buf). It was not due to a gcl/kcl/akcl bug. Bill Schelter From kbibb@qualcomm.com Mon Jan 30 15:38:46 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13210; Mon, 30 Jan 95 15:38:46 CST Received: from jafar.qualcomm.com by cli.com (4.1/SMI-4.1) id AA06381; Mon, 30 Jan 95 14:10:03 CST Received: (kbibb@localhost) by jafar.qualcomm.com (8.6.8/QC-BSD-2.4) id MAA27759; Mon, 30 Jan 1995 12:10:09 -0800 Date: Mon, 30 Jan 1995 12:10:09 -0800 From: Ken Bibb Message-Id: <199501302010.MAA27759@jafar.qualcomm.com> To: gcl@cli.com, readingj@cerfnet.com Subject: Re: Help building PCL?? > From: "John D. Reading" > > Sorry to bug you, but has anyone out there built PCL under GCL 1.1 on > Solaris 2.3? I am getting a core dump while building PCL, and I am > not progressing very well on tracking down the source of the error. > > Can anyone help me? That's the same error I'm getting :D I think I saw a post saying that the version of vprintf being used in Solaris is at fault. -- Ken Bibb "Through the fire backwards kbibb@qualcomm.com Again and Again jester@crash.cts.com Return to base." Peter Hammill--"A Headlong Stretch" From kbibb@qualcomm.com Mon Jan 30 18:00:02 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13247; Mon, 30 Jan 95 18:00:02 CST Received: from jafar.qualcomm.com by cli.com (4.1/SMI-4.1) id AA06820; Mon, 30 Jan 95 16:06:35 CST Received: (kbibb@localhost) by jafar.qualcomm.com (8.6.8/QC-BSD-2.4) id OAA17020; Mon, 30 Jan 1995 14:06:35 -0800 Date: Mon, 30 Jan 1995 14:06:35 -0800 From: Ken Bibb Message-Id: <199501302206.OAA17020@jafar.qualcomm.com> To: gcl@cli.com, wfs@fireant.ma.utexas.edu Subject: Re: Help building PCL?? > From: wfs@fireant.ma.utexas.edu (Bill Schelter) > > PCL: > On math.utexas.edu:pub/gcl/pcl-gcl-1.1.tgz > > should contain correct code for pcl and gcl-1.1 > > This was contributed by Richard Harris who has done a good deal of work > pcl. My last attempt to build this was two days ago using the version that was on fireant at that time. It had the same prob that all versions of PCL have had on Solaris 2.3--it core dumps when it reaches a certain point in the compilation. PCL appears to compile correctly under SunOS 4.x. -- Ken Bibb "Through the fire backwards kbibb@qualcomm.com Again and Again jester@crash.cts.com Return to base." Peter Hammill--"A Headlong Stretch" From readingj@cerfnet.com Mon Jan 30 19:07:50 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13264; Mon, 30 Jan 95 19:07:50 CST Received: from nic.cerf.net by cli.com (4.1/SMI-4.1) id AA07062; Mon, 30 Jan 95 17:30:53 CST Received: (from readingj@localhost) by nic.cerf.net (8.6.9/8.6.9) id PAA07035 for gcl@cli.com; Mon, 30 Jan 1995 15:30:41 -0800 Date: Mon, 30 Jan 1995 15:30:41 -0800 From: "John D. Reading" Message-Id: <199501302330.PAA07035@nic.cerf.net> To: gcl@cli.com Subject: Re: Help building PCL?? Sorry, I have been a little unresponsive on this. wfs has been helping, as some of the messages to the list have shown. The particular problem I am seeing occurs during the compilation of slots-boot.lisp, and seems to be related to the function MAKE-METHOD-FUNCTION which is called from this file. I am currently using pcl-gcl-1.1, gcc 2.6.3, and gcl 1.1 Anybody who wants to play is welcome to come along ;-) BTW: Having commented out the 4 functions that call MAKE-METHOD-FUNCTION in slots-boot.lisp I have found that the compilation proceeds then to braid.lisp, at which point it dumps core again. I haven't played with this too much, but if I didn't screw up something really bad then commenting out the function in braid.lisp which calls MAKE-METHOD-FUNCTION didn't seem to stop it from dumping core. For some reason I feel like I am swimming upstream here. John readingj@cerf.net reading@io.dsd.litton.com From raz@superior.powercerv.com Mon Jan 30 19:12:29 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13268; Mon, 30 Jan 95 19:12:29 CST Received: from riverside.mr.net by cli.com (4.1/SMI-4.1) id AA07135; Mon, 30 Jan 95 17:45:07 CST Received: from powercerv.com by riverside.mr.net (8.6.9/SMI-4.1.R931202) id RAA16887; Mon, 30 Jan 1995 17:45:16 -0600 Received: by powercerv.com (5.0/SMI-SVR4) id AA11120; Mon, 30 Jan 1995 17:44:13 +0600 Date: Mon, 30 Jan 1995 17:44:13 +0600 From: raz@superior.powercerv.com (Randal V. Zoeller) Message-Id: <9501302344.AA11120@powercerv.com> To: gcl@cli.com Subject: Unsubscribe Content-Length: 35 Please unsubscribe raz@itasca.com From @festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk Mon Jan 30 21:16:26 1995 Return-Path: <@festival.edinburgh.ac.uk:jeff@aiai.edinburgh.ac.uk> Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13292; Mon, 30 Jan 95 21:16:26 CST Received: from sun2.nsfnet-relay.ac.uk by cli.com (4.1/SMI-4.1) id AA07374; Mon, 30 Jan 95 20:25:19 CST Via: uk.ac.edinburgh.festival; Mon, 30 Jan 1995 17:51:03 +0000 Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa11495; 30 Jan 95 17:50 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Mon, 30 Jan 95 17:50:33 GMT Date: Mon, 30 Jan 95 17:50:32 GMT Message-Id: <9784.9501301750@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: Re: compiler bug w/ catch & throw To: Cornelis van der Laan , gcl@cli.com In-Reply-To: Jeff Dalton's message of Sun, 29 Jan 95 18:38:38 GMT > In-Reply-To: Cornelis van der Laan's message of Sat, 28 Jan 95 14:11 MET > > > I can't verify this here, running a (slightly patched) AKCL 1.624 > > as well as GCL 1.0 on FreeBSD 1.1.5: > > Interesting... The cases that didn't work for me were on a 386 > running FreeBSD 2.0-RELEASE #3, but it's possible the AKCL and > the GCl were compiled using some different OS (the AKCl almost > certainly was). > > The case that did work for me (AKCL 1.505) was on a Sun (though > it was also a system built before some OS upgrades). > > Anyway, this raises several possibilities... Ok, I've figured it out. Thanks to everyone who responded. It was all very helpful. And the answer is ... not a KCL or GCL bug at all. For some reason FreeBSD 2 changed the size of the area saved by setjmp. In the various BSDs used to build the various Lisps it was 10; in FreeBSD 2.0, it was 8. Why the change? Not clear. Presumably it's "more right" in some sense; but it does make otherwise compatible systems incompatible. -- jeff From root@guru.ims.uni-stuttgart.de Tue Jan 31 15:20:33 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13482; Tue, 31 Jan 95 15:20:33 CST Received: from guru.ims.uni-stuttgart.de by cli.com (4.1/SMI-4.1) id AA11085; Tue, 31 Jan 95 13:43:22 CST Received: by guru.ims.uni-stuttgart.de (Smail3.1.28.1 #16) id m0rZOXR-0003UwC; Tue, 31 Jan 95 20:46 MET Message-Id: Date: Tue, 31 Jan 95 20:46 MET From: root@guru.ims.uni-stuttgart.de (Cornelis van der Laan) To: gcl@cli.com, jeff@aiai.edinburgh.ac.uk, nils@guru.ims.uni-stuttgart.de Subject: Re: compiler bug w/ catch & throw And as far as I can see (not that far, but...) they only set the jmp_buf size to it's real size, which is 8. The code which uses it - setjmp and longjmp - is identical on FreeBSD 1.1.5 and 2.0. All OSes which use 10 bytes waste 2... So maybe editing the header file to read 10 and recompiling world helps? Nils From frkorze@srv.PacBell.COM Wed Feb 1 20:43:36 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA13729; Wed, 1 Feb 95 20:43:36 CST Received: from ns.PacBell.COM by cli.com (4.1/SMI-4.1) id AA15781; Wed, 1 Feb 95 19:48:14 CST Received: from srv.PacBell.COM (mother.srv.PacBell.COM) by ns.PacBell.COM (4.1/PacBell-11/15/94) id AA10449; Wed, 1 Feb 95 17:48:24 PST Received: from diablo.srv.PacBell.COM by srv.PacBell.COM (4.1/SMI-4.0) id AA02736; Wed, 1 Feb 95 17:48:21 PST Received: by diablo.srv.PacBell.COM (4.1/SMI-4.1) id AA00893; Wed, 1 Feb 95 17:48:21 PST Date: Wed, 1 Feb 95 17:48:21 PST From: frkorze@srv.PacBell.COM (Frank Korzeniewski) Message-Id: <9502020148.AA00893@diablo.srv.PacBell.COM> To: gcl@cli.com Subject: subscribe subscribe From ias5!c3po.iasi.com!basham@csn.net Fri Feb 3 13:58:24 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA14155; Fri, 3 Feb 95 13:58:24 CST Received: from csn.net (csn.org) by cli.com (4.1/SMI-4.1) id AA21227; Fri, 3 Feb 95 13:02:18 CST Received: from ias5.UUCP by csn.net with UUCP id AA00830 (5.65c/IDA-1.4.4 for gcl@cli.com); Fri, 3 Feb 1995 12:02:31 -0700 Received: from c3po.iasi.com by iasi.com (4.1/SMI-4.1) id AA27752; Fri, 3 Feb 95 12:02:33 MST Received: by c3po.iasi.com (4.1/SMI-4.1) id AA23706; Fri, 3 Feb 95 12:05:02 MST Date: Fri, 3 Feb 95 12:05:02 MST From: basham@c3po.iasi.com (Bryan Basham) Message-Id: <9502031905.AA23706@c3po.iasi.com> To: gcl@cli.com Subject: Status of GCL on IRIX 5.X Cc: basham@c3po.iasi.com Hello, Does any know if GCL will compile on SGI's OS IRIX 5.X? If so, which is the best FTP site to get the code? Thanks, Bryan From ajwright@acssun.pstcc.cc.tn.us Sat Feb 11 14:08:04 1995 Return-Path: Received: from cli.com by ftp (4.1/SMI-4.1) id AA01639; Sat, 11 Feb 95 14:08:04 CST Received: from PSTCC4.PSTCC.CC.TN.US by cli.com (4.1/SMI-4.1) id AA16457; Sat, 11 Feb 95 12:58:03 CST Received: from ACSSUN.PSTCC.CC.TN.US by pstcc.cc.tn.us (PMDF V4.2-11 #7176) id <01HMXBWN0KJ48ZE7U7@pstcc.cc.tn.us>; Sat, 11 Feb 1995 13:56:28 EDT Received: (from ajwright@localhost) by acssun.pstcc.cc.tn.us (8.6.9/8.6.9) id NAA03405; Sat, 11 Feb 1995 13:55:43 -0500 Date: Sat, 11 Feb 1995 13:55:43 -0500 (EST) From: "A. John Wright" Subject: GCL 1.1 and Solaris 2.3 To: gcl@cli.com Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Transfer-Encoding: 7BIT After trying everything else I can think of . . . The CS department that works on the Sun SPARCstation 20/50 I admin has requested that I get a version of LISP (preferably GCL) on this computer. So, I attempted to get gcl 1.1 to compile, but to no avail. I read all the online documentation for it and the make lives up to the point that the "raw lisp compiler" has to read in several object files (defmacro.o etc) and output the final compiler. It dies on/with: /------- ../unixport/raw_gcl ../unixport/ < foo GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o ~s is not a ~a Lisp initialization failed. \------- I don't know lisp, so I tried what I *do* know: I experimented with different C compilers, assemblers and linkers. I even tried stepping through the process by hand (That'll work! NOT!) I managed to get it to the state described above (Which is the best so far) using GCC (2.6.1), the assembler that comes with Solaris 2.3, and the defaults for "add-defs solaris". Has anyone else had any experience with this combination, (suggestions!) or know where I can find pre-compiled binaries for my system? Any assistance is appreciated! Thanks, (HELP!) --aj ajwright@acssun.pstcc.cc.tn.us ajwright@acssun.pstcc.cc.tn.us ---------------------------------------------------------------------------- A. John `AJ' Wright -- UNIX Technician -- Ext: 7068 From ajwright@acssun.pstcc.cc.tn.us Sat Feb 11 14:21:32 1995 Return-Path: Received: from cli.com by ftp (4.1/SMI-4.1) id AA01647; Sat, 11 Feb 95 14:21:32 CST Received: from PSTCC4.PSTCC.CC.TN.US by cli.com (4.1/SMI-4.1) id AA16485; Sat, 11 Feb 95 13:20:16 CST Received: from ACSSUN.PSTCC.CC.TN.US by pstcc.cc.tn.us (PMDF V4.2-11 #7176) id <01HMXCP2VDVK8ZE8BX@pstcc.cc.tn.us>; Sat, 11 Feb 1995 14:18:37 EDT Received: (from ajwright@localhost) by acssun.pstcc.cc.tn.us (8.6.9/8.6.9) id OAA04248; Sat, 11 Feb 1995 14:17:52 -0500 Date: Sat, 11 Feb 1995 14:17:52 -0500 (EST) From: "A. John Wright" Subject: GCL 1.1 and Solaris 2.3 To: gcl@cli.com Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Transfer-Encoding: 7BIT After trying everything else I can think of . . . The CS department that works on the Sun SPARCstation 20/50 I admin has requested that I get a version of LISP (preferably GCL) on this computer. So, I attempted to get gcl 1.1 to compile, but to no avail. I read all the online documentation for it and the make lives up to the point that the "raw lisp compiler" has to read in several object files (defmacro.o etc) and output the final compiler. It dies on/with: /------- ../unixport/raw_gcl ../unixport/ < foo GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o ~s is not a ~a Lisp initialization failed. \------- I don't know lisp, so I tried what I *do* know: I experimented with different C compilers, assemblers and linkers. I even tried stepping through the process by hand (That'll work! NOT!) I managed to get it to the state described above (Which is the best so far) using GCC (2.6.1), the assembler that comes with Solaris 2.3, and the defaults for "add-defs solaris". Has anyone else had any experience with this combination, (suggestions!) or know where I can find pre-compiled binaries for my system? Any assistance is appreciated! Thanks, (HELP!) --aj ajwright@acssun.pstcc.cc.tn.us ajwright@acssun.pstcc.cc.tn.us ---------------------------------------------------------------------------- A. John `AJ' Wright -- UNIX Technician -- Ext: 7068 From ajwright@acssun.pstcc.cc.tn.us Sat Feb 11 14:52:26 1995 Return-Path: Received: from cli.com by ftp (4.1/SMI-4.1) id AA01665; Sat, 11 Feb 95 14:52:26 CST Received: from acssun.pstcc.cc.tn.us by cli.com (4.1/SMI-4.1) id AA16496; Sat, 11 Feb 95 14:04:41 CST Date: Sat, 11 Feb 95 14:01:47 CST From: ajwright@acssun.pstcc.cc.tn.us Subject: GCL 1.1 and Solaris 2.3 To: gcl@cli.com Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Transfer-Encoding: 7BIT After trying everything else I can think of . . . The CS department that works on the Sun SPARCstation 20/50 I admin has requested that I get a version of LISP (preferably GCL) on this computer. So, I attempted to get gcl 1.1 to compile, but to no avail. I read all the online documentation for it and the make lives up to the point that the "raw lisp compiler" has to read in several object files (defmacro.o etc) and output the final compiler. It dies on/with: /------- ./unixport/raw_gcl ../unixport/ < foo GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o ~s is not a ~a Lisp initialization failed. \------- I don't know lisp, so I tried what I *do* know: I experimented with different C compilers, assemblers and linkers. I even tried stepping through I managed to get it to the state described above (Which is the best so far) using GCC (2.6.1), the assembler that comes with Solaris 2.3, and the defaults for "add-defs solaris". Has anyone else had any experience with this combination, (suggestions!) or know where I can find pre-compiled binaries for my system? Any assistance is appreciated! Thanks, (HELP!) --aj ajwright@acssun.pstcc.cc.tn.us ajwright@acssun.pstcc.cc.tn.us ---------------------------------------------------------------------------- A. John `AJ' Wright -- UNIX Technician -- Ext: 7068 From wfs@math.utexas.edu Sat Feb 11 18:43:26 1995 Return-Path: Received: from cli.com by ftp (4.1/SMI-4.1) id AA01716; Sat, 11 Feb 95 18:43:26 CST Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA16687; Sat, 11 Feb 95 17:55:34 CST Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA18640; Sat, 11 Feb 95 17:54:07 -0600 Date: Sat, 11 Feb 95 17:54:00 -0600 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Sat, 11 Feb 95 17:54:00 -0600 Message-Id: <9502112354.AA03498@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA03498; Sat, 11 Feb 95 17:54:00 -0600 To: ajwright@acssun.pstcc.cc.tn.us Cc: gcl@cli.com In-Reply-To: "A. John Wright"'s message of Sat, 11 Feb 1995 13:55:43 -0500 (EST) Subject: Re: GCL 1.1 and Solaris 2.3 Reply-To: wfs@math.utexas.edu It does compile correctly with gcc (100's have done so), but I dont think the sun c compiler is correct. At least it was not last time i checked. From grib@mamba.arlut.utexas.edu Wed Feb 15 09:28:27 1995 Return-Path: Received: from cli.com by ftp (4.1/SMI-4.1) id AA02438; Wed, 15 Feb 95 09:28:27 CST Received: from ns1.arlut.utexas.edu by cli.com (4.1/SMI-4.1) id AA25983; Wed, 15 Feb 95 08:22:12 CST Received: from mamba.arlut.utexas.edu (mamba.arlut.utexas.edu [129.116.176.42]) by ns1.arlut.utexas.edu (8.6.9/8.6.9) with SMTP id IAA10397 for ; Wed, 15 Feb 1995 08:22:34 -0600 Received: by mamba.arlut.utexas.edu (5.0/SMI-SVR4) id AA12398; Wed, 15 Feb 1995 08:22:34 +0600 From: grib@mamba.arlut.utexas.edu (William Gribble) Message-Id: <9502151422.AA12398@mamba.arlut.utexas.edu> Subject: Building gcl-1.1 under recent Linux? To: gcl@cli.com Date: Wed, 15 Feb 1995 08:22:32 -0600 (CST) X-Mailer: ELM [version 2.4 PL22] Content-Type: text Content-Length: 1570 Has anyone built gcl-1.1 (or 1.0, for that matter) under a recent Linux? (``recent'' meaning > 1.1.51)? I obtained a binary package of 1.0 already built which works just fine, but I wanted to rebuild it myself. I applied the patches I found on sunsite (mostly pretty minor) and proceeded to build (using gcc 2.6.2). The problem is that unexec() doesn't work right. The raw_gcl made with ld works fine, but after the dump to saved_gcl there's no a.out header on the file. od shows that the header is actually there, but there are 0x10000 zeros before it starts. I tried to debug it, but unexec() seems a little bit like black magic to me so I didn't make much progress. The nearest thing I could determine was that write_text_and_data() (?) thought the text segment consisted of 0x10000 0s, followed by the a.out header, followed by the text. It was writing all of this starting at file offset 0x00, so it was overwriting the header that a previous function had written out. At this point I decided to punt. I grabbed the unexsunos4.c from emacs-19.28 and hacked it up a little to include save.c and define N_PAGSIZ to PAGE_SIZE (as the emacs linux.h does). This fixed the header location problem, but the dumped image doesn't run properly; it just prints the startup banner and hangs without giving a prompt, taking an excessive amount of CPU but making no system calls. I dug around in the mailings list archives and saw that lots of people had similar problems with the IRIX port. Was there a canonical solution? Thanks for any advice-- Bill Gribble From n2ell!pcolsen@cs.UMD.EDU Sun Feb 19 11:36:51 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00436; Sun, 19 Feb 95 11:36:51 CST Received: from mimsy.cs.UMD.EDU by cli.com (4.1/SMI-4.1) id AA07133; Sun, 19 Feb 95 10:40:35 CST Received: from n2ell.UUCP by mimsy.cs.UMD.EDU (8.6.9/UMIACS-0.9/04-05-88) id LAA02632; Sun, 19 Feb 1995 11:40:59 -0500 Received: by n2ell (Smail3.1.29.1 #3) id m0rgERV-0003A3C; Sun, 19 Feb 95 11:25 EST Message-Id: Date: Sun, 19 Feb 95 11:25 EST From: pcolsen@n2ell (Peter C. Olsen P.E.) To: gcl@cli.com Subject: Can't compile GCL-1.1 under Linux Reply-To: pcolsen@acm.org I can't seem to compile GCL-1.1 under Linux --- and I can't figure out why. I'd really appreciate some advice. Everything seems to run just fine until I get to the part where I have to execute "../unixport/saved_gcl" --- but that file is not executable. Any suggestions? (I've included the output of my "make" below.) Peter P.S. I'd like to make the package rather than snarf down a compiled version (which I can't seem to find anyway) so I can make Maxima-5.0 (which needs the gcl *.o files). ----------------------------------------------------------------- Here's a copy of the make output. (I've broken some long lines.) Here's what I did: bash$ make >~/gcl-hist sh: ../unixport/saved_gcl: cannot execute binary file make[1]: *** [collectfn.o] Error 126 make: *** [all] Error 1 Here's what I got: (cd bin; make all) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/bin' make[1]: Nothing to be done for `all'. make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/bin' (cd mp ; make all) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/mp' make all1 "MPFILES=./mpi-386d.o ./libmport.a" make[2]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/mp' make[2]: Nothing to be done for `all1'. make[2]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/mp' make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/mp' rm -f o/cmpinclude.h ; ln h/cmpinclude.h o (cd o; make all) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/o' make[1]: Nothing to be done for `all'. make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/o' (cd lsp; make all) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/lsp' make[1]: Nothing to be done for `all'. make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/lsp' (cd cmpnew; make all) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/cmpnew' make[1]: Nothing to be done for `all'. make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/cmpnew' (cd unixport; make saved_gcl) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/unixport' make[1]: `saved_gcl' is up to date. make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/unixport' make command make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1' rm -f xbin/gcl ; echo "#!/bin/sh" > xbin/gcl; \ echo "/mnt/sdc1/local/src/gcl-1.1/unixport/saved_gcl -dir /mnt/sdc1/local/src/gcl-1.1/unixport/ \$@ " \ >> xbin/gcl chmod 755 xbin/gcl make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1' (cd cmpnew ; make collectfn.o) make[1]: Entering directory `/mnt/sdc1/local/src/gcl-1.1/cmpnew' ../unixport/saved_gcl ../unixport/ -compile collectfn make[1]: Leaving directory `/mnt/sdc1/local/src/gcl-1.1/cmpnew' Here's the listing for saved_gcl and the machine type I was using: bash$ ls -l /mnt/sdc1/local/src/gcl-1.1/unixport/saved_gcl -rwxr-xr-x 1 pcolsen bin 2134748 Feb 19 10:00 /mnt/sdc1/local/src/gcl-1.1/unixport/saved_gcl bash$ more machine 386-linux For what it's worth, "raw_gcl" does run bash$ ./raw_gcl GCL (GNU Common Lisp) April 1994 16384 pages loading ../lsp/export.lsp Initializing ../lsp/defmacro.o Initializing ../lsp/evalmacros.o Initializing ../lsp/top.o Initializing ../lsp/module.o loading ../lsp/autoload.lsp > From grib@mamba.arlut.utexas.edu Sun Feb 19 12:59:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00455; Sun, 19 Feb 95 12:59:56 CST Received: from ns1.arlut.utexas.edu by cli.com (4.1/SMI-4.1) id AA07203; Sun, 19 Feb 95 12:04:03 CST Received: from mamba.arlut.utexas.edu (mamba.arlut.utexas.edu [129.116.176.42]) by ns1.arlut.utexas.edu (8.6.9/8.6.9) with SMTP id MAA25961; Sun, 19 Feb 1995 12:04:28 -0600 Received: by mamba.arlut.utexas.edu (5.0/SMI-SVR4) id AA20928; Sun, 19 Feb 1995 12:04:27 +0600 From: grib@mamba.arlut.utexas.edu (William Gribble) Message-Id: <9502191804.AA20928@mamba.arlut.utexas.edu> Subject: Re: Can't compile GCL-1.1 under Linux To: pcolsen@acm.org Date: Sun, 19 Feb 1995 12:04:25 -0600 (CST) Cc: gcl@cli.com In-Reply-To: from "Peter C. Olsen P.E." at Feb 19, 95 11:25:00 am X-Mailer: ELM [version 2.4 PL22] Content-Type: text Content-Length: 2721 I had this same difficulty, and have solved (or at least found a way around) the problem. But whoever maintains the Linux port shoud think about changing the way the current system works. The problem: recent binutils distributions for Linux build QMAGIC binaries by default. The differences between QMAGIC and the older ZMAGIC binaries are (1) a QMAGIC binary has the struct exec header as a part of the text segment (2) the first page of the process address space is unmapped at run time for a QMAGIC binary, to help trap NULL pointer dereferences (you get a segmentation violation if you try to use a NULL pointer). While the QMAGIC binary is a nifty idea, it completely breaks the unexec() that's compiled for Linux with gcl-1.[01]. If you will use od(1) to look at your un-executable saved_gcl image, you will see a page of 0 bytes at the beginning, and then the 0314 QMAGIC magic number. This is your problem. Without the magic number at the start of the file, Linux thinks it's just a binary data file. The workaround: Force a non-QMAGIC binary. Unfortunately, QMAGIC has *replaced* ZMAGIC, which was the original Linux demand-paged executable format. Your only option with the current binutils is to use the -N option to ld, which makes an NMAGIC binary. The problem with NMAGIC is that it's not a demand-paged format and is certainly archaic. I'm not sure exactly what difference it makes to the kernel, but I'm almost certain that an NMAGIC lisp image will use memory less efficiently than a QMAGIC binary, forcing more of lisp to be in physical memory at one time. To implement this, edit gcl-1.1/unixport/makefile and change the line ``LDCC=$(CC)'' to ``LDCC=$(CC) -N''. What this does is force the original raw_gcl image to be linked as NMAGIC. unexec() (which dumps the running process to a file) copies the header of the running process for the new, dumped image. Therefore, every dump you do, including the one done by the makefile to generate saved_gcl, will create an NMAGIC binary. A real solution: patch gcl-1.1/c/unexlin.c to correctly handle QMAGIC binaries. Since the current unexec() code came from an older version of Emacs, my guess is that getting the Linux unexec() from the current Emacs distribution would be the easiest thing to do. Good luck with the compile. By the way, I'm about to mail a small patch to the gcl list which should fix the socket-stream code for Linux, if you're trying to use it. Actually, sockets are not compiled in by default; if you want it, edit gcl-1.1/h/386-linux.h and comment out the line which says #undef RUN_PROCESS. Then you can do nifty stuff like (setf sock (si:make-socket-stream 'foo.bar.com 23)). Bill Gribble From grib@mamba.arlut.utexas.edu Sun Feb 19 13:02:11 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00460; Sun, 19 Feb 95 13:02:11 CST Received: from ns1.arlut.utexas.edu by cli.com (4.1/SMI-4.1) id AA07208; Sun, 19 Feb 95 12:12:07 CST Received: from mamba.arlut.utexas.edu (mamba.arlut.utexas.edu [129.116.176.42]) by ns1.arlut.utexas.edu (8.6.9/8.6.9) with SMTP id MAA25987 for ; Sun, 19 Feb 1995 12:12:32 -0600 Received: by mamba.arlut.utexas.edu (5.0/SMI-SVR4) id AA21055; Sun, 19 Feb 1995 12:12:30 +0600 From: grib@mamba.arlut.utexas.edu (William Gribble) Message-Id: <9502191812.AA21055@mamba.arlut.utexas.edu> Subject: small patch against gcl-1.1 (run_process.c) To: gcl@cli.com Date: Sun, 19 Feb 1995 12:12:29 -0600 (CST) X-Mailer: ELM [version 2.4 PL22] Content-Type: text Content-Length: 960 This patch corrects an error in gcl-1.1/c/run_process.c. The sin_port field in struct sockaddr_in should be in network byte order, not host byte order. On Suns, I believe the two are the same; on x86 machines, they are not. The htons() macro does the right thing for the platform you're compiling on. On Linux (and x86 BSD machines, I am pretty sure) this caused calls to (si:make-socket-stream) to try to connect to the wrong port, usually generating a ``connection refused'' error. Bill Gribble --- gcl-1.1.original/c/run_process.c Fri Oct 21 11:44:10 1994 +++ gcl-1.1/c/run_process.c Sun Feb 19 11:37:37 1995 @@ -80,7 +80,7 @@ bzero((char *)&sock_add, sizeof(sock_add)); bcopy(hp->h_addr, (char *)&sock_add.sin_addr, hp->h_length); sock_add.sin_family = hp->h_addrtype; - sock_add.sin_port = server; + sock_add.sin_port = htons((short)server); sock = socket( hp->h_addrtype, SOCK_STREAM , 0); From blake@edge.ercnet.com Mon Feb 20 00:48:39 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00546; Mon, 20 Feb 95 00:48:39 CST Received: from edge.ercnet.com by cli.com (4.1/SMI-4.1) id AA07861; Sun, 19 Feb 95 23:58:24 CST Received: from ip202.ercnet.com (ip202.ercnet.com [199.0.68.202]) by edge.ercnet.com (8.6.9/8.6.9) with SMTP id XAA18619 for ; Sun, 19 Feb 1995 23:58:40 -0600 Message-Id: <199502200558.XAA18619@edge.ercnet.com> X-Sender: blake@edge.ercnet.com X-Mailer: Windows Eudora Version 1.4.4 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Sun, 19 Feb 1995 23:58:59 -0600 To: gcl@cli.com From: blake@edge.ercnet.com (Blake McBride) Subject: Re: Can't compile GCL-1.1 under Linux >The problem: recent binutils distributions for Linux build QMAGIC binaries >by default. The differences between QMAGIC and the older ZMAGIC binaries >are (1) a QMAGIC binary has the struct exec header as a part of the >text segment (2) the first page of the process address space is unmapped >at run time for a QMAGIC binary, to help trap NULL pointer dereferences >(you get a segmentation violation if you try to use a NULL pointer). > >........ I sure wish GCL had the ability to create executables via a normal link instead of loading objects and unexecing! This would make GCL useful across a much wider veriety of platforms. The way it would work is that lisp code would be compiled into objects as they are now. However, those objects could then be linked with a GCL library - producing the executable. No real need for unexec or reading in objects. --blake -- Blake McBride (615) 790-8521 voice 3020 Liberty Hills Drive (615) 791-7736 fax Franklin, TN 37064 blake@edge.ercnet.com U.S.A. From kr@shell.portal.com Wed Feb 22 20:11:59 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01185; Wed, 22 Feb 95 20:11:59 CST Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA17351; Wed, 22 Feb 95 19:10:32 CST Received: from jobe.shell.portal.com (root@jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.9/8.6.5) with ESMTP id RAA09632 for ; Wed, 22 Feb 1995 17:10:30 -0800 Received: from DialupEudora (kr@jobe.shell.portal.com [156.151.3.4]) by jobe.shell.portal.com (8.6.9/8.6.5) with SMTP id RAA06238 for ; Wed, 22 Feb 1995 17:10:24 -0800 X-Sender: kr@pop.shell.portal.com Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: \p\r\r\e%, 22 Feb 1995 17:10:41 -0800 To: gcl@cli.com From: kr@shell.portal.com (kr) Subject: what is brk needed for ? With this question I am presumably displaying my lack of knowledge about UN!X systems. But I hope that getting a quick answer from the authors of gcl will save me a lot of painful time, trying to figure this out by trial and error. In c/main.c , in the function siLsave_system() , one finds the following: >#if defined(BSD) || defined(ATT) > brk(core_end); > /* printf( "(breaking at core_end = %x in main ,)",core_end); */ >#endif What is the brk needed for ? Is there a way to work around it ? This seems to be essentially the only place in the gcl source where brk is called. I would be glad to receive advice on how to eliminate this call too. I am trying to compile gcl-1.1 on a Mac, using MacMiNT, which is a UN!X-like emulator that was originally written for Atari computers and then adapted to Macs. This sounds pretty hairy (and it is :-), but usually most C code can be ported without too much trouble to this platform. So far, after fiddling around quite a bit, I have been able to compile everything, but I am now stuck at the linking stage. I have tried to adhere to a BSD system, so "BSD" is defined. Unfortunately, the mint libraries seem to be totally unaware of any brk call, they only have sbrk. I do not understand the difference between these calls, so I do not know how to work around this problem. Thank you for helpful hints. When I have finally succeeded in getting this to run, I can contribute the configuration I used for compiling under MacMiNT, if this is desired by the gcl authors or members of this mailing list. Greetings Markus Krummenacker From kr@shell.portal.com Thu Feb 23 17:33:45 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01486; Thu, 23 Feb 95 17:33:45 CST Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA19931; Thu, 23 Feb 95 15:58:34 CST Received: from jobe.shell.portal.com (root@jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.10/8.6.5) with ESMTP id NAA13210 for ; Thu, 23 Feb 1995 13:58:32 -0800 Received: from DialupEudora (kr@jobe.shell.portal.com [156.151.3.4]) by jobe.shell.portal.com (8.6.10/8.6.5) with SMTP id NAA04253 for ; Thu, 23 Feb 1995 13:58:27 -0800 X-Sender: kr@pop.shell.portal.com Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: date\r\e%, 23 Feb 1995 13:58:44 -0800 To: gcl@cli.com From: kr@shell.portal.com (kr) Subject: what is _cleanup ? Thank you for the replies I obtained for my last question. It seems that brk() can be expressed in terms of sbrk(). However, I have an additional question concerning a problem I have run into. In the file unixsave.c , in function Lsave() , there is a call to _cleanup() . I have not found this function anywhere in the gcl source, nor does MacMiNT know about it. What does it do ? Is this a UN!X function ? Can one just leave it away with no adverse consequences ? Greetings Markus Krummenacker From mjs@s4mjs.shannon.com Sun Feb 26 13:41:47 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01978; Sun, 26 Feb 95 13:41:47 CST Received: from cnj.digex.net by cli.com (4.1/SMI-4.1) id AA26331; Sun, 26 Feb 95 12:41:24 CST Received: from shannon.com by cnj.digex.net with SMTP id AA00698 (5.67b8/IDA-1.5 for ); Sun, 26 Feb 1995 13:41:52 -0500 Received: by shannon.com (5.x/SMI-SVR4) id AA18818; Sun, 26 Feb 1995 13:41:48 -0500 Date: Sun, 26 Feb 1995 13:41:48 -0500 From: mjs@s4mjs.shannon.com (marty shannon) Message-Id: <9502261841.AA18818@shannon.com> To: gcl@cli.com Subject: gcl on Solaris 2.4 for x86 Reply-To: mjs@s4mjs.shannon.com X-Sun-Charset: US-ASCII Hi, I've been trying to get gcl-1.1 to build under Solaris 2.4 (i.e., SunOS 5.4) on a 486, and been having a devil of a time even building raw_gcl. One of the problems with the current method of configuring gcl is that a system type of, say, solaris, implies both hardware and software, even though solaris runs on wildly different hardware (sparc vs. i386/i486). It would be better if the vast majority of parameters were determined by the presense or absense of features rather than by the current method. Of course, if/as/when I get a working version, I'll send my code to Bill Schelter to be folded in to the distribution, but anyone who's more familiar with the porting issues than I am, and is willing to look at what I've done so far is more than welcome to contact me (mjs@shannon.com, in case the mail headers aren't quite right...). Marty -- ---------------+ Marty Shannon | My opinions are just that. You may share them. No one mjs@shannon.com| speaks for me, and I speak only for myself -- no matter ---------------+ where I post from. Get it? Post no flames. From DOUGHILL@BNLCL6.BNL.GOV Sun Feb 26 14:41:53 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01985; Sun, 26 Feb 95 14:41:53 CST Received: from BNLCL6.BNL.GOV by cli.com (4.1/SMI-4.1) id AA26403; Sun, 26 Feb 95 13:56:01 CST Date: Sun, 26 Feb 1995 14:56:31 -0500 (EST) From: DOUGHILL@BNLCL6.BNL.GOV To: gcl@cli.com Message-Id: <950226145631.24c41f17@BNLCL6.BNL.GOV> Subject: A Fix for GCL unexec() problem with Linux *** OUTLINE OF A FIX FOR THE GCL UNEXEC() PROBLEM ON LINUX *** SIMPTOM: GCL bombs during build. gcl-1.1 (also gcl-1.0) generates a segmentation fault during the dumping phase of the build on Slackware Linux 1.0.9. It occurs during the call to unexec(). I found a remedy by copying stuff from the current GNU emacs-19.28 (which also dumps). Take emacs/src/unexec.c, which is a somewhat modified version of unexlin.c (i. e. same author: Spencer W. Thomas). Define these macros: /* These 3 from emacs/s rc/s/linux.h */ #define A_TEXT_OFFSET(hdr) (N_MAGIC(hdr) == QMAGIC ? sizeof (struct exec) : 0) #define A_TEXT_SEEK(hdr) (N_TXTOFF(hdr) + A_TEXT_OFFSET(hdr)) #define ADJUST_EXEC_HEADER \ unexec_text_start = N_TXTADDR(ohdr) + A_TEXT_OFFSET(ohdr) Using the preprocessor I determined start_of_text () and start_of_data () to be: /* * These two functions lifted from emacs/src/sysdep.c * * Return the address of the start of the text segment prior to * doing an unexec. After unexec the return value is undefined. * See crt0.c for further explanation and _start. * */ char * start_of_text () { return ((char *) 0 ); } /* * Return the address of the start of the data segment prior to * doing an unexec. After unexec the return value is undefined. * See crt0.c for further information and definition of data_start. * * Apparently, on BSD systems this is etext at startup. On * USG systems (swapping) this is highly mmu dependent and * is also dependent on whether or not the program is running * with shared text. Generally there is a (possibly large) * gap between end of text and start of data with shared text. * * On Uniplus+ systems with shared text, data starts at a * fixed address. Each port (from a given oem) is generally * different, and the specific value of the start of data can * be obtained via the UniPlus+ specific "uvar" system call, * however the method outlined in crt0.c seems to be more portable. * * Probably what will have to happen when a USG unexec is available, * at l east on UniPlus, is temacs will have to be made unshared so * that text and data are contiguous. Then once loadup is complete, * unexec will produce a shared executable where the data can be * at the normal shared text boundry and the startofdata variable * will be patched by unexec to the correct value. * */ char * start_of_data () { extern int data_start; return ((char *) &data_start); } In unixport/makefile: # Put in stuff from emacs: crt0_stuff = $(ODIR)/pre-crt0.o /usr/lib/crt0.o raw_ $(SYSTEM): $(crt0_stuff) $(FOR_RAW) $(LDCC) -nostartfiles $(crt0_stuff) -o raw_$(SYSTEM) $(FOR_RAW) $(GCLIB) -lm It might be more portable to let the compiler pick its own libraries like this. From simons@cs.tu-berlin.de Thu Mar 2 02:49:45 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02816; Thu, 2 Mar 95 02:49:45 CST Received: from mail.cs.tu-berlin.de by cli.com (4.1/SMI-4.1) id AA07257; Thu, 2 Mar 95 01:48:08 CST Received: from wintermute.cs.tu-berlin.de (simons@wintermute.cs.tu-berlin.de [130.149.19.141]) by mail.cs.tu-berlin.de (8.6.10/8.6.10) with ESMTP id IAA19813; Thu, 2 Mar 1995 08:48:34 +0100 Received: (simons@localhost) by wintermute.cs.tu-berlin.de (8.6.10/8.6.9) id IAA03894; Thu, 2 Mar 1995 08:48:25 +0100 Date: Thu, 2 Mar 1995 08:48:25 +0100 Message-Id: <199503020748.IAA03894@wintermute.cs.tu-berlin.de> From: Martin Simons To: gcl@cli.com Subject: GCL 1.1 and NESL Content-Type: text Content-Length: 1215 Hi, I am having troubles with GCL 1.1 while compiling CMU's NESL 3.0 under SunOS5.4; everything works fine with SunOS4.1.3. After adding the definitions for Solaris (./add-defs solaris) GCL 1.1 compiles with a few warnings. I am using gcc2.5.8. Also, compiling NESL works well until I am asked to call the main function (nesl) form the GCL prompt. GCL then quits with calling `dl' function sun did not supply This error message originates from GCL's c/run_process.c where some wrappers for `dl' functions are provided. We actually do have the libdl library but naively commenting out the wrappers and compiling gcl with -ldl with dynamic loading of libraries enabled produced a saved_gcl that cored immediately. I don't want to dive any deeper into the gcl code and I would appreciate any help. Thanks and best regards, Martin Simons ------------------------------------------------------------------------- Martin Simons Internet: simons@cs.tu-berlin.de Software Engineering Research Group (Sekr. FR5-6) Tel: +49 30 314 25213 Fax: +49 30 314 73488 Technical University Berlin - Franklinstr. 28/29 - 10587 Berlin - Germany From PBC01764@niftyserve.or.jp Mon Mar 6 05:19:57 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01002; Mon, 6 Mar 95 05:19:57 CST Received: from inetnif.niftyserve.or.jp ([192.47.24.129]) by cli.com (4.1/SMI-4.1) id AA16804; Mon, 6 Mar 95 04:19:55 CST Received: by inetnif.niftyserve.or.jp (8.6.9+2.4W/3.3W8-950117-Mail-Gateway) id TAA06257; Mon, 6 Mar 1995 19:20:55 +0900 Message-Id: <199503061020.TAA06257@inetnif.niftyserve.or.jp> Date: Mon, 06 Mar 1995 19:19:00 +0900 From: Mikio Nakajima Subject: gcl-1.1 on MSDOS (djgpp) To: gcl@cli.com I got the gcl-1.1 from ftp.cli.com, tried to compile it on MSDOS + djgpp- 1.12, but I cannot finish it even now :-(. First of all, I found add-defs.bat did not work on MSDOS ver.5.0. For Japanese domestic machines, MSDOS ver.6 has been receltly available, and I don't have it now. I'd like to know what version's MSDOS does add-defs.bat support. Secondly, unixport\saved_kc.exe was not included in the archive. I am really anxious about the following message created by add-defs.bat. WARNING : unixport/saved_kcl.exe file not found _ you will not be able to recompile the .lsp files _ nor start akcl I'd like to know where to get saved_kcl.exe. Thirdly, I found out neither Smakefile nor tmpxx_.tem in the archive. I guessed Smakefile was provided as makefile, but I really didn't found out a relief for tmpxx_.tem. I'd like to know where to or how to get them. Finally, it seemed that h\dos-go32.h included . Was this dos-go32.h for djgpp-1.10 or earlier? If someone provides dos-go32.h for djgpp-1.11 or later, please tell me where to get it. Mikio Nakajima (PBC01764@niftyserve.or.jp / minakaji@mix.or.jp) From jeff@aiai.ed.ac.uk Fri Mar 24 17:07:26 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00485; Fri, 24 Mar 95 17:07:26 CST Received: from haymarket.ed.ac.uk ([129.215.128.53]) by cli.com (4.1/SMI-4.1) id AA19847; Fri, 24 Mar 95 15:36:12 CST Received: from festival.ed.ac.uk (festival.ed.ac.uk [129.215.128.24]) by haymarket.ed.ac.uk (8.6.10/8.6.10) with SMTP id VAA03612 for ; Fri, 24 Mar 1995 21:36:04 GMT Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa08313; 24 Mar 95 21:34 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Fri, 24 Mar 95 21:34:39 GMT Date: Fri, 24 Mar 95 21:34:39 GMT Message-Id: <4021.9503242134@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: possible bug w/ 0d0 To: gcl@cli.com Compiling a file that contains 0d0 and then loading the ,o file gets "Dots appeared illegally". This is running on a 386 using FreeBSD 2.0. GCL 1.1 doesn't come w/. such a port -- I just used the NetBSD stuff unchanged -- so it's possible the bug is there rather than in GCL per se. $ cat mess.lsp (defun df-zero () 0d0) $ gcl GCL (GNU Common Lisp) Version(1.1) Fri Mar 24 19:33:57 1995 Licensed under GNU Public Library License Contains Enhancements by W. Schelter >(compile-file "mess.lsp") Compiling mess.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=0 (No runtime error checking), Space=0, Speed=3 Finished compiling mess.lsp. #"mess.o" >(load *) Loading mess.o Error: Dots appeared illegally. Fast links are on: do (use-fast-links nil) for debugging Error signalled by LOAD. Broken at LOAD. Type :H for Help. >> -- jeff From toy@rtp.ericsson.se Fri Mar 24 17:23:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00489; Fri, 24 Mar 95 17:23:56 CST Received: from gwa.ericsson.com by cli.com (4.1/SMI-4.1) id AA19926; Fri, 24 Mar 95 16:08:16 CST Received: from mr2.exu.ericsson.se (mr2.exu.ericsson.com [138.85.147.12]) by gwa.ericsson.com (8.6.10/8.6.10) with ESMTP id QAA19987 for ; Fri, 24 Mar 1995 16:08:15 -0600 Received: from screamer.rtp.ericsson.se (screamer.rtp.ericsson.se [147.117.133.13]) by mr2.exu.ericsson.se (8.6.8/NAHUB-MR1.1) with ESMTP id QAA13099 for ; Fri, 24 Mar 1995 16:08:14 -0600 Received: from rcur (rcur7.rtp.ericsson.se [147.117.133.38]) by screamer.rtp.ericsson.se (8.6.8/8.6.4) with ESMTP id RAA10673; Fri, 24 Mar 1995 17:11:20 -0500 Message-Id: <199503242211.RAA10673@screamer.rtp.ericsson.se> To: gcl@cli.com Cc: toy@rtp.ericsson.se Subject: reduce with :key, and map-into Date: Fri, 24 Mar 1995 17:08:06 -0500 From: Raymond Toy I'm new to this list, so forgive if this has already been fixed. In gcl 1.x and gcl 2.0, reduce does not accept the :key keyword argument, as specified in cltl2. Also, map-into is not defined at all. I've enclosed a modified version of reduce (from seqlib.lsp) that accepts :key and a new version of map-into. My lisp knowledge is limited, so these may not be the best versions. Ray ====================================================================== -----> Raymond Toy Tel: 919-990-7480 Ericsson Inc. Fax: 919-990-7451 1 Triangle Drive E-mail: toy@rtp.ericsson.se RTP, NC 27709 (defun reduce (function sequence &key from-end start end (initial-value nil ivsp) (key #'identity)) (with-start-end start end sequence (cond ((not from-end) (when (null ivsp) (when (>= start end) (return-from reduce (funcall function))) (setq initial-value (elt sequence start)) (setf start (f+ 1 start)) ) (do ((x initial-value (funcall function x (funcall key (prog1 (elt sequence start) (setf start (f+ 1 start)) ))))) ((>= start end) x))) (t (when (null ivsp) (when (>= start end) (return-from reduce (funcall function))) (setf end (f+ end -1)) (setq initial-value (elt sequence end))) (do ((x initial-value (funcall function (funcall key (elt sequence end)) x))) ((>= start end) x) (setf end (f+ -1 end))))))) (defun map-into (result-sequence function &rest sequences) "map-into: (result-sequence function &rest sequences) Destructively modify RESULT-SEQUENCE to contain the results of applying FUNCTION to corresponding elements of the argument SEQUENCES in turn. RESULT-SEQUENCE is returned. The arguments RESULT-SEQUENCE and each element of SEQUENCES can each be either a list or a vector. The FUNCTION must accept at least as many arguments as the number of arguement SEQUENCES supplied to MAP-INTO. If RESULT-SEQUENCE and the other argument SEQUENCES are not all the same length, the iteration terminates when the shortest sequence is exhausted. If RESULT-SEQUENCE is a vector with a fill pointer, the fill pointer is ignored when deciding how many iterations to perform, and afterwards the fill pointer is set to the number of times the FUNCTION was applied. If the FUNCTION has side effects, it can count on being called first on all the elements numbered 0, then on all those numbered 1, and so on. If RESULT-SEQUENCE is longer than the shortest element of SEQUENCES, extra elements at the end of RESULT-SEQUENCE are unchanged. The function MAP-INTO differs from MAP in that it modifies an existing sequence rather than creating a new one. In addition, MAP-INTO can be called with only two arguments (RESULT-SEQUENCE and FUNCTION), while MAP requires at least three arguments. If RESULT-SEQUENCE is NIL, MAP-INTO immediately returns NIL, becaause NIL is a sequence of length zero." (let ((nel (apply #'min (if (eq 'vector (type-of result-sequence)) (array-dimension result-sequence 0) (length result-sequence)) (mapcar #'length sequences)))) ;; Set the fill pointer to the number of iterations (when (and (eq 'vector (type-of result-sequence)) (array-has-fill-pointer-p result-sequence)) (setf (fill-pointer result-sequence) nel)) ;; Perform mapping (dotimes (k nel result-sequence) (setf (elt result-sequence k) (apply function (mapcar #'(lambda (v) (elt v k)) sequences)))))) From pw@snoopy.mv.com Fri Mar 24 17:27:05 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00493; Fri, 24 Mar 95 17:27:05 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA19923; Fri, 24 Mar 95 16:06:51 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id RAA01276; Fri, 24 Mar 1995 17:03:07 -0500 Date: Fri, 24 Mar 1995 17:03:07 -0500 From: "Paul F. Werkowski" Message-Id: <199503242203.RAA01276@snoopy.mv.com> To: jeff@aiai.ed.ac.uk Cc: gcl@cli.com In-Reply-To: <4021.9503242134@subnode.aiai.ed.ac.uk> (message from Jeff Dalton on Fri, 24 Mar 95 21:34:39 GMT) Subject: Re: possible bug w/ 0d0 >>>>> "Jeff" == Jeff Dalton writes: Jeff> Compiling a file that contains 0d0 and then loading the ,o Jeff> file gets "Dots appeared illegally". Jeff> Error: Dots appeared illegally. Fast links are on: do Jeff> (use-fast-links nil) for debugging Error signalled by LOAD. Jeff> Broken at LOAD. Type :H for Help. >>> This is a known problem with the 2.0 vfprintf.c function in libc. It's been fixed in -current since mid January. Paul From leverich@rand.org Fri Mar 24 19:39:49 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00535; Fri, 24 Mar 95 19:39:49 CST Received: from rand.org by cli.com (4.1/SMI-4.1) id AA20202; Fri, 24 Mar 95 18:46:30 CST Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.10/8.6.10) with ESMTP id QAA15341; Fri, 24 Mar 1995 16:44:49 -0800 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.10/8.6.10) with ESMTP id QAA21382; Fri, 24 Mar 1995 16:44:54 -0800 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.10/8.6.9) with SMTP id QAA14580; Fri, 24 Mar 1995 16:44:52 -0800 Message-Id: <199503250044.QAA14580@atlantis.rand.org> To: "Paul F. Werkowski" Cc: jeff@aiai.ed.ac.uk, gcl@cli.com, Brian_Leverich@rand.org Reply-To: Brian_Leverich@rand.org Subject: Re: possible bug w/ 0d0 In-Reply-To: Your message of Fri, 24 Mar 95 17:03:07 EST. <199503242203.RAA01276@snoopy.mv.com> Date: Fri, 24 Mar 95 16:44:49 PST From: Brian Leverich -- Your message was: (from ""Paul F. Werkowski"") This is a known problem with the 2.0 vfprintf.c function in libc. It's been fixed in -current since mid January. Paul ------------------ Oops. I've just had one of those self-revelatory experiences where I've become aware of my own extreme cluelessness. Where are 2.0 and -current? And is there another list where gcl is being discussed besides here? BTW, if anyone else is interested in such self-abuse, there are versions of CLX and Garnet 2.2 hacked for GCL compatibility available from ftp://rand.org/pub/leverich/. (We're running GCL/CLX/Garnet on Linux boxes and HP 735s here at RAND.) We're supporting these sources in rude mode, which means we don't support them at all but we will incorporate contributed fixes into future versions. Cheers, B. From kr@shell.portal.com Sat Mar 25 06:39:58 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00617; Sat, 25 Mar 95 06:39:58 CST Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA20989; Sat, 25 Mar 95 05:12:19 CST Received: from jobe.shell.portal.com (jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.11/8.6.5) with ESMTP id DAA19500; Sat, 25 Mar 1995 03:11:43 -0800 Received: from DialupEudora (jobe.shell.portal.com [156.151.3.4]) by jobe.shell.portal.com (8.6.11/8.6.5) with SMTP id DAA25555; Sat, 25 Mar 1995 03:11:39 -0800 X-Sender: kr@pop.shell.portal.com Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: srialpop\r\e%, 25 Mar 1995 03:12:05 -0800 To: Raymond Toy , gcl@cli.com From: kr@shell.portal.com (kr) Subject: Re: reduce with :key, and map-into Cc: toy@rtp.ericsson.se At 17:08 95-03-24, Raymond Toy wrote: >I'm new to this list, so forgive if this has already been fixed. > >In gcl 1.x and gcl 2.0, reduce does not accept the :key keyword What exactly is gcl 2.0 ? Having been on this list for a couple of weeks, I have not noticed an announcement float by. Have I missed something ? Greetings Markus Krummenacker From pw@snoopy.mv.com Sat Mar 25 08:51:24 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00627; Sat, 25 Mar 95 08:51:24 CST Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA21086; Sat, 25 Mar 95 07:25:00 CST Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id IAA02697; Sat, 25 Mar 1995 08:18:19 -0500 Date: Sat, 25 Mar 1995 08:18:19 -0500 From: "Paul F. Werkowski" Message-Id: <199503251318.IAA02697@snoopy.mv.com> To: Brian_Leverich@rand.org Cc: gcl@cli.com In-Reply-To: <199503250044.QAA14580@atlantis.rand.org> (message from Brian Leverich on Fri, 24 Mar 95 16:44:49 PST) Subject: Re: possible bug w/ 0d0 >>>>> "Brian" == Brian Leverich writes: Brian> -- Your message was: (from ""Paul F. Werkowski"") Brian> This is a known problem with the 2.0 vfprintf.c Brian> function in libc. It's been fixed in -current since mid Brian> January. Brian> Oops. I've just had one of those self-revelatory Brian> experiences where I've become aware of my own extreme Brian> cluelessness. Brian> Where are 2.0 and -current? And is there another list Brian> where gcl is being discussed besides here? Sorry, I had been reading zillions of messages from the "freebsd-hackers" list where those terms are part of the lingo. 2.0 refers to FreeBSD 2.0, part of the subject in the message I was replying to. -current refers to the current state of the FreeBSD development tree which will at some point become FreeBSD 2.1(RELEASE). There is no specific gcl discussion in any of the FreeBSD mailing lists that I know of but there is a "port" of gcl-1.1 in the tree that will appear on the 2.1 CDROM. A "port" in this context is a node in the distribution where one can simply say "make" and have all the right things happen to cause the underlying application to be fetched, patched, compiled and installed. The FreeBSD gcl port adds files FreeBSD.defs and FreeBSD.h that have been tuned to the -current state of things and also arranges for the correct "unexec" variant of the system saver to be present. Having written all that I now notice that you asked "where", not "what". freebsd.cdrom.com:/pub/FreeBSD/FreeBSD-current http://www.freebsd.org -- (let ((forty++ "Sonetech Corporation") (product "Real-Time Multi-Media Recognition Systems") (IP-access :self) (OS-of-choice '|FreeBSD|) ...) From jeff@aiai.ed.ac.uk Sat Mar 25 11:00:01 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00662; Sat, 25 Mar 95 11:00:01 CST Received: from haymarket.ed.ac.uk by cli.com (4.1/SMI-4.1) id AA21227; Sat, 25 Mar 95 09:58:43 CST Received: from festival.ed.ac.uk (festival.ed.ac.uk [129.215.128.24]) by haymarket.ed.ac.uk (8.6.10/8.6.10) with SMTP id PAA08315; Sat, 25 Mar 1995 15:57:16 GMT Received: from skye-alter.aiai.ed.ac.uk by festival.ed.ac.uk id aa07751; 25 Mar 95 15:55 GMT Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Sat, 25 Mar 95 15:55:25 GMT Date: Sat, 25 Mar 95 15:55:25 GMT Message-Id: <4643.9503251555@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: Re: possible bug w/ 0d0 To: Brian_Leverich@rand.org, "Paul F. Werkowski" Cc: gcl@cli.com In-Reply-To: Brian Leverich's message of Fri, 24 Mar 95 16:44:49 PST > -- Your message was: (from ""Paul F. Werkowski"") > > This is a known problem with the 2.0 vfprintf.c function > in libc. It's been fixed in -current since mid January. > > Oops. I've just had one of those self-revelatory experiences where > I've become aware of my own extreme cluelessness. > > Where are 2.0 and -current? And is there another list where gcl is > being discussed besides here? Fear not! 2.0 and -current are versions of FreeBSD, and the "bug" turns out to be in FreeBSD rather than GCL. Indeed, I should have remembered this problem, because I've had to deal with it before. When GCL prints a floating-point number, it uses the C routine sprintf to convert it to a string and then edits the string. And this editing sassumes the string has a format that it turns out not to have for numbers such as 1.0 and 100.0. Other numbers, e.g. PI, are fine. Anyway, thanks to all who replied! -- jeff From lnm@stsn.dga.fr Mon Mar 27 06:18:08 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00931; Mon, 27 Mar 95 06:18:08 CST Received: from asterix.stsn.dga.fr (mailhost.stsn.dga.fr) by cli.com (4.1/SMI-4.1) id AA24003; Mon, 27 Mar 95 04:57:00 CST Received: (from lnm@localhost) by asterix.stsn.dga.fr (8.6.9/8.6.9) id LAA00383; Mon, 27 Mar 1995 11:40:56 +0100 Date: Mon, 27 Mar 1995 11:40:56 +0100 From: LIGNON Marc Message-Id: <199503271040.LAA00383@asterix.stsn.dga.fr> To: gcl@cli.com Subject: Pb with Xgcl on solaris I want to use Xgcl-2 on both sun4 and solaris. I built it properly for sun4 but I'm having some pbs with solaris: (gcl already works OK on solaris) Building it straight ahead I get this: Undefined first referenced symbol in file authdes_create /usr/lib/libX11.a(XConnDis.o) ld: fatal: Symbol referencing errors. No output written to raw_xgcl The symbol "authdes_create" is defined at libnsl.a which is present in one of the make variables "*LIBS*" of GCLDIR/unixport, and is present on the gcc line to build "raw_xgcl". The pb is that it figures before -lX11. If we put -lnsl after -lX11 the following appears: ld: fatal: symbol `shifts' is multiply defined: (file ../mp/libmport.a(mp2.o) and file /usr/lib/libnsl.a(des_soft.o)); ld: fatal: File processing errors. No output written to raw_xgcl Does anybody already had this problem, and knows how to overpass ? Thaks for your attention, Best regards and A bientot, Marc LIGNON ________________________________________________________________________ DGA/DCN/ST/DA/S/ES lignon@stsn.dga.fr 8, boulevard Victor tel: (33)(1) 40.59.13.14 75015 - PARIS - FRANCE fax: (33)(1) 45.79.20.40 From novak@cs.utexas.edu Mon Mar 27 14:53:10 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01138; Mon, 27 Mar 95 14:53:10 CST Received: from mail.cs.utexas.edu by cli.com (4.1/SMI-4.1) id AA25450; Mon, 27 Mar 95 13:44:18 CST Received: from ai.cs.utexas.edu (novak@ai.cs.utexas.edu [128.83.120.3]) by mail.cs.utexas.edu (8.6.10/8.6.9) with ESMTP id NAA09179; Mon, 27 Mar 1995 13:44:17 -0600 Received: by ai.cs.utexas.edu (8.6.10/Client-v1.4) id NAA11217; Mon, 27 Mar 1995 13:44:15 -0600 Message-Id: <199503271944.NAA11217@ai.cs.utexas.edu> From: novak@cs.utexas.edu (Gordon Shaw Novak Jr.) Date: Mon, 27 Mar 1995 13:44:15 -0600 X-Mailer: Mail User's Shell (7.2.5 10/14/92) To: gcl@cli.com Subject: XGCL available Cc: novak@cs.utexas.edu, boyer@cli.com Xgcl software is now available; it provides an interface to X windows from GCL (Gnu Common Lisp, formerly Austin Kyoto Common Lisp or AKCL). There is a low-level interface to the Xlib routines, and there is an easy-to-use interface that provides graphics, menus, and mouse interaction via functions that are called from Lisp. Xgcl is built on top of GCL, and it is somewhat larger (e.g. 6.7 MB for Xgcl vs. 4.9 MB for GCL) because it incorporates Xlib. To make Xgcl, you must first obtain and make GCL. The code was written by Gordon Novak, Hiep Nguyen, and William Schelter. The file is called xgcl-2.tgz . Xgcl and GCL can be FTP'd from: math.utexas.edu pub/gcl ftp.cli.com pub/gcl ftp.cs.utexas.edu pub/novak/xgcl (Xgcl file only) http://www.cs.utexas.edu/users/novak (Xgcl file only) Put the xgcl-2.tgz file in the same directory as gcl-1.1.tgz . Uncompress it with: gzip -dc xgcl-2.tgz | tar xvf - Then see the README in the directory xgcl-2. Note: the date of the xgcl-2.tgz should be 3/15/95 or later. From boyer@cli.com Mon Mar 27 16:17:19 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01168; Mon, 27 Mar 95 16:17:19 CST Received: from rita.cli.com by cli.com (4.1/SMI-4.1) id AA25658; Mon, 27 Mar 95 14:50:32 CST From: boyer@cli.com (Robert S. Boyer) Received: by rita.cli.com (4.1) id AA08369; Mon, 27 Mar 95 14:50:31 CST Date: Mon, 27 Mar 95 14:50:31 CST Message-Id: <9503272050.AA08369@rita.cli.com> To: gcl@cli.com Cc: novak@cs.utexas.edu Subject: Re: XGCL available I didn't have anything to do with Xgcl, but I have seen it operate and I think I should mention that Xgcl is an extremely fast and efficient way to communicate with X from GCL. I believe some folks have found GCL or (A)KCL, when used with other interfaces to Xlib, to be rather sluggish. I suspect this was due to the absence of sufficient attention to the declaration of arithmetic and array operations and function proclamations, something that can make a two orders of magnitude difference in the execution of KCL/GCL programs. For others who are as ignorant as me about X windows, let me make the obvious remark that Xgcl is relatively "low level" because it connects the programmer to the Xlib level of X, a level at which you can draw your own lines and place your characters at absolute locations of your choice, and so forth. This is in marked contrast to the TK interface that Bill Schelter is providing, where a plethora of higher level graphical objects are supported. The point is that both relatively high and low level X graphics facilities are becoming available to the GCL programmer. Bob From leverich@rand.org Mon Mar 27 18:14:49 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01183; Mon, 27 Mar 95 18:14:49 CST Received: from rand.org by cli.com (4.1/SMI-4.1) id AA26064; Mon, 27 Mar 95 16:57:20 CST Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.10/8.6.10) with ESMTP id OAA15969; Mon, 27 Mar 1995 14:57:18 -0800 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.10/8.6.10) with ESMTP id OAA02633; Mon, 27 Mar 1995 14:57:17 -0800 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.10/8.6.9) with SMTP id OAA18554; Mon, 27 Mar 1995 14:57:16 -0800 Message-Id: <199503272257.OAA18554@atlantis.rand.org> To: boyer@cli.com (Robert S. Boyer) Cc: gcl@cli.com, novak@cs.utexas.edu, Brian_Leverich@rand.org Reply-To: Brian_Leverich@rand.org Subject: Re: XGCL available In-Reply-To: Your message of Mon, 27 Mar 95 14:50:31 CST. <9503272050.AA08369@rita.cli.com> Date: Mon, 27 Mar 95 14:57:14 PST From: Brian Leverich Anyone know off the top of their head how much is XGCL like CLX at the Lisp funtional interface level? Cheers, B From novak@cs.utexas.edu Mon Mar 27 18:51:30 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01190; Mon, 27 Mar 95 18:51:30 CST Received: from mail.cs.utexas.edu by cli.com (4.1/SMI-4.1) id AA26100; Mon, 27 Mar 95 17:01:59 CST Received: from ai.cs.utexas.edu (novak@ai.cs.utexas.edu [128.83.120.3]) by mail.cs.utexas.edu (8.6.10/8.6.9) with ESMTP id RAA23926; Mon, 27 Mar 1995 17:01:57 -0600 Received: by ai.cs.utexas.edu (8.6.10/Client-v1.4) id RAA11686; Mon, 27 Mar 1995 17:01:55 -0600 Message-Id: <199503272301.RAA11686@ai.cs.utexas.edu> From: novak@cs.utexas.edu (Gordon Shaw Novak Jr.) Date: Mon, 27 Mar 1995 17:01:55 -0600 X-Mailer: Mail User's Shell (7.2.5 10/14/92) To: Brian_Leverich@rand.org, boyer@cli.com (Robert S. Boyer) Subject: Re: XGCL available Cc: gcl@cli.com, novak@cs.utexas.edu CLX and Xlib have very similar interfaces. My guess is that CLX is a "copy" of Xlib except that it is written in Lisp rather than C. (This does not mean that the actual calling sequences are identical, but conceptually they are the same.) CLX is a lot bigger and a lot slower than Xlib in my experience. And XGCL provides a higher-level interface than either CLX or Xlib. Regards, Gordon From leverich@rand.org Mon Mar 27 19:15:25 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01195; Mon, 27 Mar 95 19:15:25 CST Received: from rand.org by cli.com (4.1/SMI-4.1) id AA26201; Mon, 27 Mar 95 17:47:14 CST Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.10/8.6.10) with ESMTP id PAA18285; Mon, 27 Mar 1995 15:47:09 -0800 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.10/8.6.10) with ESMTP id PAA07937; Mon, 27 Mar 1995 15:47:08 -0800 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.10/8.6.9) with SMTP id PAA18645; Mon, 27 Mar 1995 15:47:07 -0800 Message-Id: <199503272347.PAA18645@atlantis.rand.org> To: novak@cs.utexas.edu (Gordon Shaw Novak Jr.) Cc: boyer@cli.com (Robert S. Boyer), gcl@cli.com, Brian_Leverich@rand.org Reply-To: Brian_Leverich@rand.org Subject: Re: XGCL available In-Reply-To: Your message of Mon, 27 Mar 95 17:01:55 CST. <199503272301.RAA11686@ai.cs.utexas.edu> Date: Mon, 27 Mar 95 15:47:05 PST From: Brian Leverich -- Your message was: (from "novak@cs.utexas.edu") CLX and Xlib have very similar interfaces. My guess is that CLX is a "copy" of Xlib except that it is written in Lisp rather than C. (This does not mean that the actual calling sequences are identical, but conceptually they are the same.) CLX is a lot bigger and a lot slower than Xlib in my experience. And XGCL provides a higher-level interface than either CLX or Xlib. Regards, Gordon ------------------ Bummer. I was hoping you were going to say that you'd taken advantage of GCL's C-friendliness to reimplement CLX as a set of direct calls to Xlib rather than emulating Xlib's functionality in Lisp. I could have used that immediately, because I have a family of apps that are CLX-dependent. I'll look at XGCL the next time I start a project from scratch. Cheers, B. From lnm@stsn.dga.fr Tue Mar 28 10:32:13 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01332; Tue, 28 Mar 95 10:32:13 CST Received: from asterix.stsn.dga.fr (mailhost.stsn.dga.fr) by cli.com (4.1/SMI-4.1) id AA27476; Tue, 28 Mar 95 09:11:03 CST Received: (from lnm@localhost) by asterix.stsn.dga.fr (8.6.9/8.6.9) id RAA02287; Tue, 28 Mar 1995 17:10:31 +0100 Date: Tue, 28 Mar 1995 17:10:31 +0100 From: LIGNON Marc Message-Id: <199503281610.RAA02287@asterix.stsn.dga.fr> To: gcl@cli.com Subject: Re: Pb with Xgcl on solaris > My guess is that you could remake putting in h/solaris.h > #define shifts shifts_p > to avoid the name conflict, with the definition in libnsl.a Less elegant, but I tried renaming all "shifts" in GCL and it overpassed THAT step all right... But trying it out I got: client13:/commun/bdt/src/Trajecto 501 > Xgcl GCL (GNU Common Lisp) Version(1.1) Mon Mar 27 14:46:37 MET DST 1995 Licensed under GNU Public Library License Contains Enhancements by W. Schelter Loading init.lsp Finished loading init.lsp >(in-package "XLIB") #<"XLIB" package> XLIB>(Xinit) calling 'dl' function sun did not supply..exitting client13:/commun/bdt/src/Trajecto 502 > In file run_process.c at line 300, we find : #ifdef __svr4__ getpagesize() { return PAGESIZE; } dlclose() {fprintf(stderr,"calling 'dl' function sun did not supply..exitting") ;exit(1);} dgettext() {dlclose();} dlopen() {dlclose();} dlerror() {dlclose();} dlsym() {dlclose();} It concerns dynamic linking... but the image was compiled with "gcc -static" (=> static linking), so I don't understand what's happening... Moreover, I found problems interfacing C and lisp at GCL level which is the subject of a new mail : Pb interfacing C and Lisp on solaris A bientot, Marc LIGNON ________________________________________________________________________ DGA/DCN/ST/DA/S/ES lignon@stsn.dga.fr 8, boulevard Victor tel: (33)(1) 40.59.13.14 75015 - PARIS - FRANCE fax: (33)(1) 45.79.20.40 From lnm@stsn.dga.fr Tue Mar 28 10:36:42 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01336; Tue, 28 Mar 95 10:36:42 CST Received: from asterix.stsn.dga.fr (mailhost.stsn.dga.fr) by cli.com (4.1/SMI-4.1) id AA27497; Tue, 28 Mar 95 09:19:39 CST Received: (from lnm@localhost) by asterix.stsn.dga.fr (8.6.9/8.6.9) id RAA02302; Tue, 28 Mar 1995 17:19:17 +0100 Date: Tue, 28 Mar 1995 17:19:17 +0100 From: LIGNON Marc Message-Id: <199503281619.RAA02302@asterix.stsn.dga.fr> To: gcl@cli.com In-Reply-To: <9503272209.AA11119@nicolas.ma.utexas.edu> (wfs@fireant.ma.utexas.edu) Subject: Pb interfacing C and Lisp on solaris Again solaris... The basic "silly" exemple from the manual which works perfectly on a sun4 (Sparc2) doesn't on a solaris (Sparc10): $cat silly.lsp (defCfun "silly(x) int x;" 0 % int y; ((+ (int x) (int "100")) (int y)) % printf("\n%d", y); % y = y+100; (print (int y)) (print (int "x+100")) % return(y); ) (defentry foo (int) (int silly)) $gcl GCL (GNU Common Lisp) Version(1.1) Tue Mar 28 15:45:26 MET DST 1995 Licensed under GNU Public Library License Contains Enhancements by W. Schelter Loading init.lsp Finished loading init.lsp >(compile-file "silly.lsp") Compiling silly.lsp. End of Pass 1. End of Pass 2. OPTIMIZE levels: Safety=1 (No runtime error checking), Space=0, Speed=3 Finished compiling silly.lsp. #"silly.o" >(load "silly.o") Loading silly.o [unknown global sym silly]start address -T 2a7c28 Finished loading silly.o 608 >(foo 3) Segmentation Fault - core dumped What we see when we load it is : [unknown global sym silly] So it seams to be a linking problem on solaris. Does anyone (perhaps Bill is better placed) have an idea where to surch to repair this pb? Thanx and A bientot, Marc LIGNON ________________________________________________________________________ DGA/DCN/ST/DA/S/ES lignon@stsn.dga.fr 8, boulevard Victor tel: (33)(1) 40.59.13.14 75015 - PARIS - FRANCE fax: (33)(1) 45.79.20.40 From wfs@math.utexas.edu Tue Mar 28 11:00:27 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01341; Tue, 28 Mar 95 11:00:27 CST Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA27606; Tue, 28 Mar 95 09:54:37 CST Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA25554; Tue, 28 Mar 95 09:53:47 -0600 Date: Tue, 28 Mar 95 09:53:38 -0600 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Tue, 28 Mar 95 09:53:38 -0600 Message-Id: <9503281553.AA11918@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA11918; Tue, 28 Mar 95 09:53:38 -0600 To: gcl@cli.com Cc: lnm@stsn.dga.fr In-Reply-To: LIGNON Marc's message of Tue, 28 Mar 1995 17:10:31 +0100 <199503281610.RAA02287@asterix.stsn.dga.fr> Subject: Re: Pb with Xgcl on solaris Reply-To: wfs@math.utexas.edu (defCfun "silly(x) int x;" 0 % int y; ((+ (int x) (int "100")) (int y)) % printf("\n%d", y); % y = y+100; (print (int y)) (print (int "x+100")) you might try (defCfun "static int silly(x) int x;" 0 ...) The loader is really designed for the output from the compiler. That has all functions being static (ie definitions not visible outside) the file. In fact you cant reference them outside the file, even if you make them global, since they wont be in the main symbol table, and I dont add them to my local entries. On ordinary BSD .o files, the load is ok whether they are static or not. hmmm XLIB>(Xinit) calling 'dl' function sun did not supply..exitting client13:/commun/bdt/src/Trajecto 502 > this means you are somehow invoking something that is invoking something that is trying to load a shared library. This is not supported, since then saving an image and knowing where they mapped the library etc is too system specific. Try to make sure you have -static on the link line (in the make in the unixport directory). I know for example that the yp stuff causes a problem, and in the new release I will have to work around this on solaris. For some reason sun does not give non shared versions of some yp referenced stuff. One sees this in reference to "~joe/.." paths in solaris. I dont know how to make the password lookup stuff work: it seems to invoke loading of shared libraries when invoked. It would seem you are hitting against a similar thing [perhaps the SAME thing]. Bill From lnm@stsn.dga.fr Wed Mar 29 10:15:23 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01625; Wed, 29 Mar 95 10:15:23 CST Received: from asterix.stsn.dga.fr (mailhost.stsn.dga.fr) by cli.com (4.1/SMI-4.1) id AA01924; Wed, 29 Mar 95 08:51:18 CST Received: (from lnm@localhost) by asterix.stsn.dga.fr (8.6.9/8.6.9) id QAA05780; Wed, 29 Mar 1995 16:50:53 +0100 Date: Wed, 29 Mar 1995 16:50:53 +0100 From: LIGNON Marc Message-Id: <199503291550.QAA05780@asterix.stsn.dga.fr> To: gcl@cli.com Subject: Re: Pb interfacing C and Lisp on solaris > (defCfun "silly(x) int x;" 0 > % int y; > ((+ (int x) (int "100")) (int y)) > % printf("\n%d", y); > % y = y+100; > (print (int y)) > (print (int "x+100")) > you might try > (defCfun "static int silly(x) int x;" 0 > ...) > The loader is really designed for the output from the compiler. That > has all functions being static (ie definitions not visible outside) > the file. In fact you cant reference them outside the file, even if > you make them global, since they wont be in the main symbol table, and > I dont add them to my local entries. On ordinary BSD .o files, the > load is ok whether they are static or not. It works fine with the "static" declaration. Thank's a lot. Marc LIGNON ________________________________________________________________________ DGA/DCN/ST/DA/S/ES lignon@stsn.dga.fr 8, boulevard Victor tel: (33)(1) 40.59.13.14 75015 - PARIS - FRANCE fax: (33)(1) 45.79.20.40 From reading@io.litton.com Wed Mar 29 18:33:02 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01671; Wed, 29 Mar 95 18:33:02 CST Received: from lcsdns1.litton.com by cli.com (4.1/SMI-4.1) id AA04691; Wed, 29 Mar 95 17:01:50 CST Received: from io.dsd.litton.com by lcsdns1.litton.com (4.1/SMI-4.1) id AA02943; Wed, 29 Mar 95 14:57:48 PST Received: by io.dsd.litton.com (Concentrix-2800 3.0/SMI-4.1) id AA07304; Wed, 29 Mar 95 15:00:20 PST Date: Wed, 29 Mar 95 15:00:20 PST From: John D. Reading Message-Id: <9503292300.AA07304@io.dsd.litton.com> To: gcl@cli.com Subject: pcl and Solaris 2 A couple of months ago I was running into troubles trying to build PCL on Solaris 2.3. Since then I have had some questions about what success I had, but no other particular progress. I find we are shortly upgrading to Solaris 2.4 (I guess within days). Has anyone had any success with PCL and Solaris 2.3? Will Solaris 2.4 make a difference (for the better, preferably)? Appreciate any help, John -- John D. Reading | Why be difficult when with just a little more reading@io.dsd.litton.com | effort you could be impossible? (818) 597-5152 | From toy@rtp.ericsson.se Sat Apr 1 18:58:57 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02996; Sat, 1 Apr 95 18:58:57 CST Received: from gwa.ericsson.com by cli.com (4.1/SMI-4.1) id AA15100; Sat, 1 Apr 95 17:45:57 CST Received: from mr2.exu.ericsson.se (mr2.exu.ericsson.com [138.85.147.12]) by gwa.ericsson.com (8.6.10/8.6.10) with ESMTP id RAA27261 for ; Sat, 1 Apr 1995 17:45:56 -0600 Received: from screamer.rtp.ericsson.se (screamer.rtp.ericsson.se [147.117.133.13]) by mr2.exu.ericsson.se (8.6.8/NAHUB-MR1.1) with ESMTP id RAA08762 for ; Sat, 1 Apr 1995 17:45:56 -0600 Received: from rcur (rcur7.rtp.ericsson.se [147.117.133.38]) by screamer.rtp.ericsson.se (8.6.8/8.6.4) with ESMTP id SAA21605 for ; Sat, 1 Apr 1995 18:49:06 -0500 Message-Id: <199504012349.SAA21605@screamer.rtp.ericsson.se> To: gcl@cli.com Subject: Fix for reduce with :key Date: Sat, 01 Apr 1995 18:45:48 -0500 From: Raymond Toy About a week ago, I posted a modified version of reduce that accepted the keyword :key as specified in CLTL2. Unfortunately, there is a bug in that if no initial value (:initial-value) is given, the key is not applied to the first element of the sequence. The following is a fix for this so that the key function is applied. Sorry if this caused problems. Ray (defun reduce (function sequence &key from-end start end (initial-value nil ivsp) (key #'identity)) (with-start-end start end sequence (cond ((not from-end) (when (null ivsp) (when (>= start end) (return-from reduce (funcall function))) (setq initial-value (funcall key (elt sequence start))) (setf start (f+ 1 start)) ) (do ((x initial-value (funcall function x (funcall key (prog1 (elt sequence start) (setf start (f+ 1 start)) ))))) ((>= start end) x))) (t (when (null ivsp) (when (>= start end) (return-from reduce (funcall function))) (setf end (f+ end -1)) (setq initial-value (funcall key (elt sequence end))) ) (do ((x initial-value (funcall function (funcall key (elt sequence end)) x))) ((>= start end) x) (setf end (f+ -1 end))))))) From lien@tam.cs.monash.edu.au Mon Apr 10 03:19:58 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04947; Mon, 10 Apr 95 03:19:58 CDT Received: from tam.cs.monash.edu.au by cli.com (4.1/SMI-4.1) id AA08750; Mon, 10 Apr 95 01:56:39 CDT Received: by tam.cs.monash.edu.au (5.67/1.37) id AA04758; Mon, 10 Apr 95 16:56:03 +1000 From: lien@tam.cs.monash.edu.au (Tam T. Lien) Message-Id: <9504100656.AA04758@tam.cs.monash.edu.au> To: gcl@cli.com Cc: lien@tam.cs.monash.edu.au Subject: GCL-1.1 on DOS X-Mailer: exmh version 1.4.1 7/21/94 Date: Mon, 10 Apr 95 16:56:03 EST Hello, Has anyone got GCL-1.1 to compile with DJGCC under DOS? Thanks, Tam. From back0003@gold.tc.umn.edu Mon Apr 10 20:01:49 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05343; Mon, 10 Apr 95 20:01:49 CDT Received: from gold.tc.umn.edu by cli.com (4.1/SMI-4.1) id AA11570; Mon, 10 Apr 95 18:58:09 CDT Received: by gold.tc.umn.edu; Mon, 10 Apr 95 15:34:46 -0500 Date: Mon, 10 Apr 1995 15:34:46 -0500 (CDT) From: Jan E Backlund Subject: ngcl-2.0beta on Slackware Linux To: gcl@cli.com Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Has anyone built gcl-2.0 on Slackware Linux? I had one minor problem with u_int32_t not being defined in inet.h and then got as far as building the raw_gcl. Now I'm stuck with deciphering why saved_gcl is not executable (I've appended the tail end of the make log). Can anyone give me a clue? Jan Erik back0003@gold.tc.umn.edu ------------- Tail end of "make &> error.log" ----------------- make install-command "INSTALL_LIB_DIR=/usr/local/src/gcl-2.0" "PREFIX_DIR=/usr/local/src/gcl-2.0" make[2]: Entering directory `/usr/local/src/gcl-2.0' (echo "#!/bin/sh" ; \ echo exec /usr/local/src/gcl-2.0/unixport/saved_gcl \\ ; \ echo ' -dir' /usr/local/src/gcl-2.0/unixport/ \\ ; \ echo ' -libdir' /usr/local/src/gcl-2.0/ \\ ; \ echo ' -eval '\''(setq si::*allow-gzipped-file* t)'\' \\ ;\ echo ' -eval '\''(setq si::*tk-library* '\"unknown\"')'\' \\;\ echo ' '\"\$@\" ) > /usr/local/src/gcl-2.0/bin/gcl; echo '#' other options: -load "/tmp/foo.o" -load "jo.lsp" -eval '"(joe 3)"' >> /usr/local/src/gcl-2.0/bin/gcl chmod a+x /usr/local/src/gcl-2.0/bin/gcl make[2]: Leaving directory `/usr/local/src/gcl-2.0' (cd xbin ; ln -s ../bin/gcl .) make[1]: Leaving directory `/usr/local/src/gcl-2.0' (cd cmpnew ; make collectfn.o) make[1]: Entering directory `/usr/local/src/gcl-2.0/cmpnew' ../unixport/saved_gcl ../unixport/ -compile collectfn.lsp ../unixport/saved_gcl: ../unixport/saved_gcl: cannot execute binary file make[1]: *** [collectfn.o] Error 126 make[1]: Leaving directory `/usr/local/src/gcl-2.0/cmpnew' make: *** [all] Error 2 From wfs@math.utexas.edu Mon Apr 10 22:52:47 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05408; Mon, 10 Apr 95 22:52:47 CDT Received: from fireant.ma.utexas.edu by cli.com (4.1/SMI-4.1) id AA11799; Mon, 10 Apr 95 21:32:30 CDT Received: from nicolas.ma.utexas.edu by fireant.ma.utexas.edu (5.64/5.51) id AA09320; Mon, 10 Apr 95 21:32:14 -0500 Date: Mon, 10 Apr 95 21:32:06 -0500 From: wfs@math.utexas.edu (Bill Schelter) Posted-Date: Mon, 10 Apr 95 21:32:06 -0500 Message-Id: <9504110232.AA07461@nicolas.ma.utexas.edu> Received: by nicolas.ma.utexas.edu (5.61/5.51) id AA07461; Mon, 10 Apr 95 21:32:06 -0500 To: back0003@gold.tc.umn.edu Cc: gcl@cli.com In-Reply-To: Jan E Backlund's message of Mon, 10 Apr 1995 15:34:46 -0500 (CDT) Subject: ngcl-2.0-beta.tgz vs gcl-2.0.tgz Reply-To: wfs@math.utexas.edu I certainly have been doing development of gcl-2.0 on slackware linux, so it should work! You probably need to take the latest released version which is ftp.cli.com:pub/gcl/gcl-2.0.tgz or ftp.ma.utexas.edu:... I am using linux 1.2.1 gcc 2.6.3, and it should work. The ngcl-2.0-beta.tgz worked in an older version of linux, but the executable format changed slightly. In any event there are other benefits to gcl-2.0, over the beta version. Among other things the socket connection stuff for the windowing has been substantially improved to avoid broken pipes. Bill From mjs@s4mjs.shannon.com Tue Apr 11 03:50:39 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05490; Tue, 11 Apr 95 03:50:39 CDT Received: from cnj.digex.net by cli.com (4.1/SMI-4.1) id AA12510; Tue, 11 Apr 95 02:36:37 CDT Received: from s4mjs (shannon.com) by cnj.digex.net with SMTP id AA14474 (5.67b8/IDA-1.5 for ); Tue, 11 Apr 1995 03:36:32 -0400 Received: by s4mjs (5.x/SMI-SVR4-mjs-950410) id AA02325; Mon, 10 Apr 1995 23:25:07 -0400 Date: Mon, 10 Apr 1995 23:25:07 -0400 From: mjs@s4mjs.shannon.com (marty shannon) Message-Id: <9504110325.AA02325@s4mjs> To: gcl@cli.com Subject: Solaris 2.4 X86 vs. gcl 1 Reply-To: mjs@s4mjs.shannon.com X-Sun-Charset: US-ASCII In a nutshell, it doesn't build. I can't figure out the right things to set in the defs and h file (and Solaris 2.4 does *not* want to build anything static -- as far as I can tell; clues on that would be appreciated, too). Has anyone done this? BTW, "Solaris" is an operating system, not a "platform". Solaris (currently) is available for both Sparc and X86 processors.... Marty Shannon From leverich@rand.org Tue Apr 11 17:28:51 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05726; Tue, 11 Apr 95 17:28:51 CDT Received: from rand.org by cli.com (4.1/SMI-4.1) id AA14652; Tue, 11 Apr 95 15:47:56 CDT Received: from monty.rand.org (monty-cc.rand.org [130.154.8.173]) by rand.org (8.6.10/8.6.10) with ESMTP id NAA22295 for ; Tue, 11 Apr 1995 13:47:45 -0700 Received: from atlantis.rand.org (atlantis.rand.org [130.154.12.124]) by monty.rand.org (8.6.10/8.6.10) with ESMTP id NAA17525 for ; Tue, 11 Apr 1995 13:47:44 -0700 Received: from localhost.rand.org (localhost.rand.org [127.0.0.1]) by atlantis.rand.org (8.6.10/8.6.10) with SMTP id NAA11641 for ; Tue, 11 Apr 1995 13:47:43 -0700 Message-Id: <199504112047.NAA11641@atlantis.rand.org> To: gcl@cli.com Reply-To: Brian_Leverich@rand.org Subject: GCL-2.0 Breaks CLX Date: Tue, 11 Apr 95 13:47:40 PDT From: Brian Leverich The new socket code (and the TK interface) is way cool, but it breaks CLX in a couple of ways. Also, it looks like the new socket code might allow some improvements in CLX performance. Any chance that one of the folks familiar with the socket code could be persuaded to hack CLX to fix the breakage and best exploit the new code? It would improve compatibility with existing apps ... Thanks, B. --------- Dr. Brian Leverich Informations Systems Scientist The RAND Corporation From jsc@esther.mit.edu Wed Apr 12 23:21:48 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA06134; Wed, 12 Apr 95 23:21:48 CDT Received: from esther.mit.edu by cli.com (4.1/SMI-4.1) id AA19409; Wed, 12 Apr 95 22:17:20 CDT Received: by esther.mit.edu id AA06628 (5.67b/IDA-1.5 for gcl@cli.com); Wed, 12 Apr 1995 23:17:50 -0400 Date: Wed, 12 Apr 1995 23:17:50 -0400 From: jsc@esther.mit.edu Message-Id: <199504130317.AA06628@esther.mit.edu> To: gcl@cli.com Subject: bugs in gcl-1.1 Reply-To: jsc@mit.edu I recently tried recompiling gcl 1.1 under Linux 1.2.3 using the latest ELF gcc compiler for that platform. This uncovered a few bugs. 386-linux.h should define DONT_NEED_MALLOC. Here is a backtrace from a segfault: #0 0x800357c in alloc_object (t=t_string) at ../c/alloc.c:201 #1 0x80443fd in alloc_simple_string (l=0) at ../c/string.d:41 #2 0x801eb9a in init_symbol () at ../c/symbol.d:616 #3 0x8001c22 in initlisp () at ../c/main.c:409 #4 0x80019d9 in main (argc=1, argv=0xbffffbd0, envp=0xbffffbd8) at ../c/main.c:275 The problem is Vignore_maximum_pages gets referenced at this point, but isn't ever set except in init_alloc_function(), which gets called in initlisp() *after* init_symbol() gets called. I tried to work around this, but it turns out that Vignore_maximum_pages gets set using make_special(), which depends on string_register, which gets set using alloc_simple_string(), which depends on Vignore_maximum_pages. This compiled fine under gcc 2.6.3, so I can only assume that some default initialization behavior or something changed. Anyways, at a quick glance this looks wrong. Please let me know if it is indeed incorrect, or if I've just overlooked something. On an entirely different subject, does anyone know if CLX and garnet work under gcl? -Jin PS Please Cc any replies to the mailing list directly to me as well, as I've only recently joined. From leverich@rand.org Thu Apr 13 02:19:42 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA06163; Thu, 13 Apr 95 02:19:42 CDT Received: from rand.org by cli.com (4.1/SMI-4.1) id AA19953; Thu, 13 Apr 95 01:01:34 CDT Received: (from Uclarita@localhost) by rand.org (8.6.10/8.6.10) with UUCP id XAA08971; Wed, 12 Apr 1995 23:01:25 -0700 Received: (from leverich@localhost) by saugus.rand.org (8.6.9/8.6.9) id WAA02174; Wed, 12 Apr 1995 22:58:11 -0700 From: Brian Leverich Message-Id: <199504130558.WAA02174@saugus.rand.org> Subject: Re: bugs in gcl-1.1 To: jsc@mit.edu Date: Wed, 12 Apr 1995 22:58:11 -0700 (PDT) Cc: gcl@cli.com, leverich@rand.org In-Reply-To: <199504130317.AA06628@esther.mit.edu> from "jsc@esther.mit.edu" at Apr 12, 95 11:17:50 pm Organization: RAND, Santa Monica, CA Reply-To: Brian_Leverich@rand.org X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 1215 > On an entirely different subject, does anyone know if CLX and garnet > work under gcl? > > -Jin They did. )-8 If you use GCL-1.0 or 1.1, anonymous ftp to ftp://rand.org/pub/leverich and check out the clx and garnet you'll find there. Mike Harm (USC/RAND) and my port of garnet is a little crufty, but it mostly works. Preallocate some extra memory, or you'll be living in GC hell. General performance of GCL/CLX/garnet is satisfactory on a Linux 486DX2-66, satisfactory on a SPARC 2, and primo on a HP 735. We've also integrated RAND's ROSS simulation language with garnet's KR object system, and we're currently working on a GUI for general use with ROSS. Anyway, the new socket code in GCL-2.0 sort of gratuitously breaks CLX. I'm hoping somebody who knows what they're doing (this is a thinly veiled canonical beg and grovel) will release a CLX that's compatible with the new socket code. If no kind soul will do that, I'll eventually do it myself (but don't hold your breath waiting, and it won't be as nicely done as it could have been by somebody who has expertise in the area). Cheers, B. ---------- Dr. Brian Leverich Information Systems Scientist, The RAND Corporation leverich@rand.org From jsc@gwar.mit.edu Thu Apr 13 21:47:25 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA06528; Thu, 13 Apr 95 21:47:25 CDT Received: from gwar.mit.edu by cli.com (4.1/SMI-4.1) id AA22740; Thu, 13 Apr 95 20:34:53 CDT Received: by gwar.mit.edu id AA00555 (5.67b/IDA-1.5 for gcl@cli.com); Thu, 13 Apr 1995 21:34:52 -0400 Date: Thu, 13 Apr 1995 21:34:52 -0400 From: "Jin S. Choi" Message-Id: <199504140134.AA00555@gwar.mit.edu> To: gcl@cli.com Subject: gcl 2.0 and Linux w/ ELF I pulled over gcl 2.0 and tried to get it running under Linux w/ ELF. 1) Again, DONT_NEED_MALLOC has to be defined. 2) Compiling mpi-386d.o fails with undefined symbols, must use mpi.o. 3) raw_gcl dies on startup in alloc_object(), again because sSAignore_maximum_pagesA is referenced before being used in line 223 of alloc.c. It's declared to be an object in a DEFVAR macro at the top of this file, but the macro that gets used is the one in notcomp.h, which just declares it as an `object' (a pointer). Has anyone else run into this problem? It would appear to be a general one... From bslawson@phobos.astro.uwo.ca Mon Apr 17 18:15:59 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00204; Mon, 17 Apr 95 18:15:59 CDT Received: from phobos.astro.uwo.ca by cli.com (4.1/SMI-4.1) id AA01993; Mon, 17 Apr 95 16:52:10 CDT Message-Id: <9504172149.AA03593@phobos.astro.uwo.ca> Date: Mon, 17 Apr 1995 17:49:18 +0500 From: bslawson@phobos.astro.uwo.ca (Bob Slawson) To: gcl@cli.com Subject: bug in gcl-2.0 : aref Reply-To: bslawson@phobos.astro.uwo.ca Content-Length: 962 There is a bug in aref when accessing arrays with rank > 1. Transcript: GCL (GNU Common Lisp) Version(2.0) Mon Apr 17 14:07:56 EDT 1995 Licensed under GNU Public Library License Contains Enhancements by W. Schelter >*features* (:COMPILER :NUMLIB :SDEBUG TRUNCATE_USE_C BSD SPARC SUN SGC IEEE-FLOATING-POINT UNIX GCL AKCL COMMON KCL) >(defvar test-array (make-array '(2 3) :initial-element 0)) TEST-ARRAY >test-array #2A((0 0 0) (0 0 0)) >(setf (aref test-array 1 1) 4) 4 >test-array #2A((0 0 0) (0 4 0)) --- This looks OK >(aref test-array 1 1) 0 --- Ooops! >(setf (aref test-array 0 0) 9) 9 >test-array #2A((9 0 0) (0 4 0)) --- This is OK too. >(aref test-array 0 1) 9 --- Oops should be 0 >(aref test-array 2 0) 0 --- Should be an errror! >(setf (aref test-array 2 0) 20) Error: Index 0 to array is too large Fast links are on: do (use-fast-links nil) for debugging Error signalled by EVAL. Broken at SETF. Type :H for Help. >> BobS From wfs@math.utexas.edu Tue Apr 18 01:45:24 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00480; Tue, 18 Apr 95 01:45:24 CDT Received: from parthenon.ma.utexas.edu (parthenon.pi.ma.utexas.edu) by cli.com (4.1/SMI-4.1) id AA02784; Tue, 18 Apr 95 00:33:05 CDT Received: (from wfs@localhost) by parthenon.ma.utexas.edu (8.6.10/8.6.10) id AAA10754; Tue, 18 Apr 1995 00:31:56 -0500 Date: Tue, 18 Apr 1995 00:31:56 -0500 From: Bill Schelter Message-Id: <199504180531.AAA10754@parthenon.ma.utexas.edu> To: gcl@cli.com Subject: patch for 2.0 Reply-To: wfs@fireant.ma.utexas.edu Here is a patch for the multidimensional array aref bug, for ./saved_gcl -compile, and for the gcl-2.0/makefile in case you do not have tcl/tk or X. There will be a new release shortly but in the meantime: To apply the patch do % cd gcl-2.0/.. % patch -p < patch-1 % (cd gcl-2.0 ; make) ==================begin patch-1=============================== *** /tmp/gcl-2.0/makefile Sun Apr 9 22:14:48 1995 --- gcl-2.0/makefile Thu Apr 13 12:42:22 1995 *************** *** 106,112 **** then echo "Make of GCL `cat majvers`.`cat minvers` completed." ; \ xbin/notify "CC=${CC}" ; \ else echo "The new image failed to compile collectfn.lsp" ;fi ! if [ "${X11_LIBS}" != unknown ] ; then if [ "${TK_LIBS}" != unknown ]; \ then (cd gcl-tk ; make) ; fi ; fi (cd info ; make) --- 109,115 ---- then echo "Make of GCL `cat majvers`.`cat minvers` completed." ; \ xbin/notify "CC=${CC}" ; \ else echo "The new image failed to compile collectfn.lsp" ;fi ! if [ "${X11_LIBS}" != "-Lunknown -lX11" ] ; then if [ "${TK_LIBS}" != "-Lunknown -ltk" ]; \ then (cd gcl-tk ; make) ; fi ; fi (cd info ; make) *** /tmp/gcl-2.0/o/array.c Sat Mar 4 07:50:37 1995 --- gcl-2.0/o/array.c Tue Apr 18 00:13:12 1995 *************** *** 70,76 **** FEerror(" ~a has wrong rank",x); if (rank == 1) return fSaref1(x,i); va_start(ap); ! m = 0; k = i; /* index into 1 dimensional array */ i1 = 0; --- 70,76 ---- FEerror(" ~a has wrong rank",x); if (rank == 1) return fSaref1(x,i); va_start(ap); ! m = 1; k = i; /* index into 1 dimensional array */ i1 = 0; *************** *** 80,86 **** if (k > x->a.a_dims[m]) FEerror("Index ~a to array is too large",1,make_fixnum (m)); i1 += k; ! if (m < rank) { i1 = i1 * x->a.a_dims[m]; if (m < N_FIXNUM_ARGS) { k = va_arg(ap,int);} --- 80,86 ---- if (k > x->a.a_dims[m]) FEerror("Index ~a to array is too large",1,make_fixnum (m)); i1 += k; ! if (m <= rank) { i1 = i1 * x->a.a_dims[m]; if (m < N_FIXNUM_ARGS) { k = va_arg(ap,int);} *************** *** 87,96 **** else {object x = va_arg(ap,object); check_type(x,t_fixnum); k = Mfix(x);} ! m++;} else break;} va_end(ap); ! return fSaref1(x,i); } if (n > 2) { FEerror("Too many args (~a) to aref",1,make_fixnum(n));} --- 87,97 ---- else {object x = va_arg(ap,object); check_type(x,t_fixnum); k = Mfix(x);} ! m++; ! } else break;} va_end(ap); ! return fSaref1(x,i1); } if (n > 2) { FEerror("Too many args (~a) to aref",1,make_fixnum(n));} *** /tmp/gcl-2.0/unixport/init_gcl.lsp Sat Apr 1 19:47:23 1995 --- gcl-2.0/unixport/init_gcl.lsp Wed Apr 12 15:50:20 1995 *************** *** 54,66 **** :output-file (or (si::get-command-arg "-o") (si::get-command-arg "-compile")) ! :o-file (not ! (or (equalp (si::get-command-arg "-o-file") ! "nil") ! (not (si::get-command-arg "-no-o" ! t))) ! ) :c-file (si::get-command-arg "-c-file" t) :h-file (si::get-command-arg "-h-file" t) :data-file (si::get-command-arg "-data-file" t) --- 54,65 ---- :output-file (or (si::get-command-arg "-o") (si::get-command-arg "-compile")) ! :o-file ! (cond ((equalp (si::get-command-arg "-o-file") ! "nil") nil) ! ((si::get-command-arg "-o-file" t)) ! (t t)) :c-file (si::get-command-arg "-c-file" t) :h-file (si::get-command-arg "-h-file" t) :data-file (si::get-command-arg "-data-file" t) ------------------------------end patch-1 ---------------- From gene@ccs.neu.edu Tue Apr 18 14:38:34 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00627; Tue, 18 Apr 95 14:38:34 CDT Received: from amber.ccs.neu.edu by cli.com (4.1/SMI-4.1) id AA04006; Tue, 18 Apr 95 13:00:33 CDT Received: from highpoint.ccs.neu.edu (gene@highpoint.ccs.neu.edu [129.10.112.80]) by amber.ccs.neu.edu (8.6.10/8.6.4) with ESMTP id OAA15258; Tue, 18 Apr 1995 14:00:28 -0400 From: Gene Cooperman Received: (gene@localhost) by highpoint.ccs.neu.edu (8.6.10/8.6.4) id OAA16854; Tue, 18 Apr 1995 14:00:27 -0400 Date: Tue, 18 Apr 1995 14:00:27 -0400 Message-Id: <199504181800.OAA16854@highpoint.ccs.neu.edu> To: gcl@cli.com, wfs@fireant.ma.utexas.edu Subject: Re: another make-array bug ? I encountered another make-array bug for arrays of rank 1. I don't know if this is covered by the previous patch, but I would guess it's different. When I type: (make-array 256 :element-type 'fixnum ":fill-pointer t) I get an error: Correctable error: Expected a FIXNUM Signalled by EVAL. If continued: Supply right type Broken at MAKE-ARRAY. Type :H for Help. When I type: (make-array 256 :element-type 'fixnum ":fill-pointer 5) it results in an infinite loop. - Gene Cooperman From bslawson@phobos.astro.uwo.ca Tue Apr 18 23:09:47 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00724; Tue, 18 Apr 95 23:09:47 CDT Received: from phobos.astro.uwo.ca by cli.com (4.1/SMI-4.1) id AA05291; Tue, 18 Apr 95 22:01:42 CDT Message-Id: <9504190258.AA25164@phobos.astro.uwo.ca> Date: Tue, 18 Apr 1995 22:58:37 +0500 From: bslawson@phobos.astro.uwo.ca (Bob Slawson) To: wfs@fireant.ma.utexas.edu Cc: gcl@cli.com In-Reply-To: <199504180531.AAA10754@parthenon.ma.utexas.edu> (message from Bill Schelter on Tue, 18 Apr 1995 00:31:56 -0500) Subject: Re: patch for 2.0 Reply-To: bslawson@phobos.astro.uwo.ca Content-Length: 2603 Patch-1 didn't seem to fix the aref problem so I worked on it myself. It's been tested with interpreted forms and compiled code at all possible safety levels. A context diff follows (it applies on top of patch-1). I don't understand the necessity of changing the flag in the DEFUN ("AREF1" ... from ONE_VAL to NONE, but this is what was necessary for aref to work in compiled code. I merely paralleled the DEFUN with DEFUN ("ASET1". (All compiled code I tried calls `fSaref1' irrespective of safety level because of a macro in `cmpinclude.h'). BobS ------------------------------------------------- *** o/array.c-patchd-1 Tue Apr 18 21:49:05 1995 --- o/array.c Tue Apr 18 22:08:38 1995 *************** *** 59,65 **** (x, i, va_alist) object x; int i; ! { int n = VFUN_NARGS; int i1; va_list ap; --- 59,65 ---- (x, i, va_alist) object x; int i; ! va_dcl { int n = VFUN_NARGS; int i1; va_list ap; *************** *** 70,76 **** FEerror(" ~a has wrong rank",x); if (rank == 1) return fSaref1(x,i); va_start(ap); ! m = 1; k = i; /* index into 1 dimensional array */ i1 = 0; --- 70,76 ---- FEerror(" ~a has wrong rank",x); if (rank == 1) return fSaref1(x,i); va_start(ap); ! m = 0; k = i; /* index into 1 dimensional array */ i1 = 0; *************** *** 77,85 **** rank-- ; while(1) { ! if (k > x->a.a_dims[m]) FEerror("Index ~a to array is too large",1,make_fixnum (m)); i1 += k; if (m <= rank) { i1 = i1 * x->a.a_dims[m]; if (m < N_FIXNUM_ARGS) --- 77,86 ---- rank-- ; while(1) { ! if (k >= x->a.a_dims[m]) FEerror("Index ~a to array is too large",1,make_fixnum (m)); i1 += k; + m++; if (m <= rank) { i1 = i1 * x->a.a_dims[m]; if (m < N_FIXNUM_ARGS) *************** *** 87,93 **** else {object x = va_arg(ap,object); check_type(x,t_fixnum); k = Mfix(x);} - m++; } else break;} va_end(ap); --- 88,93 ---- *************** *** 131,137 **** } DEFUN("AREF1", object, fSaref1, SI, 2, 2, ! ONE_VAL, OO, IO, OO,OO, "For array X and index I it returns (aref x i) as if x were \ 1 dimensional, even though its rank may be bigger than 1") (x, i) --- 131,137 ---- } DEFUN("AREF1", object, fSaref1, SI, 2, 2, ! NONE, OO, IO, OO,OO, "For array X and index I it returns (aref x i) as if x were \ 1 dimensional, even though its rank may be bigger than 1") (x, i) From bf@next.ipl.fr Wed Apr 19 04:44:42 1995 Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00962; Wed, 19 Apr 95 04:44:42 CDT Received: from cpe.ipl.fr by cli.com (4.1/SMI-4.1) id AA05979; Wed, 19 Apr 95 03:26:20 CDT Return-Path: bf@next.ipl.fr Received: from next.ipl.fr (next.ipl.fr [193.48.221.6]) by cpe.ipl.fr (8.6.5/8.6.5) with SMTP id KAA03268 for ; Wed, 19 Apr 1995 10:25:40 +0100 Received-Date: Wed, 19 Apr 1995 10:25:40 +0100 Received: by next.ipl.fr (NX5.67e/NX3.0S) id AA00428; Wed, 19 Apr 95 09:22:23 +0100 Date: Wed, 19 Apr 95 09:22:23 +0100 From: Beatrice Fuchs Message-Id: <9504190822.AA00428@next.ipl.fr> Received: by NeXT.Mailer (1.100) Received: by NeXT Mailer (1.100) To: gcl@cli.com Subject: unsubscribe Please, remove my address from the mailing list. Bests regards Beatrice Fuchs-- Beatrice Fuchs, CPE LYON, Laboratoire LISA,| 31, Place Bellecour, F-69288 Lyon Cedex 02 | bf@cpe.ipl.fr Tel : (33) 72 32 50 63 | -------------------------------------------------------------------------------- From pw@snoopy.mv.com Sat Apr 22 20:28:37 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02363; Sat, 22 Apr 95 20:28:37 CDT Received: from snoopy.mv.com by cli.com (4.1/SMI-4.1) id AA16054; Sat, 22 Apr 95 19:23:39 CDT Received: (from pw@localhost) by snoopy.mv.com (8.6.9/8.6.6) id TAA27534; Sat, 22 Apr 1995 19:54:14 -0400 Date: Sat, 22 Apr 1995 19:54:14 -0400 From: "Paul F. Werkowski" Message-Id: <199504222354.TAA27534@snoopy.mv.com> To: gcl@cli.com In-Reply-To: <9504190258.AA25164@phobos.astro.uwo.ca> (bslawson@phobos.astro.uwo.ca) Subject: Re: patch for 2.0 >>>>> "Bob" == Bob Slawson writes: Bob> Patch-1 didn't seem to fix the aref problem so I worked on it Bob> myself. It's been tested with interpreted forms and compiled Bob> code at all possible safety levels. Not wanting to confuse this issue more, but, having just grabbed the gcl-2.0 kit from its home along with patch and attempted to build it on FreeBSD 2.0. I found that while the unpatched installation built and ran OK (except for possibly the array bug) but ANY patch to array.c resulted in >Loading ../gcl-tk/tk-package.lsp Finished loading ../gcl-tk/tk-package.lsp T >Initializing ../lsp/predlib.o Initializing ../lsp/setf.o Initializing ../lsp/arraylib.o Initializing ../lsp/assert.o Initializing ../lsp/defstruct.o Error: ARRAY-HAS-FILL-POINTER-P is invalid as a function. Error signalled by ADJUST-ARRAY. Broken at BREAK-LEVEL. Type :H for Help. . To repeat: . I can build a vanilla system no problem. . Patch for no tcl/tk (part of patch-1) works ok. . Any of the changes to array.c produces the above abort. I don't have the time to persue this more right now. Please consider it a heads-up. Paul From etokna@etn.ericsson.se Mon Apr 24 04:10:35 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02845; Mon, 24 Apr 95 04:10:35 CDT Received: from mailgate.ericsson.se by cli.com (4.1/SMI-4.1) id AA17821; Mon, 24 Apr 95 02:36:43 CDT Received: from hisoy0 (hisoy0.etn.ericsson.se [193.161.188.2]) by mailgate.ericsson.se (8.6.11/1.0) with SMTP id JAA17681 for ; Mon, 24 Apr 1995 09:36:28 +0200 Received: from hisoy1.etn.ericsson.se by hisoy0 (5.0/LME-DOM-2.2.3) id AA16226; Mon, 24 Apr 1995 09:36:46 +0200 From: etokna@etn.ericsson.se (Knut-Haavard Aksnes) Received: by hisoy1.etn.ericsson.se (4.1/client-1.5) id AA12903; Mon, 24 Apr 95 09:34:01 +0200 Date: Mon, 24 Apr 95 09:34:01 +0200 Message-Id: <9504240734.AA12903@hisoy1.etn.ericsson.se> To: gcl@cli.com Subject: PCL for gcl-2.0 Comments: Hyperbole mail buttons accepted, v3.18.3. Content-Length: 336 Has anybody had any success in building PCL for GCL-2.0? -- ------------------------------------------------------------------------------- Name: Knut-Håvard Aksnes (ECMA 94) or Knut-Haavard Aksnes (ASCII) Ericsson signature: HI/ETO/X/I KNA Phone: +47 37 05 14 81 Email: etokna@hisoy.etn.ericsson.se (internet) ETO.ETOKNA (memo) From thorsten@arch.su.EDU.AU Fri Apr 28 03:06:07 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA04515; Fri, 28 Apr 95 03:06:07 CDT Received: from arch.su.EDU.AU (archsci.arch.su.EDU.AU) by cli.com (4.1/SMI-4.1) id AA01283; Fri, 28 Apr 95 01:42:59 CDT Received: from glass (glass.arch.su.EDU.AU) by arch.su.EDU.AU (4.1/SMI-4.1) id AA00544; Fri, 28 Apr 95 16:34:48 EST From: thorsten@arch.su.EDU.AU (Thorsten Schnier) Received: by glass (5.x) id AA05146; Fri, 28 Apr 1995 16:39:55 +1000 Date: Fri, 28 Apr 1995 16:39:55 +1000 Message-Id: <9504280639.AA05146@glass> To: gcl@cli.com Subject: Problems with defCfun in gcl-2.0 X-Sun-Charset: US-ASCII After installing gcl 2.0 (on solaris and linux w/o ELF), some inline c-functions didn't work anymore. While loading a file containing a defCfun, the loader complained about not knowing the symbols Creturn and CMPfuncall. The new documentation doesn't mention creturn anymore, and I was able to rewrite the code without it. When comparing the intermediate C-Code with the output from gcl-1.1, I noticed a macro for CMPfuncall: #define CMPfuncall funcall with this line added in the function code, loading works again. Is this a bug ? And is Creturn obsolete ? Or do I have to add the macro for creturn from gcl-1.1 code ? BTW. Someone had problems with linux with ELF. Is this solved ? thorsten From anders.vinjar@notam.uio.no Sat Apr 29 15:31:29 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA05081; Sat, 29 Apr 95 15:31:29 CDT Received: from mons.uio.no by cli.com (4.1/SMI-4.1) id AA05759; Sat, 29 Apr 95 14:17:34 CDT Received: from ulrik.uio.no by mons.uio.no with local-SMTP (PP) id <13480-0@mons.uio.no>; Sat, 29 Apr 1995 21:17:25 +0200 Received: by alban.notam.uio.no ; Sat, 29 Apr 1995 19:14:52 GMT Date: Sat, 29 Apr 1995 19:14:52 GMT From: anders.vinjar@notam.uio.no Message-Id: <199504291914.TAA15011@alban.notam.uio.no> To: gcl@cli.com Subject: gcl-2.0 on irix5 (patch-level 2). PCL stops it! I've successfully built and saved an executing version of gcl-2.0 on the SGI (irix-5.2). The gcl-tcl interface works seemingly allright. When trying to build pcl-gcl-2.1 (from ftp.ma.utexas.edu) it goes all the way through the compilation and loading without warnings, but si::save-system fails to save the image. The error message is like this: Error: Sorry, Cant save an image with dynamically loaded .o files Fast links are on: do (use-fast-links nil) for debugging Error signalled by EVAL. Broken at SYSTEM:SAVE-SYSTEM. Type :H for Help. >> What can i do to avoid the problem, or to find some bugs to fix? -- Anders Vinjar | NoTAM - | Addr: Ole Jacob Brochs gt. 1 | http://www.notam.uio.no 0454 Oslo, Norway | Vox: +47 22857970 Vox: + 47 22607968 | Fax: +47 22857974 From ahe@cs.unh.edu Mon May 1 11:28:54 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA06725; Mon, 1 May 95 11:28:54 CDT Received: from cs.unh.edu by cli.com (4.1/SMI-4.1) id AA10514; Mon, 1 May 95 10:11:46 CDT Received: from mica.cs.unh.edu by cs.unh.edu (5.x/SMI-SVR4) id AA27480; Mon, 1 May 1995 11:11:45 -0400 Received: by mica.cs.unh.edu (5.x/SMI-SVR4) id AA02994; Mon, 1 May 1995 11:11:44 -0400 Date: Mon, 1 May 1995 11:11:44 -0400 From: ahe@cs.unh.edu (Andy Evans) Message-Id: <9505011511.AA02994@mica.cs.unh.edu> To: gcl@cli.com Subject: Ultrix 4.4 on a 5000 - any success? Hello folks - I've been trying to build gcl-2.0 and gcl-1.1 on a DECstation 5000 with Ultrix 4.4 - without success. I'm using either DEC's cc or gcc 2.6.3. The process has gotten as far as "loading auto.lsp" but it has never finished making saved_gcl. The raw_gcl has core-dumped or complained of illegal instruction (last one was "GBC not enabled" on 1.1). Does anyone have any tips or reccomendations? I'd be grateful! Thanks - - Andy ==================================================== Andrew Evans Computer Science Department Computer Specialist University of New Hampshire ahe@cs.unh.edu Durham, NH 03824 ==================================================== From haleysj@HIRAMF.hiram.edu Mon May 1 17:02:45 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00222; Mon, 1 May 95 17:02:45 CDT Received: from HIRAMF.hiram.edu by cli.com (4.1/SMI-4.1) id AA11894; Mon, 1 May 95 15:47:15 CDT Received: by HIRAMF.hiram.edu (5.65/DEC-Ultrix/4.3) id AA15562; Mon, 1 May 1995 16:53:14 -0400 Date: Mon, 1 May 1995 16:53:13 -0400 (EDT) From: Steve To: gcl@cli.com Subject: GCL on DecStation 3100 / VAXstation 3100 Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hello! I am having a lot of trouble compiling GCL-2.0 on both a DECstation 3100 and a VAXstation 3100. When compiling with the stock cc, I get many compiler warnings and it eventually stops. Changing references in the dec3100.defs from cc to gcc (version 2.6.3) makes compiling go well until raw_gcl. When running raw_gcl, it reports an illegal instruction and core dumps. This happens both on the vaxstation using vax.defs and the decstation using dec3100.defs. They are running Ultrix 4.4 and 4.3 respectively. Any ideas? Thanks in advance. I hope I can get it to work, for gcl looks like a nice package. -- Steve Haley ...................*****************************************.................... haleysj@hiramf.hiram.edu haleysj@hiram.edu bm711@cleveland.freenet.edu =) :) 8-) :> =D -------------------------------------------------------------------------------- From tkunze@ccrma.Stanford.EDU Mon May 1 18:22:02 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00239; Mon, 1 May 95 18:22:02 CDT Received: from ccrma.Stanford.EDU (ccrma.Stanford.EDU) by cli.com (4.1/SMI-4.1) id AA12120; Mon, 1 May 95 17:04:46 CDT Received: from cmn10 by ccrma.Stanford.EDU (NX5.67e/NeXT-1.0) id AA06121; Mon, 1 May 95 15:04:44 -0700 From: tkunze@ccrma.Stanford.EDU (Tobias Kunze) Message-Id: <9505012204.AA06121@ ccrma.Stanford.EDU > Received: by cmn10.Stanford.EDU (NX5.67e/NX3.0X) id AA21169; Mon, 1 May 95 15:04:43 -0700 Date: Mon, 1 May 95 15:04:43 -0700 Received: by NeXT.Mailer (1.100) Received: by NeXT Mailer (1.100) To: gcl@cli.com Subject: gcl2.0 on irix5 Hi- has anyone been able to successfully build gcl2 with pcl 2.1? Also: does anyone happen to know why it is that the irix5 port can't save an image with dynamically loaded code? TIA, -Tobias From erik@naggum.no Tue May 2 18:47:28 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00836; Tue, 2 May 95 18:47:28 CDT Received: from naggum.no by cli.com (4.1/SMI-4.1) id AA16658; Tue, 2 May 95 17:22:26 CDT Received: by naggum.no id for gcl@cli.com; Tue, 2 May 1995 22:22:23 UT Date: 02 May 1995 21:48:49 UT From: Erik Naggum Organization: Naggum Software; +47 2295 0313 Message-Id: <19950502.12274@naggum.no> To: gcl@cli.com Subject: compile problems I retrieved GCL-2.1, and appreciate the developments, but I'm a little surprised by the difficulties I'm having in changing the prompting and exiting behavior of top-level and break-level. I also don't want tcl/tk and find it hard to build without them. here's the desired diff to top-level (I redefine it in init_gcl.lsp): --- top.lsp Sun Apr 2 01:40:14 1995 +++ ../unixport/init_gcl.lsp Tue May 2 17:14:47 1995 @@ -100,7 +20,7 @@ (loop (setq +++ ++ ++ + + -) (if *no-prompt* (setq *no-prompt* nil) - (format t "~%~a>" + (format t "~&~a> " (if (eq *package* (find-package 'user)) "" (package-name *package*)))) (reset-stack-limits) @@ -109,55 +29,19 @@ (when (catch *quit-tag* (setq - (locally (declare (notinline read)) (read *standard-input* nil *top-eof*))) - (when (eq - *top-eof*) (bye)) + (when (eq - *top-eof*) (terpri) (bye)) (let ((values (multiple-value-list (locally (declare (notinline eval)) (eval -))))) (setq /// // // / / values *** ** ** * * (car /)) (fresh-line) (dolist (val /) (locally (declare (notinline prin1)) (prin1 val)) - (terpri)) + (princ " ")) nil)) (setq *evalhook* nil *applyhook* nil) (terpri *error-output*) (break-current))))) however, when I made these trivial changes, I got a barrage of errors, and gcl failed to build. the amazing part is that compiling an _unmodified_ lsp/top.lsp crashes equally spectacularly. (the copy to top1.lsp is only to prevent destruction of top.*.) I also note that there's a stray ) in this file, but removing it had no effect whatsoever. I'm not sure that's a good sign, either. $ cp top.lsp top1.lsp $ make top1.o Compiling top1.lsp. Loading ./cmpinit.lsp Finished loading ./cmpinit.lsp Warning: WARN is being redefined. Warning: BREAK is being redefined. End of Pass 1. Error: Caught fatal error [memory may be damaged] Fast links are on: do (use-fast-links nil) for debugging Error signalled by CATCH. Backtrace: gcl-top-level > if > let > error-set > error-set > progn > compile-file > loop > setq > apply > apply > compiler::compile-file1 > let > unwind-protect > progn > let* > let > unwind-protect > progn > when > compiler::compiler-pass2 > let > unwind-protect > progn > let > let > unwind-protect > progn > catch > compiler::ctop-write > universal-error-handler > break-level > let* > UNLESS top1.c: In function `init_top1': top1.c:4: `VV' undeclared (first use this function) top1.c:4: (Each undeclared identifier is reported only once top1.c:4: for each function it appears in.) top1.c: In function `LI1': top1.c:12: `VV' undeclared (first use this function) top1.c:38: `Lnk140' undeclared (first use this function) top1.c:44: `Lnk141' undeclared (first use this function) top1.c:167: `Lnk145' undeclared (first use this function) top1.c: In function `L2': top1.c:176: `VM2' undeclared (first use this function) top1.c:187: `VV' undeclared (first use this function) top1.c:191: `Lnk140' undeclared (first use this function) top1.c:199: `Lnk146' undeclared (first use this function) top1.c: In function `L3': top1.c:227: `VM3' undeclared (first use this function) top1.c:243: `VV' undeclared (first use this function) top1.c:290: `Lnk147' undeclared (first use this function) top1.c:295: `Lnk148' undeclared (first use this function) top1.c: In function `L4': top1.c:316: `VM4' undeclared (first use this function) top1.c:333: `VV' undeclared (first use this function) top1.c:350: `Lnk149' undeclared (first use this function) top1.c:358: `Lnk150' undeclared (first use this function) top1.c:366: `Lnk151' undeclared (first use this function) top1.c:405: `Lnk152' undeclared (first use this function) top1.c:409: `Lnk153' undeclared (first use this function) top1.c:414: `Lnk154' undeclared (first use this function) top1.c:419: `LnkLI155' undeclared (first use this function) top1.c:427: `Lnk156' undeclared (first use this function) top1.c:492: `Lnk157' undeclared (first use this function) top1.c:596: `Lnk145' undeclared (first use this function) top1.c: At top level: top1.c:605: parse error at end of input $ tail top1.c vs_base=vs_top; (void) (*Lnk145)(); vs_top=sup; T210:; goto T209;} } } /* local entry for function WARN */ static object LI5(V30 compiling top1.lsp anywhere but in the distribution lsp directory succeeds. I have made the changes in unixport/init_gcl.lsp and can live with that, but I find this more than disturbing, so I thought you might be interested in a report on the behavior. # From johng@jolt.mpx.com.au Wed May 3 01:36:59 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01027; Wed, 3 May 95 01:36:59 CDT Received: from kyoko.mpx.com.au by cli.com (4.1/SMI-4.1) id AA17798; Wed, 3 May 95 00:30:57 CDT Received: from jolt.mpx.com.au by kyoko.mpx.com.au with smtp (Smail3.1.29.1 #6) id m0s6X1s-0006NXC; Wed, 3 May 95 15:31 EST Received: by jolt.mpx.com.au (Smail3.1.28.1 #2) id m0s6Wzq-0005tZC; Wed, 3 May 95 15:29 EST Message-Id: From: johng@jolt.mpx.com.au (John Gibbons) Subject: GCL 2.x and Solaris 2.3/4 To: gcl@cli.com Date: Wed, 3 May 1995 15:29:14 +1000 (EST) X-Mailer: ELM [version 2.4 PL21] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 187 Has anoyone succeeded in building PCL-GCL under Solaris 2.3 or 2.4 with gcl2.0. I have built gcl itself but pcl always fails to build in a dramatic way. Any help would be appreciated. From anders.vinjar@notam.uio.no Wed May 3 04:30:24 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA01068; Wed, 3 May 95 04:30:24 CDT Received: from mons.uio.no by cli.com (4.1/SMI-4.1) id AA18213; Wed, 3 May 95 03:20:36 CDT Received: from ulrik.uio.no by mons.uio.no with local-SMTP (PP) id <28573-0@mons.uio.no>; Wed, 3 May 1995 10:20:11 +0200 Received: by alban.notam.uio.no ; Wed, 3 May 1995 08:17:34 GMT Date: Wed, 3 May 1995 08:17:34 GMT From: anders.vinjar@notam.uio.no Message-Id: <199505030817.IAA06857@alban.notam.uio.no> To: johng@jolt.mpx.com.au Cc: gcl@cli.com In-Reply-To: (johng@jolt.mpx.com.au) Subject: Re: GCL 2.x and Solaris 2.3/4 >>>>> johng@jolt.mpx.com.au (John Gibbons) writes: John> Has anoyone succeeded in building PCL-GCL under Solaris 2.3 or John> 2.4 with gcl2.0. John> I have built gcl itself but pcl always fails to build in a John> dramatic way. John> Any help would be appreciated. Could you please describe the situation a bit closer? From saruman.lb.bawue.de!shendi@delos.s.bawue.de Sat May 6 11:11:00 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02740; Sat, 6 May 95 11:11:00 CDT Received: from delos.s.bawue.de by cli.com (4.1/SMI-4.1) id AA29016; Sat, 6 May 95 10:01:08 CDT Received: from saruman.lb.bawue.de by delos.s.bawue.de with uucp (Smail3.1.28.1 #3) id m0s7hUW-000BHmC; Sat, 6 May 95 12:53 MET DST Received: by saruman.lb.bawue.de (Smail3.1.28.1 #1) id m0s7gci-0001b9C; Sat, 6 May 95 11:58 MET DST Message-Id: From: shendi@saruman.lb.bawue.de (Alexander Shendi) Subject: Prevent Loading of init.lsp in GCL 1.1 To: gcl@cli.com Date: Sat, 6 May 1995 11:58:07 +0200 (MET DST) In-Reply-To: from "John Gibbons" at May 3, 95 03:29:14 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 80 Is there a way to prevent loading init.lsp in GCl 1.1. Thanx in advance Alex From kr@shell.portal.com Sat May 6 17:55:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA02826; Sat, 6 May 95 17:55:56 CDT Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA29558; Sat, 6 May 95 16:52:26 CDT Received: from jobe.shell.portal.com (jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.11/8.6.5) with ESMTP id OAA14876; Sat, 6 May 1995 14:51:43 -0700 Received: from DialupEudora (jobe.shell.portal.com [156.151.3.4]) by jobe.shell.portal.com (8.6.11/8.6.5) with SMTP id OAA05899; Sat, 6 May 1995 14:51:40 -0700 X-Sender: kr@pop.shell.portal.com Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: srialpop\r\e%, 6 May 1995 14:52:13 -0800 To: shendi@saruman.lb.bawue.de (Alexander Shendi), gcl@cli.com From: kr@shell.portal.com (kr) Subject: Re: Prevent Loading of init.lsp in GCL 1.1 Sender: gcl-request@cli.com At 11:58 5/6/95, Alexander Shendi wrote: >Is there a way to prevent loading init.lsp in GCl 1.1. How about just removing that file or keeping it empty ? :-) Greetings Markus Krummenacker From yuasa@ylab.tutics.tut.ac.jp Sun May 7 20:40:48 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03206; Sun, 7 May 95 20:40:48 CDT Received: from katura.ylab.tutics.tut.ac.jp by cli.com (4.1/SMI-4.1) id AA01623; Sun, 7 May 95 19:27:50 CDT Received: (from yuasa@localhost) by katura.ylab.tutics.tut.ac.jp (8.6.9+2.4W/3.3W9) id JAA15531 for gcl@cli.com; Mon, 8 May 1995 09:23:52 +0900 Date: Mon, 8 May 1995 09:23:52 +0900 From: Taiichi Yuasa Message-Id: <199505080023.JAA15531@katura.ylab.tutics.tut.ac.jp> To: gcl@cli.com Subject: RE: Prevent Loading of init.lsp in GCL 1.1 >> Is there a way to prevent loading init.lsp in GCl 1.1. In the original KCL, the instruction to load init.lsp is given in the file lsp/top.lsp. (defun top-level () ... (catch *quit-tag* (when (probe-file "init.lsp") (load "init.lsp"))) ... ) Commenting out the entire CATCH form will prevent loading init.lsp. -- Taiichi From kr@shell.portal.com Sun May 7 20:42:39 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03210; Sun, 7 May 95 20:42:39 CDT Errors-To: sandel@cli.com Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA01624; Sun, 7 May 95 19:30:37 CDT Errors-To: sandel@cli.com Received: from jobe.shell.portal.com (jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.11/8.6.5) with ESMTP id RAA11193 for ; Sun, 7 May 1995 17:29:54 -0700 From: kr@shell.portal.com Received: (kr@localhost) by jobe.shell.portal.com (8.6.11/8.6.5) id RAA06468; Sun, 7 May 1995 17:29:53 -0700 Date: Sun, 7 May 1995 17:29:53 -0700 Message-Id: <199505080029.RAA06468@jobe.shell.portal.com> To: gcl@cli.com Subject: onscure format quirks Sender: gcl-request@cli.com Errors-To: gcl-request@cli.com I have repeatedly used the (format ...) function to print out numbers in a flexible and variable fashion. I think that the syntax of this construct is pretty brain-dead and user-unfriendly, and many of the options are pretty obscure (and so not well tested and debugged, it seems). So I probably only have myself to blame for having run into problems with various LISPs by using format. But OTOH there must be reason why it was included, and there is no reasonable alternative I know of, so I'll have to use what is available. I have encountered the following oddities in gcl-1.1 : * extra comma, but empty field: ------------------------------ (format nil "~v,'0,X" 8 1234) yields this: ========== >(format nil "~v,'0,X" 8 1234) (format nil "~v,'0,X" 8 1234) Error: Format error: illegal ,. V "~v,'0,X" Fast links are on: do (use-fast-links nil) for debugging Error signalled by FORMAT. Broken at FORMAT. Type :H for Help. >> ========== With my reading of CLtL2, p. 586, it should be legal to have all these commas in there, even if a field happens to be empty (I even read it as if though it were required). This works with no problem in MCL2.0 . So I do not know whether this is a bug in GCL or in MCL, or whether I just should code in a more defensive style. * no comma, empty field: ----------------------- (format nil "~v,'0X" 8 1234) works correctly: ========== >(format nil "~v,'0X" 8 1234) "000004D2" > ========== This also works in the same way in MCL2.0 . * with comma, and specified field with "commachar": -------------------------------------------------- (format nil "~v,'0,' X" 8 1234) does not do quite what one would expect: ========== >(format nil "~v,'0,' X" 8 1234) "000004D2" > ========== It should have divided the output in blocks with a length of three (as a default). In MCL2.0 it gives the following as expected: ========== ? (format nil "~v,'0,' X" 8 1234) " 000 4D2" ? ========== I just wanted to toss out these observations. Maybe other people here have encountered these issues, and maybe know definitive answers as to what is correct.. Many Greetings Markus Krummenacker From marks@earth.sarnoff.com Mon May 8 13:39:19 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03578; Mon, 8 May 95 13:39:19 CDT Received: from groucho.sarnoff.com by cli.com (4.1/SMI-4.1) id AA03269; Mon, 8 May 95 11:58:49 CDT Received: (from marks@localhost) by groucho.sarnoff.com (8.6.9/8.6.9) id NAA26578; Mon, 8 May 1995 13:03:55 -0400 Date: Mon, 8 May 1995 13:03:55 -0400 From: "Aaron J. Marks" Message-Id: <199505081703.NAA26578@groucho.sarnoff.com> To: gcl@cli.com Subject: GCL 2.0 and PCL 2.1 on BSDI 2.0 Has anyone successfully compiled GCL 2.0 and PCL 2.1 on a BSDI 2.0 box? I've been able to compile gcl (using FreeBSD mach type) but pcl does not compile. It looks like it's the infamous sprintf fcn bug that FreeBSD had. I am under this assumption because I get the "Dots appeared illegally" error message when trying to load the compiled version of DEPS: ... Loading binary of DEFCLASS... Loading binary of DEFS... Error: Dots appeared illegally. ... Comments/suggestions? Thanks, -Aaron. -- Aaron J. Marks - marks@sarnoff.com "If you steal from one author it's David Sarnoff Research Center plagiarism; if you steal from many Voice: (609)734-2704; Fax: (609)734-2049 it's research." --Unknown From jeff@aiai.ed.ac.uk Mon May 8 15:15:56 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03647; Mon, 8 May 95 15:15:56 CDT Received: from haymarket.ed.ac.uk by cli.com (4.1/SMI-4.1) id AA03619; Mon, 8 May 95 13:36:42 CDT Received: from aiai.ed.ac.uk (skye-alter.aiai.ed.ac.uk [192.41.104.6]) by haymarket.ed.ac.uk (8.6.10/8.6.10) with SMTP id TAA28234; Mon, 8 May 1995 19:36:00 +0100 Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Mon, 8 May 95 19:35:05 BST Date: Mon, 8 May 95 19:35:04 BST Message-Id: <27938.9505081835@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: Re: onscure format quirks To: kr@shell.portal.com, gcl@cli.com In-Reply-To: kr@shell.portal.com's message of Sun, 7 May 1995 17:29:53 -0700 > I have repeatedly used the (format ...) function to print out numbers > in a flexible and variable fashion. I think that the syntax of this > construct is pretty brain-dead and user-unfriendly, and many of the > options are pretty obscure (and so not well tested and debugged, it > seems). So I probably only have myself to blame for having run into > problems with various LISPs by using format. But OTOH there must be > reason why it was included, and there is no reasonable alternative I > know of, so I'll have to use what is available. The short answer is that FORMAT is in Common Lisp because it was in MacLisp and in the various MacLisp successors that lead to Common Lisp, and because it didn't generate enough dislike to get it eliminated or extensively modified. If I recall correctly, Guy Steele is the originator of some of the more elaborate aspects. Anyway, some people think it's neat, and indeed cool, to be able to write loops and case statements in format strings. ^_^ I don't think it's much worse than some of the similar things used in C (printf(), strftime()), but it's true that it has some problems. Anyway, useing the XP pretty-printer's format in AKCL, I get: -> (format nil "~v,'0,X" 8 1234) "000004D2" -> (format nil "~v,'0X" 8 1234) "000004D2" -> (format nil "~v,'0,' X" 8 1234) "000004D2" You say the last one should work like this: > ? (format nil "~v,'0,' X" 8 1234) > " 000 4D2" As you can see, that's not what I get above. Nor is it what I get using the native FORMAT (ie, as opposed to the XP one) in Lucid CL 4.1 or Allegro CL 4.1. -- jeff From kr@shell.portal.com Mon May 8 18:27:04 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03711; Mon, 8 May 95 18:27:04 CDT Received: from nova.unix.portal.com by cli.com (4.1/SMI-4.1) id AA04458; Mon, 8 May 95 17:11:31 CDT Received: from jobe.shell.portal.com (jobe.shell.portal.com [156.151.3.4]) by nova.unix.portal.com (8.6.11/8.6.5) with ESMTP id PAA11907; Mon, 8 May 1995 15:10:45 -0700 Received: from DialupEudora (jobe.shell.portal.com [156.151.3.4]) by jobe.shell.portal.com (8.6.11/8.6.5) with SMTP id PAA15964; Mon, 8 May 1995 15:10:41 -0700 X-Sender: kr@pop.shell.portal.com Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: shell\r\elogin:, 8 May 1995 15:11:15 -0800 To: Jeff Dalton , gcl@cli.com From: kr@shell.portal.com (kr) Subject: Re: obscure format quirks At 19:35 5/8/95, Jeff Dalton wrote: >I don't think it's much worse than some of the similar things >used in C (printf(), strftime()), but it's true that it has >some problems. I think those are also pretty brain-dead. :-) (I did not even mention Fortran :-) >Anyway, useing the XP pretty-printer's format in AKCL, I get: Thanx for sending those datapoints. But what is "XP pretty-printer" and where would I get it ? Greetings Markus Krummenacker From richards@igor.music.qc.edu Mon May 8 20:27:14 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA03729; Mon, 8 May 95 20:27:14 CDT Received: from aaron.music.qc.edu by cli.com (4.1/SMI-4.1) id AA04714; Mon, 8 May 95 19:05:32 CDT Received: from igor by aaron.music.qc.edu (NX5.67e/NX3.0M) id AA04757; Mon, 8 May 95 20:02:17 -0400 Received: by igor.music.qc.edu (NX5.67e/NX3.0S) id AA00326; Mon, 8 May 95 20:03:47 -0400 Date: Mon, 8 May 1995 20:03:44 -0400 (EDT) From: "David N. Richards" X-Sender: richards@igor To: Gcl Mailing List Subject: make breaks Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hi - Trying to build on a NeXTStep 3.2 m68k platform. got the following error during a make (see below) - any help much appreciated!! Dave ========================================== David Richards The Aaron Copland School of Music at Queens College E-mail: richards@aaron.music.qc.edu Voice : 1-(718)-997-3874 ========================================== NeXTunixsave.c:121: `va_alist' undeclared (first use this function) NeXTunixsave.c:121: (Each undeclared identifier is reported only once NeXTunixsave.c:121: for each function it appears in.) NeXTunixsave.c:296: `i' undeclared (first use this function) NeXTunixsave.c:296: `the_commands_len' undeclared (first use this function) NeXTunixsave.c:297: `the_commands' undeclared (first use this function) NeXTunixsave.c:299: `segment' undeclared (first use this function) NeXTunixsave.c:305: `data_size' undeclared (first use this function) NeXTunixsave.c:305: `data_address' undeclared (first use this function) NeXTunixsave.c:306: `bss_size' undeclared (first use this function) NeXTunixsave.c:307: `fdatastart' undeclared (first use this function) NeXTunixsave.c:308: `fdatasize' undeclared (first use this function) NeXTunixsave.c:309: `fgrowth' undeclared (first use this function) NeXTunixsave.c:314: `section' undeclared (first use this function) NeXTunixsave.c:315: `seg' undeclared (first use this function) NeXTunixsave.c:352: `outfd' undeclared (first use this function) NeXTunixsave.c:352: `the_header' undeclared (first use this function) NeXTunixsave.c:355: warning: `return' with a value, in function returning void NeXTunixsave.c:366: warning: `return' with a value, in function returning void NeXTunixsave.c:373: `infd' undeclared (first use this function) NeXTunixsave.c:376: warning: `return' with a value, in function returning void NeXTunixsave.c:378: `size' undeclared (first use this function) NeXTunixsave.c:380: `buf' undeclared (first use this function) NeXTunixsave.c:388: warning: `return' with a value, in function returning void NeXTunixsave.c:399: warning: `return' with a value, in function returning void NeXTunixsave.c:411: `st' undeclared (first use this function) NeXTunixsave.c:414: warning: `return' with a value, in function returning void NeXTunixsave.c:422: warning: `return' with a value, in function returning void NeXTunixsave.c:427: warning: `return' with a value, in function returning void NeXTunixsave.c:430: warning: `return' with a value, in function returning void unixsave.c: In function `init_unixsave': unixsave.c:174: `Lsave' undeclared (first use this function) *** Exit 1 From etokna@etn.ericsson.se Tue May 9 10:14:17 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00420; Tue, 9 May 95 10:14:17 CDT Received: from mailgate.ericsson.se by cli.com (4.1/SMI-4.1) id AA06328; Tue, 9 May 95 08:44:49 CDT Received: from hisoy0 (hisoy0.etn.ericsson.se [193.161.188.2]) by mailgate.ericsson.se (8.6.11/1.0) with SMTP id PAA03844 for ; Tue, 9 May 1995 15:44:39 +0200 Received: from hisoy1.etn.ericsson.se by hisoy0 (5.0/LME-DOM-2.2.3) id AA19866; Tue, 9 May 1995 15:45:08 +0200 From: etokna@etn.ericsson.se (Knut-Haavard Aksnes) Received: by hisoy1.etn.ericsson.se (4.1/client-1.5) id AA02791; Tue, 9 May 95 15:43:45 +0200 Date: Tue, 9 May 95 15:43:45 +0200 Message-Id: <9505091343.AA02791@hisoy1.etn.ericsson.se> To: gcl@cli.com Subject: Bug in gcl-2.1 installation Comments: Hyperbole mail buttons accepted, v3.18.3. Content-Length: 444 The tcl/tk and X11 include and library directories from add-defs is not propagated properly to the makefile in gcl-tk. This makefile has to be edited manually. -- ------------------------------------------------------------------------------- Name: Knut-Håvard Aksnes (ECMA 94) or Knut-Haavard Aksnes (ASCII) Ericsson signature: HI/ETO/X/I KNA Phone: +47 37 05 14 81 Email: etokna@hisoy.etn.ericsson.se (internet) ETO.ETOKNA (memo) From etokna@etn.ericsson.se Tue May 9 10:25:14 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00426; Tue, 9 May 95 10:25:14 CDT Received: from mailgate.ericsson.se by cli.com (4.1/SMI-4.1) id AA06409; Tue, 9 May 95 09:01:36 CDT Received: from hisoy0 (hisoy0.etn.ericsson.se [193.161.188.2]) by mailgate.ericsson.se (8.6.11/1.0) with SMTP id QAA06092 for ; Tue, 9 May 1995 16:01:32 +0200 Received: from hisoy1.etn.ericsson.se by hisoy0 (5.0/LME-DOM-2.2.3) id AA20254; Tue, 9 May 1995 16:02:00 +0200 From: etokna@etn.ericsson.se (Knut-Haavard Aksnes) Received: by hisoy1.etn.ericsson.se (4.1/client-1.5) id AA02945; Tue, 9 May 95 16:00:36 +0200 Date: Tue, 9 May 95 16:00:36 +0200 Message-Id: <9505091400.AA02945@hisoy1.etn.ericsson.se> To: gcl@cli.com Subject: gcl-2.1 fails to compile pcl-gcl-2.1 on solaris. Comments: Hyperbole mail buttons accepted, v3.18.3. Content-Length: 367 GCL core dumps during compilation of slots-boot.lisp on a Sparc 1000 running SunOS 5.3. -- ------------------------------------------------------------------------------- Name: Knut-Håvard Aksnes (ECMA 94) or Knut-Haavard Aksnes (ASCII) Ericsson signature: HI/ETO/X/I KNA Phone: +47 37 05 14 81 Email: etokna@hisoy.etn.ericsson.se (internet) ETO.ETOKNA (memo) From etokna@etn.ericsson.se Tue May 9 10:34:58 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00433; Tue, 9 May 95 10:34:58 CDT Received: from mailgate.ericsson.se by cli.com (4.1/SMI-4.1) id AA06421; Tue, 9 May 95 09:05:12 CDT Received: from hisoy0 (hisoy0.etn.ericsson.se [193.161.188.2]) by mailgate.ericsson.se (8.6.11/1.0) with SMTP id QAA06626 for ; Tue, 9 May 1995 16:05:09 +0200 Received: from hisoy1.etn.ericsson.se by hisoy0 (5.0/LME-DOM-2.2.3) id AA20314; Tue, 9 May 1995 16:05:38 +0200 From: etokna@etn.ericsson.se (Knut-Haavard Aksnes) Received: by hisoy1.etn.ericsson.se (4.1/client-1.5) id AA03004; Tue, 9 May 95 16:04:15 +0200 Date: Tue, 9 May 95 16:04:15 +0200 Message-Id: <9505091404.AA03004@hisoy1.etn.ericsson.se> To: gcl@cli.com Subject: Path to libX11 should be stored for TCL/TK Comments: Hyperbole mail buttons accepted, v3.18.3. Content-Length: 426 (si::tkconnect) fails for me if libX11.so.6.0 isn't in the LD_LIBRARY_PATH, this path should ideally be remembered from the settings in add-defs. -- ------------------------------------------------------------------------------- Name: Knut-Håvard Aksnes (ECMA 94) or Knut-Haavard Aksnes (ASCII) Ericsson signature: HI/ETO/X/I KNA Phone: +47 37 05 14 81 Email: etokna@hisoy.etn.ericsson.se (internet) ETO.ETOKNA (memo) From jeff@aiai.ed.ac.uk Tue May 9 11:19:45 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00470; Tue, 9 May 95 11:19:45 CDT Received: from haymarket.ed.ac.uk by cli.com (4.1/SMI-4.1) id AA06649; Tue, 9 May 95 10:06:51 CDT Received: from aiai.ed.ac.uk (skye-alter.aiai.ed.ac.uk [192.41.104.6]) by haymarket.ed.ac.uk (8.6.10/8.6.10) with SMTP id QAA19507; Tue, 9 May 1995 16:05:42 +0100 Received: from subnode.aiai.ed.ac.uk (bute) by aiai.ed.ac.uk; Tue, 9 May 95 16:04:45 BST Date: Tue, 9 May 95 16:04:45 BST Message-Id: <4003.9505091504@subnode.aiai.ed.ac.uk> From: Jeff Dalton Subject: Re: obscure format quirks To: kr , gcl@cli.com In-Reply-To: kr's message of shell\r\elogin:, 8 May 1995 15:11:15 -0800 > >Anyway, useing the XP pretty-printer's format in AKCL, I get: > > Thanx for sending those datapoints. But what is "XP pretty-printer" and > where would I get it ? XP is a pretty printer written by Dick Waters at MIT. It's the basis for the pretty printer specification in Chapter 27 of Common Lisp the Language 2nd edition. The easiest way to make GCL conform to that spec (which I think is also in the ANSI standard) would be to use XP. It is (or at least was) available on the net. The version I use in AKCL is dated November, 26 1991 (rather old) and slightly modified (by me) for AKCL to avoid calling COMPILE at run-time. Now, _where_ on the net. ... The latest info I have is from 8 feb 93: Dick Waters' XP Lisp Pretty Printer is available by anonymous ftp from merl.com:/pub/xp/ as the files xp-code.lisp, xp-doc.txt, and xp-test.lisp. (It is also available from the Lisp Utilities Repository.) For further information, contact Dick Waters, or . For something more recent, check the Comp.lang.lisp FAQ. For the utilities repository, try http://www.cs.cmu.edu/Web/Groups/AI/html/air.html -- jeff From wavehh!wavehh.hanse.de!cracauer@mail.hanse.de Tue May 9 12:02:42 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00503; Tue, 9 May 95 12:02:42 CDT Received: from ki1.chemie.fu-berlin.de by cli.com (4.1/SMI-4.1) id AA06844; Tue, 9 May 95 10:51:06 CDT Received: by ki1.chemie.fu-berlin.de (Smail3.1.28.1) from mail.hanse.de (134.100.239.2) with smtp id ; Tue, 9 May 95 16:43 MEST Received: from wavehh.UUCP by mail.hanse.de with UUCP for gcl@cli.com id ; Tue, 9 May 95 16:43 MET DST Received: by wavehh.hanse.de (4.1/SMI-4.1) id AA02969; Tue, 9 May 95 16:42:01 +0200 From: cracauer@wavehh.hanse.de (Martin Cracauer) Message-Id: <9505091442.AA02969@wavehh.hanse.de> Subject: gcl@cli.com: Please post announcements To: gcl@cli.com Date: Tue, 9 May 1995 16:42:00 +0200 (MET DST) Reply-To: cracauer@wavehh.hanse.de X-Mailer: ELM [version 2.4 PL24] Content-Type: text Content-Length: 377 Hello, I'd suggest that announcements for new releases of gcl are placed on this list (gcl@cli.com). I see versions passing by in bug reports, but never see any noification of a new release or beta release. Thanks. -- %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Martin Cracauer (private address) Tel.: +49 40 / 522 18 29 Fax.: +49 40 / 522 85 36 From richards@igor.music.qc.edu Tue May 9 12:43:23 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00543; Tue, 9 May 95 12:43:23 CDT Received: from aaron.music.qc.edu by cli.com (4.1/SMI-4.1) id AA07005; Tue, 9 May 95 11:27:29 CDT Received: from igor by aaron.music.qc.edu (NX5.67e/NX3.0M) id AA00751; Tue, 9 May 95 12:24:04 -0400 Received: by igor.music.qc.edu (NX5.67e/NX3.0S) id AA02536; Tue, 9 May 95 12:25:44 -0400 Date: Tue, 9 May 1995 12:25:41 -0400 (EDT) From: "David N. Richards" X-Sender: richards@igor To: Gcl Mailing List Subject: make breaks Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hi - Building gcl2.0 on a NeXT m68k running 3.2 - and get the following: cc -DVOL=volatile -I/Temp/Gcl/gcl-2.1/o -fwritable-strings -c -O -I../gcl-tk -I../h/ main.c main.c:554: only 3 args to macro 'RETURN' (4 expected) *** Exit 1 Stop. What should be the 4th value?? Many thanks in advance - Dave ========================================== David Richards The Aaron Copland School of Music at Queens College E-mail: richards@aaron.music.qc.edu Voice : 1-(718)-997-3874 ========================================== From tkunze@ccrma.Stanford.EDU Tue May 9 16:15:00 1995 Return-Path: Received: from cli.com by ftp.cli.com (4.1/SMI-4.1) id AA00576; Tue, 9 May 95 16:15:00 CDT Received: from ccrma.Stanford.EDU (ccrma.Stanford.EDU) by cli.com (4.1/SMI-4.1) id AA07668; Tue, 9 May 95 14:43:27 CDT Received: from cmn18 by ccrma.Stanford.EDU (NX5.67e/NeXT-1.0) id AA16996; Tue, 9 May 95 12:43:25 -0700 From: tkunze@ccrma.Stanford.EDU (Tobias Kunze) Message-Id: <9505091943.AA16996@ ccrma.Stanford.EDU > Received: by cmn18.Stanford.EDU (NX5.67d/NX3.0X) id AA12106; Tue, 9 May 95 12:43:24 -0700 Date: Tue, 9 May 95 12:43:24 -0700 Received: by NeXT.Mailer (1.100) Received: by NeXT Mailer (1.100) To: "David N. Richards" Subject: Re: make breaks Cc: Gcl Mailing List > Hi - Building gcl2.0 on a NeXT m68k running 3.2 - and get the following: > cc -DVOL=volatile -I/Temp/Gcl/gcl-2.1/o -fwritable-strings -c -O > -I../gcl-tk -I../h/ main.c > main.c:554: only 3 args to macro 'RETURN' (4 expected) > *** Exit 1 > Stop. > > What should be the 4th value?? Many thanks in advance - > Dave > Dave- it is a typo as far as i remember. And--bad news--it's only the first error in a row (about 50 or so) :( I finally gave up. I couldn't get it to compile under Next 3.3 (black). I tweaked the sources here and there, but no way. However, if somebody managed to get it to compile, I'd be glad to hear about! -Tobias