The Project Revision Control System, PRCS, is the front end to a set of tools that (like CVS) provide a way to deal with sets of files and directories as an entity, preserving coherent versions of the entire set.
Abstractly, PRCS presents you ("the user") with the abstraction of named projects that have multiple (project) versions, each of which is a set of files, arranged into a directory subtree. A project is the collection of all project versions of that project. Each project version contains one distinguished file, called the (project) version descriptor, in the top directory of the subtree, which contains a description of the files included in that particular version. For a project named `P', this file has the name `P.prj' (and so is also called a "`.prj' file"). The operations provided allow you to perform the following actions:
In addition, you may add and subtract the files contained in a new project version, or move files from one subdirectory to another by editing the working copy of the `.prj' file before checking in the new version. See section Descriptors.
PRCS keeps all checked-in project versions in a projects
repository. See section Repository. At any given
time, you fix a prevailing repository by setting the environment
variable PRCS_REPOSITORY (which defaults to `$HOME/PRCS'),
and do not explicitly mention it in any
commands. In general, you should modify the projects repository only with
the PRCS command or other designated utilities, or the results
will be unpredictable.
PRCS uses the approach found in CVS (Concurrent Version System): unlike version control systems such as RCS and SCCS, PRCS locks projects in the repository only while it is actually running (in order to make operations on projects atomic). One simply checks in new versions as desired; the system will warn you if others have made parallel changes to the major version of the project you are working on that may have to be merged with changes you have made.
Where PRCS differs from CVS (and, indeed, other source-code control systems of the author's acquaintance) is in its attempt to provide a particularly simple model to the user. A project is a collection of files, period. Although the current implementation of PRCS actually uses RCS in its implementations, none of the details of RCS are of any consequence to the user.
PRCS is somewhat unusual in that I wrote the bulk of this reference manual before implementation started. I am greatly indebted to Josh MacDonald for condensing it from vaporware into real software, and to the Berkeley PATH project and its sponsors for initially paying his way.
Emphasized terms within a definition are defined later in this section.
# % ^ - _ + = , .
that does not begin with a hyphen, equal-sign, or period.
"In DOS, you write this \"C:\\BIN\\FOO\""
prcs
checkin records a list of project versions from which it was derived
(in the Parent-Version and Merge-Parents attributes of the descriptor).
This establishes a graph of project versions, each being derived from
its parents. A project version P0 is an ancestor of version
P1, if one can reach P0 from P1 by zero or more steps
along this directed, acyclic ancestry graph (thus, P1 is an
ancestor of itself). See section Parent-Version attribute, section Merge-Parents attribute, and section Prcs merge.
prcs checkin replaces with appropriate text depending on
the keyword (such as the time and date of checkin, the login of the
account that checks in a version of a file, the name of the file, or the
name of the project). See section Keywords.
For an overview of the entire process of project source-code maintenance with PRCS, read the sections below in order. It uses a continuing example of a project called P1.
Before you can use PRCS at all, you must establish a PRCS
repository (a directory).
PRCS will create and initialize the repository if it
does not exist the first time you attempt to use it. PRCS takes
the name of the repository from the environment variable
PRCS_REPOSITORY, which must be defined before running PRCS (if
undefined, it defaults to `$HOME/PRCS').
For example, to have a repository named `PRCS_FILES' in your home directory, execute
% setenv PRCS_REPOSITORY ~/PRCS_FILES
in the C-shell or
$ export PRCS_REPOSITORY
$ PRCS_REPOSITORY=${HOME}/PRCS_FILES
in the Bourne shell. You may want to put this command into your .cshrc
or .profile initialization file. In the rest of this document, we will
assume that PRCS_REPOSITORY is properly defined.
When you first run PRCS after
having set this environment variable, it will create directory
`PRCS_FILES' in
your home directory, assuming it does not already exist.
If you attempt to run PRCS with
PRCS_REPOSITORY set to a directory not in the right format, it will report an
error.
Once your repository is established, you should generally be careful not to modify it except through PRCS, except that changing the group, protections, or name of the repository is harmless.
To create the first version of a completely new project named P1, go to the desired working directory and use the following command:
% prcs checkout P1
Assuming there is no prior project P1, the only result of this command is to create a working file `P1.prj' in the current directory; the projects directory and its contents are not modified. In `P1.prj', you will initially find the following lines:
;; -*- Lisp -*- (Created-By-Prcs-Version 1 1 0) (Project-Description "") (Project-Version P1 0 0) (Parent-Version -*- -*- -*-) (Version-Log "Empty project.") (New-Version-Log "") (Checkin-Time "Sun, 31 Dec 1995 01:54:11 -0700") (Checkin-Login jmacd) (Ignore ()) (Project-Keywords) (Files ; This is a comment. Fill in files here. ; For example: (prcs/checkout.cc ()) ) (Merge-Parents) (New-Merge-Parents)
The second line identifies the project version from which these files came: in this case from a version 0.0 of the project named P1. Minor version 0 exists implicitly for every major version of a project; it is empty, containing no files and a standard project descriptor, as shown above. The third line indicates the (in this case, nonexistent) version from which version 0.0 was derived.
When you create working files, say main.c and doc/foobar.1, you can add them to the version of P1 you are creating by editing the Files entry of `P1.prj' to look like this:
(Files (main.c ()) (doc/foobar.1 ()) )
The Files list identifies the names of the files as they appear relative to the working directory in which you start. The empty lists after each name contain information indicating where these files came from in the repository; since these are new files, the lists are empty to indicate that your files did not come from the repository, but are new. The information in these lists has a format internal to PRCS, and you need not worry about it. Even though the project descriptor `P1.prj' itself is not explicitly listed in the Files list, it is implicitly considered one of the files in the project.
When you have many files to add to the Files list--as for example when you have an existing directory structure full of files that you wish to bring under version control--it is inconvenient to do so by hand. PRCS can add them for you automatically with the command
% prcs populate P1
which finds all files in the current directory and its subdirectories and adds them to the Files list in `P1.prj'. At this point, you can edit `P1.prj' to remove files that you don't want to save (e.g., compiled object files or backup files).
It is a good idea at this point to add some commentary to the `.prj' file before check-in:
(Project-Description "Sample PRCS Project") (New-Version-Log "checkin main.c and doc/foobar.1")
The Project-Description information, by default, is carried over from version to version of a project (so that you usually won't change it after the first version). The New-Version-Log becomes the Version-Log upon check-in and any old version log from a prior version is deleted. Thus, you may use the Version-Log entries of the checked-in versions for any notes about the reasons for the new version or its contents. You can use newlines in the string constants used to hold the commentary, and make them arbitrarily long. See section Definitions.
To check in a new project version of project P1, use the command
% prcs checkin P1
If there is only one `.prj' file in the current directory, you may
even leave off the P1 argument. This command looks at the file
`P1.prj' to find out the major version of P1 from which the current
working files were copied and adds a new minor version to it. In this
case, since `P1.prj' records no major version name, PRCS uses
the default major version name, `0'. Thus, this check-in will create
version 0.1 (minor version 1 of major version 0) of P1. The command
prcs checkin modifies `P1.prj', before checking it in, to
look something like this:
;; -*- Lisp -*- (Created-By-Prcs-Version 0 13 8) (Project-Description "Sample PRCS Project") (Project-Version P1 0 1) (Parent-Version P1 0 0) (Version-Log "checkin main.c and doc/foobar.1") (New-Version-Log "") (Checkin-Time "Sun, 31 Dec 1995 02:10:24 -0700") (Checkin-Login jmacd) (Ignore ()) (Files (main.c (P1/0_main.c 1.1 644)) (doc/foobar.1 (P1/0_foobar.1 1.1 644)) ) (Merge-Parents) (New-Merge-Parents)
As you can see, PRCS has added the version identification, a timestamp, a record of the account performing the checkin, and some internal identification information about the two files now in the project. It has also changed the New-Version-Log declaration to the Version-Log (which is removed when you modify this version and check-in a new one).
Your working directory now contains copies of the files in version 0.1 of P1. You might now add a new working file, say `header.h', and modify `main.c'. You add the line
(header.h ())
to the Files list. You will also want to create a New-Version-Log entry and perform another checkin. The new version (0.2) of `P1.prj' might now contain
;; -*- Lisp -*- (Created-By-Prcs-Version 0 13 8) (Project-Description "Sample PRCS Project") (Project-Version P1 0 2) (Parent-Version P1 0 1) (Version-Log "") (New-Version-Log "") (Checkin-Time "Sun, 31 Dec 1995 02:13:00 -0700") (Checkin-Login jmacd) (Ignore ()) (Files (main.c (P1/0_main.c 1.2 644)) (doc/foobar.1 (P1/0_foobar.1 1.1 644)) (header.h (P1/0_header.h 1.1 644)) ) (Merge-Parents) (New-Merge-Parents)
The files of version 0.2 of P1 have now been saved; you may destroy
the working directory and later restore it with prcs checkout.
It is not necessary to check in all your working files. By providing a list of files and directories after the project specifier, you can limit the working files considered to those listed:
% prcs checkin P1 doc
copies only the working files in doc into the new version of P1; for other files listed in `P1.prj', PRCS uses the versions identified by the internal file identifiers in the Files list of `P1.prj'. Thus, if these identifiers are unchanged in your working `P1.prj' file since it was checked out, the same files versions that they identify will be used in the new version of P1 (the working versions, however, may differ in the values of certain keywords. See section Keywords.)
If there is any file which is not in (Files) entriy in .prj file,
prcs checkin aborts. In order to disable this feature, you
have to add the following line into .prj file.
(CompleteCheckin "false")
If you want to exclude specific files from checkin, such as .o files or backup files of editor, you can specify a rule in (Ignore (pattern)) attribute in .prj file. See section Ignore attribute.
The command prcs checkout creates working copies of the files
(and directories) in a project version, usually rooted in the current working
directory. For example,
% prcs checkout -r0.1 P1
checks out version 0.1 (minor version 1 of major version 0) of project P1, creating the subdirectory doc, if needed. If you didn't specify a major version name when you created the project it defaulted to '0' and this is the initial version with no files. You needn't be quite so specific in specifying the version:
% prcs checkout -r1 P1
or
% prcs checkout -r1.@ P1
checks out the latest minor version of major version 1 of P1. If you use only numerals for your major versions, you may even leave off the -r altogether to get the latest version checked into the repository.
It is not necessary to check out an entire version at once; you may specify selected directories or files instead. For example
% prcs checkout -r0.2 P1 main.c header.h
checks out just the file main.c from version 0.2 of P1, and
% prcs checkout -r0.1 P1 doc
checks out the doc directory (and all files contained in and below it) from version 0.1 of P1.
If you have modified a working file, say main.c, and want to see how it differs from the file from which it was checked out, use
% prcs diff P1 main.c
By specifying versions, you can perform other comparisons. To compare the working version of main.c with that in version 0.1:
% prcs diff -r0.1 P1 main.c
To compare main.c in versions 0.1 and 0.3:
% prcs diff -r0.1 -r0.3 P1 main.c
To compare the working version of main.c with the latest in version 0:
% prcs diff -r0.@ P1 main.c
In place of `main.c', you may use any list of files and directories; each will be compared with files from the indicated versions. Mentioning a directory causes comparison of all files under that directory. Leaving off the file name list (e.g.,
% prcs diff P1 % prcs diff -r0.1 P1 % prcs diff -r0.@ P1
) performs the comparison for all files in the project versions.
The output of prcs diff indicates differences between files,
and also cases where a file is present in one version (or working
version) but not another.
If you have a working version that you have been modifying, you may have
added files to it that are not yet reflected in the project descriptor
(`.prj' file). Normally, prcs diff will ignore these
files. To get a full comparison, you must first add these additional
file names to the working
project descriptor (see section Adding files to a project). If you don't really intend to add
these files immediately, you can save and restore the original working project
descriptor:
% cp P.prj tmp-P.prj % prcs populate P % prcs diff ... output from diff % mv tmp-P.prj P.prj
or (if you haven't touched the project descriptor since checking it out), you can simply restore the project descriptor:
% prcs populate P % prcs diff ... output from diff % prcs checkout P P.prj
After modifying the working files created by checking out the
then-latest version of P1 (1.15, let us say), you may try to check them
back
in and discover that other minor versions of P1 version 1 have been
added since 1.15 (prcs checkin will tell you this, and prcs
diff or prcs merge -n will give details).
One way to deal with these differences is to merge the contents of your working files with the new version in the repository:
% prcs merge P1
By default, the checked-in version used by this command is the latest
minor revision of the major revision of P1 indicated in your working
`.prj' file (for this example, let's say version 1.17). The
command prcs merge will prompt you for what action you want to
take in cases where there are changes between the latest version and the
version from which you checked out the working files.
The merge does not create and check in a new version; it merely updates
your working files. It will also update your `P1.prj' file to
indicate that it came from the version specified--implicitly or
explicitly--by the merge subcommand, and you may perform a
check-in as a new version.
Successive checkins will produce a sequence of minor versions of the same major version of a project. The major version name is useful for indicating such things as releases of a piece of software or "branches" of a project on which development will proceed independently. To create a new major version, specify it during check-in with the -r option. For example, to convert your working files into the first minor version of a new branch of project P1 called `MyBranch', use the command
% prcs checkin -rMyBranch P1
Modify at will; your modifications will be part of major version `MyBranch', and will be invisible to others using the repository unless they specifically ask for that major version. To merge back into the main branch (or any other), just use `prcs diff' and `prcs merge', as above:
% prcs merge -r0.@ P1
To copy an entire checked-in version into a new major version, use a sequence of commands like the following.
% prcs checkout -r0.@ P1 P1.prj % prcs checkin -r1 P1 P1.prj
Only the project-version descriptor actually gets checked out and in. Because the file list for `P1.prj' remains unchanged, the two versions share all the same files (except for those files that contain certain keywords. See section Keywords.)
The simplest possible use of PRCS is just to use it to keep a single thread of project versions, thus allowing you to "roll back" at any time, and allowing multiple people to do maintenance on a single system simultaneously.
To start things off with a new project, P1, type
% prcs checkout P1
in the working directory where you are developing your project. (If you haven't yet established a repository, PRCS will create one.) If the project was already underway before you created a PRCS project for it, you will probably want to start off by bringing your existing files under version control. To do so, type
% prcs populate P1
Edit the file `P1.prj' to remove any files you don't want to control (like `.o' files, back-up files, and executables). You can also specify the list of files "in the population" on the command line. The files in any directories in this list will be included recursively. In effect, specifying no files is equivalent to specifying "." (the current directory).
% prcs populate P1 file1 file2
Whenever you want to checkpoint your project, type
% prcs checkin
from the root working directory (the one with `P1.prj' in it). If someone else has checked in files in the meantime, you will be notified of discrepancies, which you can handle by typing
% prcs merge
from the root working directory and (after it's done), editing the files.
Whenever it comes time to begin working on a deposited version of the project, you can go to a clean directory and check out the latest version with
% prcs checkout P1
Make your modifications and check back in as before.
First checks for the existence of a PRCS repository, and prompts
you to create one if there is none present. If the environment variable
PRCS_REPOSITORY is set, it is taken as the name of the
repository. Otherwise, PRCS looks for a directory called
`PRCS' in your home directory. For descriptions of the possible
subcommands, see section Subcommands.
Every version in a repository has a version name, which is unique within that version's project. A version name consists of a major version name and a minor version name, separated by a period. The major version name may be any valid label. The minor version name is always a positive integer numeral, without leading 0's. The -r option specifies versions within a project by a version specifier.
A version specifier has a form similar to that of a version name: M.N, where M is a major version name and N is a minor version name. Certain values for M and N have special meanings:
@ means "the latest in the repository (or 0 if none)."
When @ is used for
the major version, it indicates the numerically largest major version
name (of the given project) among all those major version names that are
non-negative integers with no leading zeroes.
In subcommand descriptions, arguments appearing in square brackets ([]) are optional. An argument followed by `...' indicates "one or more".
Options have both a short form (hyphen followed by one letter) and a long form (introduced by two hyphens). Certain options, defined in the first subsection below, are standard: valid on all commands. Other options are listed with the individual subcommands to which they apply, with their long forms in parentheses.
A project operand identifies the name of the project to be acted on (a simple label, denoted P below), the name of the project-version descriptor file (`P.prj'), and a root working directory (denoted D below). In the usual case, project is just the name of a project (a simple label), in which case P is project and D is the current directory (`.'). The full story is as follows:
Some subcommands allow you to omit the project operand. In those cases, it defaults to the current directory (`.'), so that by the preceding rules, there must be a single `.prj' file in the current directory, which then implicitly supplies the project name.
When the directory D determined by these rules is not the current working directory `.', the effect is as follows:
The file-or-dir operands restrict a subcommand to operate only on specific files within a project version. Each must be the name of a file or directory in the relevant project version. A directory operand stands for all files under that directory (i.e., in it or, recursively, under its subdirectories). The default file-or-dir operand is the working directory specified by the project operand (see section Specifying Projects). Usually, this is simply the current directory (`.'), specifying everything under the project directory.
This section collects definitions of options used in
PRCS subcommands. Many options have both a short form (e.g., -f)
and a GNU-format long form (e.g., --force); in these cases, the
long form is listed immediately below the short form. In the
descriptions below, a Standard option is one that is legal on all
subcommands, and an Option is valid on one or more subcommands.
prcs execute to indicate that the specified command is to
be executed just once, with all files as arguments.
prcs populate to indicate that files that are not present
in a working directory are to be removed from the Files list.
prcs uncompress, causes the files in a project to be
uncompressed on the spot, rather than incrementally as they are needed.
prcs diff to compare keywords, instead of stripping them
before comparison.
prcs execute to indicate that the names of the files to be
processed, or the options attached to them, must match pattern.
prcs diff, produces comparisons against empty ("new") files
where a file is missing in one of the versions being compared. If you
intend to use the -N option and later apply the diffs you should make
sure to always use one of the context diff formats. See the
PRCS_DIFF_OPTIONS environment variable See section Environment Variables.
prcs populate to add all files with the :no-keywords
attribute.
prcs execute to indicate that the names of files to be
processed, or the options attached to them, must not match pattern.
prcs execute to supply file data as the standard input.
prcs diff not to output changes
in the project files of each version.
prcs execute to indicate that directories are to be
listed in pre-order: directory names first, then their non-directory
contents, then subdirectories.
PRCS_REPOSITORY environment variable.
prcs merge, allowing a
skilled user finer control over the process. See section Prcs merge.
prcs checkin to set the Version-Log on the command line.
You will be prompted if there is an existing New-Version-Log already.
This option overrides the PRCS_LOGQUERY environment variable.
ADDITIONAL OPTIONS: `-r', `-p', `-u', `-P'
Create a working copy of a version of the specified files and
directories in project rooted in the current directory (creating
any subdirectories that don't exist). You may use the option `-r'
to select a version. The default major and minor version specifiers are
both @.
You will be warned when a check-out would change the contents of an existing working file, and interactively given the choice of replacing the file with the version being checked out, or leaving the existing file alone. Existing working files that do not differ from corresponding project files are not modified (their changed and modified dates do not change). The `-f' option will cause checkout to overwrite all files that differ without asking.
When PRCS overwrites a working file that is a
symbolic link, then (unless the :symlink option is specified for
that file), it will overwrite the file that the link references, without
changing the link itself.
With the `-u' option, PRCS will instead remove the link
and create a new file.
If a working project-version descriptor for project exists, then the project-version descriptor is treated as an ordinary file for purposes of check-out. A check-out with a non-defaulted file-or-dir list that does not contain the descriptor therefore leaves the descriptor unchanged. Further, if the `-P' option is supplied, the project descriptor will not be created at all.
Files that are checked out are restored to the protections recorded for them at check-in time, usually modified (on UNIX systems) by the user's file mode creation mask (see the UNIX manual entry for umask(1)). With the `-p' option, this mask is ignored, and the command restores the files' protections to exactly their state at check-in.
Returns a status code of 0 if all goes well, and non-zero if there are errors.
ADDITIONAL OPTIONS: `-d', `-u', `--no-keywords'
Add each file in the file-or-dir list to the Files list of the `.prj' file identified by project, without modifying the repository. For each directory in file-or-dir, recursively adds all files in that directory structure (or the directory itself, if it is empty). The file-or-dir operand list defaults to the current directory.
Without the `-d' option, any files already in the Files list remain there unchanged. With `-d', populate will prompt you about removing files from the Files list that are listed there as appearing under one of the directories given in the file-or-dir list, but are not found in the working directory. With `-d' and `-f', it will remove them silently.
For example, suppose the file `P.prj' lists files `C', `D1/A', `D1/B', `D2/A', `D2/B'. Suppose a working directory contains files `A', `D1/B', and `D2/B'. Then the command
prcs populate -d P
would remove files `C', `D1/A', and `D2/A' from the Files list of `P.prj', and add file `A'. For the same initial `P.prj' file, the command
prcs populate -d P D1
would remove `D1/A' from `P.prj', and would add nothing.
New files are added with null internal file identifiers.
Populate attempts to fill in the :symlink and
:directory options.
It also applies a heuristic test to each file
added to guess whether it is a non-text file; for each file it judges to be a
non-text file, it includes the :no-keywords option for that file
(see section Files attribute).
ADDITIONAL OPTIONS: `-u'
Remove each file in the file-or-dir list from the Files list of the `.prj' file identified by project, without modifying the repository. For each directory in file-or-dir, recursively removes all files in that directory structure (or the directory itself, if it is empty). The file-or-dir operand list defaults to the current directory, but since removing all files is probably not something the user wants to do, it queries first if no file-or-dir is supplied.
ADDITIONAL OPTIONS: `-r', `-u', `-j', `--version-log'
Create a new version of the specified project in the repository. Copy all files listed in the Files attribute of the descriptor (`project.prj'), taking those specified in file-or-dir operands and the descriptor itself from the working files. Also, modify the descriptor to reflect the contents of the new version. See section Descriptors.
If file-or-dir arguments are provided, they must name files and directories listed in the Files attribute. Only those working files are checked in; other files listed in the project descriptor are carried over from the file versions listed for them in the Files attribute. It is an error for these other files to have null internal file identifiers in the `.prj' file.
If you want to exclude specific files from checkin, such as .o files or
backup files of editor, you can specify a rule in (Ignore (pattern))
attribute in .prj file. See section Ignore attribute.
When prcs checkin finds any file which is not either (Files) files
list or (Ignore) files list, the command shows warning and aborts.
This will prevent your newly created but not yet prcs add'ed
`.c' or `.h' files in your working directory from being
forgotten in checkin phase. If you do not want prcs checkin
to look at the (Ignore) list, you have to add:
(CompleteCheckin "false")
line in your .prj file.
A check-in records the access protection on each file (reading it from
the working file).
It records symbolic links with
the :symlink option and empty directories with a
:directory option. You could have, for example, a Files
attribute as follows:
(Files (empty () :directory) ; an empty directory (pointer () :symlink) ; a symbolic link (realfile () :no-keywords) ; the only real file in the list )
PRCS reads
the contents of a symbolic link from the link at check-in time.
It is thus wise to use relative pathnames inside the project.
Directories with a :directory option
need not be actually empty, but this ensures their creation at checkout.
It is an error to have a :no-keywords option in the same file
as a :symlink, or :directory
option.
The new version will have the latest minor version number for the specified major version. The `-r' option may specify a major version; any minor version specified is ignored. The `-r' option defaults to `-r.@', thus taking the major version from the `.prj' file.
Let LV denote the last minor version on the branch being checked into. A checkin is considered safe if the nearest common ancestor of LV and the working version is LV (see section Parent-Version attribute). This is to insure that the version being checked in is a descendant of the version at the head of the branch before checkin. Otherwise, changes at the head of the branch are clobbered. Unless the `-f' option is present, PRCS will warn you if a checkin is not safe and allow you to abort.
With `-l' present, PRCS will list which files are modified, inserted, deleted, or renamed, relative to the real working version prior to checkin. With `-L', PRCS will list all these changes in addition to listing which files are unmodified.
Returns a status code of 0 if all goes well, and non-zero if there are errors.
ADDITIONAL OPTIONS: `-r' (once or twice), `-k', `-N', `-P'
Compare the (partial) contents of two repository versions of a project or of a repository version and a working version. When file-or-dir arguments are present, they restrict comparison to the specified files or subdirectory trees. The project argument may be defaulted if there is a single file with extension `.prj' in the current working directory, in which case PRCS takes its name (without `.prj') as the project name.
The options may specify zero, one, or two versions (using
`-r' options). Specifying no `-r' options is equivalent to supplying
the single option `-r.', the version from which the current working
directory was checked out. A `-r' option that specifies only a major
version (leaving off the minor version and its preceding period)
implicitly has a minor version of @.
When one version specifier is given, the
files in the indicated version are compared to the working files. The
command's output notes any discrepancies in the contents of
identically-named files, or in files present in one version but not the
other. Two
version specifiers cause the same comparison, but between two checked-in
versions of the project.
Normally, diff canonicalizes keyword instances in the files it
compares, removing the keyword values and leaving just the keyword
name. As a result, two different versions of a file that differ only
in their keyword values will compare as equal. The `-k' option
causes diff to compare keyword values as well.
If a file is present only in one of the project versions being compared,
diff will normally just announce that such a file occurs only
in one of the versions. With the `-N' option, it will produce the
same output it would if it instead treated the non-existent version as
an empty file. This is useful when producing patch files that include
new files as well as changed ones. Its effect is essentially the same
as that of the same option in GNU diff when it compares directories.
Files in the two versions being compared are paired by their names and their internal-file families (see section Files attribute, for a description of internal-file families). That is, PRCS will assume that two files in the different versions correspond if they have the same name or the same internal-file family. This makes it possible to have conflicts (for example, when one switches the names of two files in a project descriptor); in those cases PRCS will ask the user to resolve the conflict.
The diff-options, if provided, are any options acceptable to
GNU diff. @xref{(diff)Top}. The environment variable
PRCS_DIFF_OPTIONS, if set, supplies a default set of whitespace
separated diff-options.
If the file-or-dir is defaulted and `-P' is not supplied, or if file-or-dir contains the project descriptor, PRCS outputs differences between project descriptors.
Returns a status code of 0 if there were no differences, 1 if there were differences, and 2 if there were problems.
ADDITIONAL OPTIONS: `-r', `-u', `-s'
Find differences between working files and those of the selected version of project, and (unless `-n' is specified) try to modify the working files to reconcile them. The default major and minor versions are `-r.@'.
More precisely, the comparison takes place between three versions of project: the working version (called PW in what follows), the common version (PC) and the selected version (PS).
PW is the working version specified in the checked-out descriptor and consists of all files listed therein. PS is the version against which PRCS will reconcile changes, identified by the (possibly defaulted) `-r' option. A common version is chosen according to the following algorithm:
Each version, including the working version, has a list of parent versions. The parent versions establish an ancestry graph that is directed and contains no cycles. A version is an ancestor of another if it can be reached by following zero or more parents (see section Parent-Version attribute and section Merge-Parents attribute). PC is chosen from the set of all versions that are ancestors of PW and PS. If amongst this intersection, there exists a unique nearest common ancestor, then it is PC. If there is no such version, PRCS will complain and ask the user for help.
Files in the three versions are matched up both by name and by internal-file family (see section Files attribute). That is, PRCS will assume that two files in different versions correspond if they have the same name or the same internal-file family. This makes it possible to have conflicts (for example, when one switches the names of two files in a project descriptor); in those cases PRCS will ask the user to resolve the conflict.
A prcs merge will report discrepancies and (unless `-n' is
specified) take actions as directed. The possible actions are as
follows.
The `-f' option causes default actions to each discrepancy (see below). Otherwise, PRCS will ask the user to specify one of the possible actions. All of the actions except n modify the descriptor, either to remove the file from the Files list, or to modify its internal file identifier to the one from PS.
Like prcs diff, prcs merge canonicalizes files to strip
keyword values before comparing them for discrepancies (see section Prcs diff).
In the descriptions that follow, saying that files are "equivalent" means
that they are identical when canonicalized.
PRCS detects the following discrepancies:
When PRCS overwrites a working file that is a
symbolic link, then (unless the :symlink option is specified for
that file), it will overwrite the file that the link references, without
changing the link itself.
With the `-u' option, PRCS will instead remove the link
and create a new file.
Whenever PRCS is going to replace a working file as a result of one of the actions above, it first moves the original file into the `obsolete' directory and gives it a unique name.
A partial merge takes place when not all files in the three versions are considered for merging. This can happen if a merge is aborted or the file-or-dir option limits the files under consideration. When this happens, PRCS records which files have been merged and their names in the working project's New-Merge-Parents attribute (see section Merge-Parents attribute).
A merge is considered complete when each file in each version has been considered once. A complete merge may be achieved by several partial merges. To prevent reconsidering the same file more then once, PRCS will not consider any set of files containing a file that was already considered in a previous merge against the same selected version unless the `-s' option is specified.
In general, the `-s' option allows the user to turn off all the
safety features built into prcs merge. It will allow a user to
merge files more than once, override the choice of an effective working
version, restart a complete merge, and start a new merge without
completing a previous merge.
By default, PRCS uses the GNU diff3 command to perform 2- or 3-way
file merges when you choose the m action (@xref{(diff)top}). It
is possible, however, to supply your own merge command. By setting the
PRCS_MERGE_COMMAND environment variable to a program name, PRCS will call the
named command instead of diff3 (see section Environment Variables).
ADDITIONAL OPTIONS: `-u'
Replace keywords in the selected files, according to the version named in the project file. Files whose contents would not change under keyword substitution are not modified. With the `-n' option, simply reports files that will change as a result of keyword replacement. See section Keywords.
When PRCS overwrites a working file that is a
symbolic link, then (unless the :symlink option is specified for
that file), it will overwrite the file that the link references, without
changing the link itself.
With the `-u' option, PRCS will instead remove the link
and create a new file.
You can use this command to update keywords in working files after
performing a prcs checkin, which does not modify keyword values.
ADDITIONAL OPTIONS: `-r', `--sort'
Print information about versions of the named project. For each version listed, prints the project name, major and minor version identifiers, date and time checked in, and the user who checked in the version. With the `-l' option, also prints out the version logs and each project description. With `-L', also prints out the files constituting the version, their attributes, and their unkeyed MD5 checksum; the file-or-dir options in this case restrict the files or directories that are listed.
The `-r' option is special in this command in that it may contain
any of the shell filename-generation pattern characters defined for the
Bourne shell,
sh: *, which matches any string of 0
or more characters; ?, which matches any single character;
[...], which matches any of the enclosed characters; and
[!...], which matches any character except one of the enclosed
characters. These special pattern characters must be escaped to get
them past the shell. The default for
both the major and minor versions is
*. The `-r' option is still interpreted as a separate
major and minor version; that is, `-rFoo.*' matches `Foo.1', but
not `Foo.Bar.1'.
The `--sort=TYPE' option sets the type of sorting that the info command should use. If TYPE is `date', versions are sorted by individual version date, otherwise the default, `version', is to sort by major version creation date.
ADDITIONAL OPTIONS: `-r' (once or twice), `-l'
The changes subcommand helps you find what versions modify a specific file or files. For example, suppose you fixed a bug a long time ago and now it shows up again in the current version. You would like to know what project versions modify a certain file so you can find out who undid your bug-fix and blame them!
Like prcs diff, the options may specify zero, one, or two
versions (using `-r' options). Specifying no `-r' options is
equivalent to supplying the single option `-r.', the version from
which the current working directory was checked out. A `-r' option
that specifies only a major version (leaving off the minor version and
its preceding period) implicitly has a minor version of @. When
one version specifier is given, only that version is scanned for
changes. Two version specifiers cause a range of versions to be
scanned. In the two version case, one version must be an ancestor of
the other, and reversing the order of the two versions will reverse the
order of the output.
When file-or-dir arguments are present, they restrict the change report to the specified files or subdirectory trees. Otherwise all files are reported.
For each specified version that modifies a file (in the reporting set)
relative to one of its parent versions, changes prints the
project name, major and minor version identifiers, date and time checked
in, and the user who checked in the version. Then it prints the parent
version and a listing of all the selected files that were modified,
including: addition, deletion, rename, type change, symlink change, or
version change. For regular files, changes also prints the
number of lines added and removed by the modification. If a project
version does not modify any of the selected files, nothing is printed.
When the `-l' option is present, changes also prints the
version-log associated with each version.
Perform various administrative subfunctions that generally have no effect on the abstract state of the repository. The available functions are as follows.
Encourages PRCS to save space in storing the data for project, even at the expense of slower processing. This command may take some time to execute. Interrupting it in the middle is safe, although it may lead to somewhat erratic space-saving behavior.
ADDITIONAL OPTIONS: `-i'
Encourages PRCS to save time in processing commands against project, even at the expense of using more space--perhaps considerably more. With the `-i' option, PRCS will immediately expand the entire representation of project. Without the `-i' option, PRCS will increase its use of space only as needed. If disk space should run out, PRCS will leave the repository in a correct state.
With the optional operand, interactively sets read and write permissions to all versions of project, in the repository, and (on systems that provide it) sets the group to which the files belong. This command allows the owner of a project in a repository to restrict the access that others have to its contents. One can specify that these files may only be read or may not be accessed at all.
Without a project operand, interactively sets read and write permissions to the repository itself, both for members of the repository directory's group, and for all other users. One must have at least read access to the repository to look at or check out any project (in addition to having appropriate access to the project itself). You must have write access to the repository to create new projects in it.
Rebuilds internal repository data on project, possibly removing file storage for deleted project versions and verifying the integrity of all project versions. With the `-l' option, reports deletions. Unless there has been damage to the repository, this command generally has no visible effect on the notional contents of the repository.
Create a new project in the repository containing no (non-empty) versions.
Performing a prcs checkout on a non-existent project creates a
blank working project descriptor without modifying the repository. The
first prcs checkin for that project then creates a new entry in
the repository. Normally, no other explicit initialization for the
project's entry in the repository is
necessary. However, with this method it is impossible to perform a
prcs access before the first check-in, because there is no record
of the new project
in the repository. With prcs init, you can create a repository
entry (containing no versions) that you can modify with prcs
access.
Deletes a repository entry, permanently removing all versions and data from the repository.
Lists all projects in the repository.
Renames a repository entry, similar to renaming during prcs
unpackage.
Print out all compiled-in values and their values, possibly altered by
environment variables or command-line arguments.
Also verifies that the executables named by $RCS_PATH, or its
compiled-in default, are valid and checks their version numbers to see
that they are not outdated.
Sometimes, it is desirable to package up and move or copy an entire
PRCS project, rather than a single version of a project (discussed
elsewhere in section Importing Project Versions.) One can move an entire
repository using the usual UNIX tools for copying, moving, distributing,
and packaging directories. For individual projects, it is preferable to
use the package and unpackage subcommands in order to
insure that all of PRCS's internal structure is maintained.
ADDITIONAL OPTIONS: `-z'
Creates a plain file named packagedfile, containing all the data in project. A value of `-' for packagedfile denotes the standard output. With `-z', the output file is compressed (in gzip format).
Creates project in the repository to be a copy of the project from
which packagedfile was created.
The file packagedfile must have been created by prcs
package (with or without the `-z' option). If project is
omitted, it defaults to the name of the project from which
packagedfile was created. If
project exists in the repository, the user is prompted to delete
the old copy from the repository.
A value of `-' for packagedfile denotes
the standard input.
With `-n', it merely reports the name of
the project that would be added.
If project differs from the original name as stored in packagedfile, the project is renamed. Renaming a project revises the history of the project; the result is as if the project had always been named project, with one exception: When checked out, project descriptor will have a comment inserted making note of the original project name.
ADDITIONAL OPTIONS: `-r', `-P', `--pre', `--all', `--pipe' `--match pattern', `--not pattern'
Execute `command arg ...,' suitably modified as described below, for the name of each file and directory in file-or-dir in the specified version, filtered by any `--match' and `--not' operators as described below. This elaborate command is intended to facilitate efficient, open-ended extension of the functions of PRCS. Each directory, whether or not explicitly named in the project, is included once, either in pre- or post-order.
With the `--all' option, command is executed once. Otherwise, it is executed once per file name.
The pattern options filter the selected files. To be selected, the entry in the Files attribute for a given file must satisfy the `--match' pattern (if any) and not satisfy the `--not' pattern, if any. Each pattern is a regular expression; an entry satisfies the pattern if either the name or any of the `:' options matches the pattern.
For each execution of command, PRCS first replaces the following strings wherever they appear in command and in each arg:
{}
{options}
:implicit-directory option, whether or not it appears separately in
the Files attribute. These directories appear whether or not the files
containing them were on the command line (otherwise, it would be
impossible to get just the directories).
The project file is included with the :project-file attribute.
With the `--all' option, all of these colon-option strings are
concatenated together, with the strings for different files separated by
unquoted blanks.
{file}
{file} is the same as
{}.
It is the empty string
for a directory or a file carrying the :symlink option.
With the `--all' option, all of these file names are
concatenated together, separated by unquoted blanks.
(In contexts where curly braces are special to your shell, you will have to escape them.) After these substitutions, PRCS invokes command, which must be an executable file. It looks for command in the same directories as the shell (using the PATH environment variable). When `--' and what follows are omitted, they default to
-- /bin/echo {}
which simply echoes the names of all files and directories in the selected version, one per line.
One small glitch: prcs execute uses the current directory in
effect when it is invoked as the current directory seen by
command. This is true, even when the project operand
specifies a subdirectory. For example,
% cd /usr/users/foo % prcs execute D/P -- pwd
will print `/usr/users/foo' once for each file listed in `/usr/users/foo/D/P.prj', and not `/usr/users/foo/D', as the general description of the project operand might otherwise suggest (see section Specifying Projects). This allows you to do something like the following:
% ln -s . P-1.0 % tar -cvf P-1.0.tar `prcs execute --not :.*directory P-1.0`
for making tarfiles containing all files in a project in an appropriately named subdirectory for distribution.
If the `--pipe' option is present, then the contents of the file (as would be contained in the file named by `{file}') is supplied as the standard input. The `--pipe' and `--all' options are incompatible. Any arg whose replacement contains unquoted blanks (introduced by the `--all' option) is divided at those blanks into separate arguments (even in the case where substitution results in a null string, that null string will be broken out as a separate argument). Each of the resulting arguments, is passed directly to the program specified by command as a single string, with no further processing by any shell.
With the `--all' option, PRCS will invoke command only once. The order of the lists of arguments replacing each pattern in command is consistent, so that, for example, the options for the first file name in the replacement for `{}' is the first string in the replacement for `{options}'. For example, if a project contains the files `A', `B/Q', and `B/R', then
prcs execute --all . -- foo -a{}-b "{options}"
will execute (in effect) the single command
foo -aA B/Q B/R B-b "" "" "" ":directory"
With the `--pre' (for pre-order) option the directory name is listed first, then the non-directory file names within that directory, then the subdirectories and their contents. Otherwise, subdirectories and their contents are listed first, then the names of its non-directory files, then the name of the directory itself.
With `-n', the command prints the commands that will be executed without actually executing them.
Files can be moved(renamed) within a project and within the working directory tree simultaneously. It works basically like mv except that files can't be moved from outside of the working directory tree into the working directory tree.
Note: prcs move was implemented and is maintained by 15-410 class staff after version 1.3.2.
Delete the indicated version of project. It will no longer be possible to retrieve the given version. Therefore, this is a command that you should use very sparingly, if at all.
Deleting a version does not necessarily release the space used by the files it contains, since these may be mentioned in other versions. At various times, PRCS will note what space can actually be released and do so.
Minor version numbers increase monotonically, regardless of deletions, and are never re-used.
The projects repository is a directory that stores the checked-in copies
of some set of projects. You usually specify it to PRCS by
setting the PRCS_REPOSITORY environment variable, or by using the
`-R' (`--repository') option. It defaults to `$HOME/PRCS'.
The precise contents of the repository--the relationship between the files that appear there and the abstract set of version contents presented by PRCS---is deliberately left undefined. You should not count on it being stable from release to release.
To increase the speed of checkin and other subcommands, PRCS
maintains redundant information, both in the repository and in working
directories. Specifically, you may find files with names such as
`.P.prcs_aux' in your working directory, which contain information
that allow PRCS to make a quick, heuristic determination of which
files you have modified from their repository versions. It is safe to
remove (but not to modify) these auxiliary files.
A project-version descriptor is a file that defines the contents of the
project. For project with name P, the project-version descriptor
(or just `descriptor') is called `P.prj' and is kept at the
root of the project directory tree. When you check in a project, the
project descriptor indicates what files need to be included, which have
keywords that need to be expanded, and from which previous versions of
the those files (if any) the working files were derived. A prcs
checkin then creates and checks in an updated descriptor file.
The sections below detail the syntax of the project descriptor file and the standard attribute entries that may appear in it.
The syntax of the project descriptor file is a slight variant on Lisp. It consists of `S-expressions' (this is the terminology established by Lisp, so we'll use it here). In general, an S-expression has one of the following forms
label string constant (s-exprs)
Where `s-exprs' is (recursively) a sequence of 0 or more s-expressions, separated from each other by whitespace and optionally separated from the surrounding parentheses by whitespace (`whitespace' is any non-empty sequence of blanks, tabs, form feeds, and newlines). We've described the syntax of labels and string constants elsewhere (see section Definitions).
You may include comments in the descriptor; these consist of a semicolon (outside of a string) followed by all characters up to and not including the next newline or the end of the file. Comments have no effect on the behavior of PRCS and are copied from one version of a project descriptor to the next.
The non-nested S-expressions that appear in a descriptor (those that aren't part of some larger expression) each start with an attribute name (a label). The valid attribute names are described in the following sections.
This indicates that the descriptor is (or is modified from) the
descriptor for version M.N of project P.
M must be a valid
major version name and N must be a valid minor version name
(see section Version Names and Specifiers).
The prcs checkin command sets the Project-Version of the
checked-in descriptor. See section Version Names and Specifiers.
This attribute is intended to provide a title or short description of
the project. By default, prcs checkin leaves it undisturbed.
For example,
(Project-Description "Sample PRCS Project")
The version log is stored with the descriptor of each version of a
project. A prcs checkin initializes it from the New-Version-Log
attribute, removing any existing Version-Log attribute, and setting the
New-Version-Log to the empty string. The version log is intended to
describe the changes made in this version from its predecessor, or to
contain other comments about it.
For example,
(Version-Log "Added header.h") (New-Version-Log "Modified main.c")
These attributes record the login of the user performing the PRCS
check-in that created this version and the time at which it was
performed. They are set by prcs checkin.
For example,
(Checkin-Time "Sun, 12 Jun 1994 20:00:00 -0700") (Checkin-Login hilfingr)
Each of the 0 or more arguments of the Files attribute denotes one of the constituents of a version. The file arguments give the pathnames of files relative to some root directory. The internal-file-id arguments contain information used by PRCS to identify versions of individual files.
If F is one of the file arguments, then
no component of F may be `.' or `..'. Unless one of the options
listed for F is :directory (see below), no other file
argument in the Files attribute may begin with `F/'.
The precise format of an internal file identifier is not defined, and is subject to change whenever PRCS is revised. Two properties, however, are guaranteed:
prcs checkin will assign it an internal file identifier
(possibly unchanged from that of the working version) in the same
internal-file family. If a working file has a null internal file
identifier (`()'), prcs checkin will put it in a new
internal-file family, otherwise unpopulated. Internal-file families are
useful
for certain kinds of version comparisons and merges.
The options arguments denote 0 or more labels indicating special properties of the file. At the moment, four are defined.
:no-keywords
:symlink
:directory
:tag=label
prcs execute
subcommand (see section Prcs execute). There may be any number of
:tag options attached to a file.
The project descriptor (`.prj' file) is not explicitly included in the Files attribute. Nevertheless, it is implicitly one of the files in each project version.
For example,
(Files (empty () :directory) ; an empty directory (pointer () :symlink) ; a symbolic link (realfile () :no-keywords) ; the only real file in the list )
This attribute is used to inform prcs populate of files that it
should not add to the Files list of the project descriptor. It is
included as a convenience; you
can accomplish the same effect by editing the Files list after running
populate.
Each of the pattern arguments is a regular expression (as
used by grep). Any path name that prcs populate finds
that matches one or more pattern arguments is not added to the Files
list (existing names on the list are not affected).
For example, the following filters out object (`.o') files, library archive (`.a') files, core files, files generated by TeX, Postscript (`.ps') files that occur in subdirectory `doc', Emacs backup files, and the executable file `a.out':
(Ignore ("\\.o$" "\\.a$" "~$" "^a.out$" "^core$"
"\\.dvi$" "\\.aux$" "\\.log"
"^doc/.*\\.ps"))
Also, this attribute is used to inform prcs checkin of files
that it should ignore. Probably it is a good idea to include `.o'
files or backup files into this list, since those files should
not be checked into the repository. When prcs checkin finds any
file which is not either (Files) files list or (Ignore) files list,
the command shows warning and aborts. This will prevent your newly
created but not yet prcs add'ed `.c' or `.h' files
in your working directory from being forgotten in checkin phase.
If you do not want prcs checkin to look at the (Ignore) list,
you have to add:
(CompleteCheckin "false")
line in your .prj file.
Note: This attribute was previously named Populate-Ignore, but has been renamed Ignore by Tadashi Okoshi after version 1.3.2.
The prcs merge command uses the New-Merge-Parents attribute to
record the history of updates to each file.
Each merged-against-version indicates a selected version that was
merged into the
working files, each effective-working-version indicates the
effective working version used to select the common version,
complete indicates whether the merge against
merged-against-version was completed,
and each file-action records what action
(delete, add, replace, merge, no action) merge took for each file.
This
information is useful in preventing confusion when the user interrupts a
merge operation and
then re-executes the merge.
The file-actions entries contain data recording each merge action that takes place. Since the filenames in each merge action need not be the same (see section Files attribute), PRCS records each of up to three filenames for each merge action or an empty list (`()') where a file was not present. In addition, PRCS records which action was taken on each set of files.
The prcs checkin command copies any New-Merge-Parents attribute in the working project
descriptor into a Merge-Parents attribute in the new (checked-in)
project descriptor. It discards any existing Merge-Parents attribute in
the working file.
Each merge-parent in a checked-in version is considered as a parent version. Similarly, each new-merge-parent in a working version is considered as a parent version (see section Parent-Version attribute).
This attribute, along with the Merge-Parents attribute described above, tracks the history of a project version: Parent-Version records the version from which a project version originally descended, and Merge-Parents records versions against which it was merged.
This has the same form as the Project-Version attribute
(see section Project-Version attribute), except that for an initial version
(unrelated to any prior version),
each of P, M, and N is -*-.
The Parent-Version defines the ancestors of
a version in the repository as follows:
Given two versions in a repository, any version that is an ancestor to both is called a common ancestor. A common ancestor whose distance from both is minimal is called a nearest common ancestor. It is possible for there to be more than one nearest common ancestor. If so, commands attempting to determine the nearest common ancestor will either abort or ask the user for help.
Each
prcs checkin sets Parent-Version to name the Project-Version
attribute of the version from
which the newly
checked-in version is derived. Additionally, it sets the New-Merge-Parents
attribute to the Merge-Parents attribute.
The Project-Keywords attributes allows the user to define new keywords and their values, which will behave exactly like the builtin keywords (see section Keywords). Each keyword must be unique and must not match one of the builtin keywords. All files contained in the project version described by a descriptor containing a Project-Keywords list will be keyed with each keyword. Each value is either a label or a string and may not contain a newline character. For example, you could include such project meta-data as state labels, release labels, and version numbers:
(Project-Keywords (ReleaseMajorVersion 1)
(ReleaseMinorVersion 1)
(ReleaseMicroVersion 0)
(ReleaseVersion "$ReleaseMajorVersion$.$ReleaseMinorVersion$.$ReleaseMicroVersion$"))
(ProjectState "Release")
(RealAuthor "Josh")) ; in case the login doesn't
; provide enough information.
Files in a project may contain instances of keywords that (in effect)
prcs checkin will replace with certain information.
Conceptually, upon checkin, each keyword instance in every file
of the project (not just
those listed in the checkin subcommand)
is updated as described below with the value of its
keyword in the new project version. As a performance consideration,
however, this is implemented in such a way that partial checkins only
have to examine the files that are listed in the checkin subcommand (see section Keywords and File Identifiers).
There are two forms of keyword instance: simple and formatted. A simple keyword instance has one of the forms
$keyword$ $keyword: data$
This syntax is strict: PRCS does not recognize a keyword when there
is whitespace on either side (separating it from the dollar sign that
precedes it or the dollar sign or colon that follows). There may not be
newlines anywhere in the keyword instance. If the keyword is one that
PRCS recognizes, other than Format, prcs checkin
replaces the instance with an instance of the second form, with the data
part depending on the particular keyword. Formatted keyword instances
allow a more flexible replacement format.
PRCS understands the following built-in keywords, in addition to any declared in the Project-Keywords attribute. Others are ignored (their instances are left unchanged). Instances of the Source and ProjectHeader keywords change whenever the file is moved from one directory to another within a project. Instances of Project, ProjectDate, ProjectAuthor, ProjectMajorVersion, ProjectMinorVersion, ProjectVersion, and ProjectHeader can change even when the file containing them does not, since they depend on the project, not the file. These keywords are known collectively as project-related keywords.
Keyword replacement is recursive.
Replacement of a keyword takes place on a string which is either the
quoted contents of a Format keyword or the defined value of a
regular keyword instance. When this string contains an instance of a
keyword which is not currently being replaced, it recursively outputs
the value of that keyword. Processed text is not reexamined. A keyword
instance is not replaced when it is appears during a recursive
replacement of the same keyword, to prevent loops.
One minor glitch in the above description. Since the `$' character delimits keyword instances, tt is necessary to prevent `$' characters from being inserted as the value of a non-formatted keyword. All text after the first `$' would otherwise end up being inserted into the document. For example, if the keyword K had value "I have a `$' character!" were replaced:
$K: I have a $ character!$
Then, the next replacement would result in:
$K: I have a $ character!$ character!$
and so on. Therefore, when spurious `$' characters are found
inside the contents of a regular keyword value (not Format), they
are replaced with the `|' character.
Sometimes, you want keyword replacement data to appear without the
leading $keyword: and trailing `$'. A formatted keyword
instance allows this. An appearance in your file of
$Format: "string"$ string
will cause prcs checkin to replace the next line of the source
file (i.e., the line after the second `$') with string, after
first making modified keyword substitutions in it. Specifically, each
instance of $keyword$ in string (for which
keyword is one of the recognized keywords) is replaced by just the
data that would be placed to the right of the `:' in a simple
keyword instance. The Format instance itself is not
altered--only the subsequent line. For example, if a file in a project contains
/* $Format: "static char* version = \"$ProjectVersion$\";"$ */ static char* version = "x.x";
then the version of the file checked in for version 1.2 of the project will contain
/* $Format: "static char* version = \"$ProjectVersion$\";"$ */ static char* version = "1.2";
The format string may not contain newlines. It may contain quotation
marks if they are stropped with a backslash, as in the example above.
In general, the backslash quotes the succeeding character. Any keyword
instances on a list after a Format instance are ignored (that is,
they are left unchanged by keyword substitution). Any nested instance
of Format is also left unchanged.
The actual content of a working file upon checkout depends on two things: the internal file identifier recorded for it and the contents of the project descriptor. The internal file identifier designates a particular "raw" file contents stored in the repository that never changes once it is first created. The checked out file consists of this contents modified by keyword replacement. If the same file identifier is used for two files in two different projects, the checked-out working files will differ in the values supplied for these keywords. This is a subtle point, but it is significant when you copy a file identifier from one project to another.
For example, suppose that file F in version 1.2 of project P contains the line
/* $ProjectVersion: 1.2$ */
and that its entry in the Files attribute of P.prj is
(F (P/0_F 1.1))
If you decide to rename F to G for the next version of P, you can do so by changing its line in the Files attribute to
(G (P/0_F 1.1))
When you check the project back in, creating version 1.3, this same line will be retained as its file identifier (assuming, that is, you made no changes to G or its former incarnation, F). However, in version 1.3 of P, G will contain
/* $ProjectVersion: 1.3$ */
That is, PRCS actually performs all keyword replacement on check-out (think of it as lazy replacement). It only makes a user-visible difference in the handling of internal identifiers. It also has a considerable effect on the performance of partial checkins (those that contain a list of specific files and directories to check in, and copy all unmentioned files from file identifiers listed in the Files attribute). Because PRCS does not have to examine the files excluded from the file list for keywords, it can process them very quickly.
You can use PRCS to "bundle together" a set of projects (so that they may be released together, for example). Suppose that P is the name you want to give to the (new) combined project, with P1, P2, and P3 the names of the subprojects. To create the initial version of P, take the following steps.
% prcs checkout -r... Pi Pi.prj(replacing `...' with the appropriate version specifier). This checks out the `.prj' file for Pi into Pi.
To create a working copy of this version of P (including working copies of all its subprojects' files), use the following (Bourne shell) commands (I'll assume that 0.1 is the version name for P).
$ prcs checkout -r0.1 P $ for i in P1 P2 P3; do > (cd $i; prcs checkout $i) > done
Obvious modifications of this procedure will serve to create updated versions of P.
With a little care, we can automate this process a bit further. Suppose that we establish a convention that we tag each subproject descriptor file in the Files list of P:
(Files ... (P1 (...) :tag=project) (P2 (...) :tag=project) (P3 (...) :tag=project) ... )
This allows a general-purpose version of the checkout procedure above, one that does not require that we remember the names of all the subprojects:
% prcs execute -r0.1 --match :tag=project P -- prcs checkout {}
PRCS allows you to create a sequence of project versions out of a sequence of releases of a source directory tree. You may want to do this, for example, if you are installing successive versions of some system imported from elsewhere and wish to be able to retrieve old versions, while still saving space by storing only the changes between successive versions. (See section Moving and Copying Projects, if you want to move or copy an entire project from one repository to another.)
The process is simple. For concreteness, let us use the name P
as the project name. For the first imported version, we create P
and populate it with the first version of the imported source files. In
the directory that contains the imported files, use the commands
% prcs checkout P % prcs populate P # Edit P.prj # % prcs checkin P
The editing of `P.prj' might consist of setting the
Project-Description and New-Version-Log attributes, and
possibly adding
:no-keywords options to some files.
When you receive a new release of P, place it in a directory
structure and, at the top directory of that structure, use the commands
% prcs checkout P P.prj % prcs populate -d P # Edit P.prj # % prcs checkin P
With the `-d' option, the prcs populate command will add any
new files found in the working directory, but not listed in
`P.prj', and will remove any files listed in `P.prj', but not
present in the working directory.
Again, you will probably want to edit the New-Version-Log
attribute.
Imported software sources often have their own version designations, and
you may want to have the PRCS major version identifiers include
all or part of these designations. Suppose the first such imported release is
called 1.0-beta. In the working directory containing this release,
you might type
% prcs checkout -r1.0-beta.0 P % prcs populate P # Edit P.prj # % prcs checkin P
On receiving a new release from outside, say 1.1-beta, you would put
it under a working directory and, from that directory, issue
the commands
% prcs checkout -r1.0-beta. P P.prj % prcs populate -d P # Edit P.prj # % prcs checkin -r1.1-beta. P
Sometimes, you will have local modifications or enhancements to a piece
of imported software, and want to bring your local version up to date
with successive parallel releases of the imported software.
To do this,
use the procedure above for keeping the imported
releases. Put your own local releases in another major version, let's
say Local.
To start with,
you'll use the initial imported release:
% prcs checkout -rV P P.prj # Edit the project # % prcs checkin -rLocal P.prj
Here, V is whatever PRCS version number you have assigned to the
vendor's (the imported) version.
Continue making your local modifications in major revision Local.
When a
new external release comes in, first make
sure the most recent version of your local version is checked in.
Then put the new imported version into a fresh directory structure,
populate a project file, and check it in as outlined above.
For concreteness,
let's say that your most recent local version is designated Local.5, and
the newly-checked-in imported version is designated Vendor.3.
In the working directory for Vendor.3 (which you have just checked in), issue the command
% prcs merge -rLocal.@ -n P
to see where the discrepancies are (or use prcs diff), and use
% prcs merge -rLocal.@ P
to merge in the revisions since the last external release with the local changes since the last release. Edit the files to resolve conflicts, and edit `P.prj' as usual. Finally, check the result back in with
% prcs checkin -rLocal P
which will create version Local.6. Its parent versions are Local.5 and Vendor.3. Alternatively, you could check out the local version, Local.5, and merge against the vendor branch
% prcs merge -rVendor.@ P
and checkin as Local.6. The resulting parent versions are the same in either case, with the sense of all the merge operations reversed.
Sometimes, you will find it desirable to rename the files you import and re-arrange their directory structure in your local version. You probably only want to perform this restructuring on the Local branch of the project, not the Vendor branch. Initially, you will create your first local version like this:
% prcs checkout -rV P P.prj # Rename files # # Edit P.prj to change the names in the Files attribute so # as to correspond to your renaming. % prcs checkin -rLocal P.prj
Editing `P.prj' here consists simply in changing the file names, not the internal file identifiers. For example, if the file `doc/mogrify.1' in branch V is to be called `man/mogrify2.1' in the Local branch, you would rename the file:
% mkdir man % mv doc/mogrify.1 man/mogrify2.1
and then you would edit the line for `doc/mogrify.1' in `P.prj', which might read
(doc/mogrify.1 (P/0_foobar.1 1.1 644))
originally, into
(man/mogrify2.1 (P/0_foobar.1 1.1 644))
The internal file identifier, therefore, records the correspondence between `doc/mogrify.1' in the V branch and `man/mogrify2.1' in the Local branch. When either of these files is updated and checked in, its new internal file identifier will be in the same internal-file family; thus the correspondence between the files is not lost.
The concept of an internal-file family (see section Files attribute) was
introduced to allow prcs diff and prcs merge to "do
the right things" (that is, compare or merge against the appropriate files)
even in the face of such restructuring. For example, when comparing a Local
version to a Vendor version, prcs diff will compare `doc/mogrify.1'
from the Vendor version with `man/mogrify.1' in the working files of your
Local version.
Certain information that PRCS needs is not likely to be the same
for different users and installations, and is inconvenient or
inappropriate to specify with every command. In such cases, PRCS
takes the information from environment variables, where they are
defined, and uses some implementation-defined default otherwise.
PRCS checks first the users's environment, then the builder's
compile-time environment, and finally uses the documented default. Here
is a listing of the environment variables that PRCS queries, and
their meanings. Run prcs config to see a listing of all the
environment variables PRCS is aware of and their values.
-j option is not specified.
/dev/null is used. To get
the default behavior, you could use the script performing the
following:
diff3 -maE -Lworkinglabel working -Lcommonlabel common \ -Lselectedlabel selected > output
The return value is interpreted as with diff3
(@xref{(diff)Top}). Return 0 to indicate no conflicts, 1 to indicate
conflicts, and 2 to indicate trouble. Warning: as the result of
a symlink, output and working
files may be the same! So, the above
command will sometimes result in an exploding file because gdiff
overwrites itself as it merges.
--plain-format on the command
line.
RCS_PATH directories it looks for in an installation-specific place.
Version 2, June 1991
Copyright (C) 1989, 1991 Free Software Foundation, Inc. 675 Mass Ave, Cambridge, MA 02139, USA Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed.
The licenses for most software are designed to take away your freedom to share and change it. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change free software--to make sure the software is free for all its users. This General Public License applies to most of the Free Software Foundation's software and to any other program whose authors commit to using it. (Some other Free Software Foundation software is covered by the GNU Library General Public License instead.) You can apply it to your programs, too.
When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for this service if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs; and that you know you can do these things.
To protect your rights, we need to make restrictions that forbid anyone to deny you these rights or to ask you to surrender the rights. These restrictions translate to certain responsibilities for you if you distribute copies of the software, or if you modify it.
For example, if you distribute copies of such a program, whether gratis or for a fee, you must give the recipients all the rights that you have. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights.
We protect your rights with two steps: (1) copyright the software, and (2) offer you this license which gives you legal permission to copy, distribute and/or modify the software.
Also, for each author's protection and ours, we want to make certain that everyone understands that there is no warranty for this free software. If the software is modified by someone else and passed on, we want its recipients to know that what they have is not the original, so that any problems introduced by others will not reflect on the original authors' reputations.
Finally, any free program is threatened constantly by software patents. We wish to avoid the danger that redistributors of a free program will individually obtain patent licenses, in effect making the program proprietary. To prevent this, we have made it clear that any patent must be licensed for everyone's free use or not licensed at all.
The precise terms and conditions for copying, distribution and modification follow.
NO WARRANTY
If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms.
To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively convey the exclusion of warranty; and each file should have at least the "copyright" line and a pointer to where the full notice is found.
one line to give the program's name and an idea of what it does. Copyright (C) 19yy name of author This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
Also add information on how to contact you by electronic and paper mail.
If the program is interactive, make it output a short notice like this when it starts in an interactive mode:
Gnomovision version 69, Copyright (C) 19yy name of author Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'. This is free software, and you are welcome to redistribute it under certain conditions; type `show c' for details.
The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, the commands you use may be called something other than `show w' and `show c'; they could even be mouse-clicks or menu items--whatever suits your program.
You should also get your employer (if you work as a programmer) or your school, if any, to sign a "copyright disclaimer" for the program, if necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the program `Gnomovision' (which makes passes at compilers) written by James Hacker. signature of Ty Coon, 1 April 1989 Ty Coon, President of Vice
This General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Library General Public License instead of this License.
This document was generated on 17 September 2003 using texi2html 1.56k.