Re: CVS updates on committers list...

Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:

> "Larry Rosenman" <ler(at)lerctr(dot)org> writes:
> > I'm referring to the actual commit messages.
>
> It *would* be awfully nice if the pgsql-committers traffic were one
> message per commit, instead of one per directory touched per commit.
> This has been suggested before, but nothing got done ...

This is easy to set up, for an appropriate definition of ``easy''.

I have included two files below, commit_prep and log_accum. Do
cvs co CVSROOT
Copy the files into the CVSROOT directory. Check the definition of
$MAILER near the top of log_accum. Everything else should be fine.
Do this:
cvs add commit_prep
cvs add log_accum

Edit the file checkoutlist, and add these lines:

commit_prep Won't be able to do mail logging.
log_accum Won't be able to do mail logging.

Edit the file commitinfo, and add something like this line:

DEFAULT /usr/bin/perl $CVSROOT/CVSROOT/commit_prep -r

Edit the file loginfo, and add something like this line (replace
MAILINGLIST with the mailing address to which you want log messages to
be sent):

DEFAULT /usr/bin/perl $CVSROOT/CVSROOT/log_accum -m MAILINGLIST -s %s

Then:
cvs commit

Then test it.

Good luck. I didn't write these scripts, but I've used them in a
number of places.

Ian

commit_prep:
==================================================
#!/usr/bin/perl
# -*-Perl-*-
#
# $Id: commit_prep,v 1.1 1998/12/01 03:18:27 ian Exp $
#
# Perl filter to handle pre-commit checking of files. This program
# records the last directory where commits will be taking place for
# use by the log_accum.pl script. For new files, it forces the
# existence of a RCS "Id" keyword in the first ten lines of the file.
# For existing files, it checks version number in the "Id" line to
# prevent losing changes because an old version of a file was copied
# into the direcory.
#
# Possible future enhancements:
#
# Check for cruft left by unresolved conflicts. Search for
# "^<<<<<<<$", "^-------$", and "^>>>>>>>$".
#
# Look for a copyright and automagically update it to the
# current year. [[ bad idea! -- woods ]]
#
#
# Contributed by David Hampton <hampton(at)cisco(dot)com>
#
# Hacked on lots by Greg A. Woods <woods(at)web(dot)net>

#
# Configurable options
#

# Constants (remember to protect strings from RCS keyword substitution)
#
$LAST_FILE = "/tmp/#egcscvs.lastdir"; # must match name in log_accum.pl
$ENTRIES = "CVS/Entries";

# Patterns to find $Log keywords in files
#
$LogString1 = "\\\$\\Log: .* \\\$";
$LogString2 = "\\\$\\Log\\\$";
$NoLog = "%s - contains an RCS \$Log keyword. It must not!\n";

# pattern to match an RCS Id keyword line with an existing ID
#
$IDstring = "\"@\\(#\\)[^:]*:.*\\\$\Id: .*\\\$\"";
$NoId = "
%s - Does not contain a properly formatted line with the keyword \"Id:\".
I.e. no lines match \"" . $IDstring . "\".
Please see the template files for an example.\n";

# pattern to match an RCS Id keyword line for a new file (i.e. un-expanded)
#
$NewId = "\"@(#)[^:]*:.*\\$\Id\\$\"";

$NoName = "
%s - The ID line should contain only \"@(#)module/path:\$Name\$:\$\Id\$\"
for a newly created file.\n";

$BadName = "
%s - The file name '%s' in the ID line does not match
the actual filename.\n";

$BadVersion = "
%s - How dare you!!! You replaced your copy of the file '%s',
which was based upon version %s, with an %s version based
upon %s. Please move your '%s' out of the way, perform an
update to get the current version, and them merge your changes
into that file, then try the commit again.\n";

#
# Subroutines
#

sub write_line {
local($filename, $line) = @_;
open(FILE, ">$filename") || die("Cannot open $filename, stopped");
print(FILE $line, "\n");
close(FILE);
}

sub check_version {
local($i, $id, $rname, $version);
local($filename, $cvsversion) = @_;

open(FILE, "<$filename") || return(0);

@all_lines = ();
$idpos = -1;
$newidpos = -1;
for ($i = 0; <FILE>; $i++) {
chop;
push(@all_lines, $_);
if ($_ =~ /$IDstring/) {
$idpos = $i;
}
if ($_ =~ /$NewId/) {
$newidpos = $i;
}
}

if (grep(/$LogString1/, @all_lines) || grep(/$LogString2/, @all_lines)) {
print STDERR sprintf($NoLog, $filename);
return(1);
}

if ($debug != 0) {
print STDERR sprintf("file = %s, version = %d.\n", $filename, $cvsversion{$filename});
}

if ($cvsversion{$filename} == 0) {
if ($newidpos != -1 && $all_lines[$newidpos] !~ /$NewId/) {
print STDERR sprintf($NoName, $filename);
return(1);
}
return(0);
}

if ($idpos == -1) {
print STDERR sprintf($NoId, $filename);
return(1);
}

$line = $all_lines[$idpos];
$pos = index($line, "Id: ");
if ($debug != 0) {
print STDERR sprintf("%d in '%s'.\n", $pos, $line);
}
($id, $rname, $version) = split(' ', substr($line, $pos));
if ($rname ne "$filename,v") {
print STDERR sprintf($BadName, $filename, substr($rname, 0, length($rname)-2));
return(1);
}
if ($cvsversion{$filename} < $version) {
print STDERR sprintf($BadVersion, $filename, $filename, $cvsversion{$filename},
"newer", $version, $filename);
return(1);
}
if ($cvsversion{$filename} > $version) {
print STDERR sprintf($BadVersion, $filename, $filename, $cvsversion{$filename},
"older", $version, $filename);
return(1);
}
return(0);
}

#
# Main Body
#

$id = getpgrp(); # You *must* use a shell that does setpgrp()!

# Check each file (except dot files) for an RCS "Id" keyword.
#
$check_id = 0;

# Record the directory for later use by the log_accumulate stript.
#
$record_directory = 0;

# parse command line arguments
#
while (@ARGV) {
$arg = shift @ARGV;

if ($arg eq '-d') {
$debug = 1;
print STDERR "Debug turned on...\n";
} elsif ($arg eq '-c') {
$check_id = 1;
} elsif ($arg eq '-r') {
$record_directory = 1;
} else {
push(@files, $arg);
}
}

$directory = shift @files;

if ($debug != 0) {
print STDERR "dir - ", $directory, "\n";
print STDERR "files - ", join(":", @files), "\n";
print STDERR "id - ", $id, "\n";
}

# Suck in the CVS/Entries file
#
open(ENTRIES, $ENTRIES) || die("Cannot open $ENTRIES.\n");
while (<ENTRIES>) {
local($filename, $version) = split('/', substr($_, 1));
$cvsversion{$filename} = $version;
}

# Now check each file name passed in, except for dot files. Dot files
# are considered to be administrative files by this script.
#
if ($check_id != 0) {
$failed = 0;
foreach $arg (@files) {
if (index($arg, ".") == 0) {
next;
}
$failed += &check_version($arg);
}
if ($failed) {
print STDERR "\n";
exit(1);
}
}

# Record this directory as the last one checked. This will be used
# by the log_accumulate script to determine when it is processing
# the final directory of a multi-directory commit.
#
if ($record_directory != 0) {
&write_line("$LAST_FILE.$id", $directory);
}
exit(0);
==================================================

log_accum:
==================================================
#!/usr/bin/perl
# -*-Perl-*-
#
# Perl filter to handle the log messages from the checkin of files in
# a directory. This script will group the lists of files by log
# message, and mail a single consolidated log message at the end of
# the commit.
#
# This file assumes a pre-commit checking program that leaves the
# names of the first and last commit directories in a temporary file.
#
# Contributed by David Hampton <hampton(at)cisco(dot)com>
#
# hacked greatly by Greg A. Woods <woods(at)web(dot)net>

# Usage: log_accum.pl [-d] [-s] [-M module] [[-m mailto] ...] [-f logfile]
# -d - turn on debugging
# -m mailto - send mail to "mailto" (multiple)
# -M modulename - set module name to "modulename"
# -f logfile - write commit messages to logfile too
# -s - *don't* run "cvs status -v" for each file

#
# Configurable options
#

# Set this to something that takes "-s"
$MAILER = "/bin/mail";

# Constants (don't change these!)
#
$STATE_NONE = 0;
$STATE_CHANGED = 1;
$STATE_ADDED = 2;
$STATE_REMOVED = 3;
$STATE_LOG = 4;

$LAST_FILE = "/tmp/#egcscvs.lastdir";

$CHANGED_FILE = "/tmp/#egcscvs.files.changed";
$ADDED_FILE = "/tmp/#egcscvs.files.added";
$REMOVED_FILE = "/tmp/#egcscvs.files.removed";
$LOG_FILE = "/tmp/#egcscvs.files.log";

$FILE_PREFIX = "#egcscvs.files";

#
# Subroutines
#

sub cleanup_tmpfiles {
local($wd, @files);

$wd = `pwd`;
chdir("/tmp") || die("Can't chdir('/tmp')\n");
opendir(DIR, ".");
push(@files, grep(/^$FILE_PREFIX\..*\.$id$/, readdir(DIR)));
closedir(DIR);
foreach (@files) {
unlink $_;
}
unlink $LAST_FILE . "." . $id;

chdir($wd);
}

sub write_logfile {
local($filename, @lines) = @_;

open(FILE, ">$filename") || die("Cannot open log file $filename.\n");
print FILE join("\n", @lines), "\n";
close(FILE);
}

sub format_names {
local($dir, @files) = @_;
local(@lines);

if ($dir =~ /^\.\//) {
$dir = $';
}
if ($dir =~ /\/$/) {
$dir = $`;
}
if ($dir eq "") {
$dir = ".";
}

$format = "\t%-" . sprintf("%d", length($dir) > 15 ? length($dir) : 15) . "s%s ";

$lines[0] = sprintf($format, $dir, ":");

if ($debug) {
print STDERR "format_names(): dir = ", $dir, "; files = ", join(":", @files), ".\n";
}
foreach $file (@files) {
if (length($lines[$#lines]) + length($file) > 65) {
$lines[++$#lines] = sprintf($format, " ", " ");
}
$lines[$#lines] .= $file . " ";
}

@lines;
}

sub format_lists {
local(@lines) = @_;
local(@text, @files, $lastdir);

if ($debug) {
print STDERR "format_lists(): ", join(":", @lines), "\n";
}
@text = ();
@files = ();
$lastdir = shift @lines; # first thing is always a directory
if ($lastdir !~ /.*\/$/) {
die("Damn, $lastdir doesn't look like a directory!\n");
}
foreach $line (@lines) {
if ($line =~ /.*\/$/) {
push(@text, &format_names($lastdir, @files));
$lastdir = $line;
@files = ();
} else {
push(@files, $line);
}
}
push(@text, &format_names($lastdir, @files));

@text;
}

sub accum_subject {
local(@lines) = @_;
local(@files, $lastdir);

$lastdir = shift @lines; # first thing is always a directory
@files = ($lastdir);
if ($lastdir !~ /.*\/$/) {
die("Damn, $lastdir doesn't look like a directory!\n");
}
foreach $line (@lines) {
if ($line =~ /.*\/$/) {
$lastdir = $line;
push(@files, $line);
} else {
push(@files, $lastdir . $line);
}
}

@files;
}

sub compile_subject {
local(@files) = @_;
local($text, @a, @b, @c, $dir, $topdir);

# find the highest common directory
$dir = '-';
do {
$topdir = $dir;
foreach $file (@files) {
if ($file =~ /.*\/$/) {
if ($dir eq '-') {
$dir = $file;
} else {
if (index($dir,$file) == 0) {
$dir = $file;
} elsif (index($file,$dir) != 0) {
@a = split /\//,$file;
@b = split /\//,$dir;
@c = ();
CMP: while ($#a > 0 && $#b > 0) {
if ($a[0] eq $b[0]) {
push(@c, $a[0]);
shift @a;
shift @b;
} else {
last CMP;
}
}
$dir = join('/',@c) . '/';
}
}
}
}
} until $dir eq $topdir;

# strip out directories and the common prefix topdir.
chop $topdir;
@c = ($modulename . '/' . $topdir);
foreach $file (@files) {
if (!($file =~ /.*\/$/)) {
push(@c, substr($file, length($topdir)+1));
}
}

# put it together and limit the length.
$text = join(' ',@c);
if (length($text) > 50) {
$text = substr($text, 0, 46) . ' ...';
}

$text;
}

sub append_names_to_file {
local($filename, $dir, @files) = @_;

if (@files) {
open(FILE, ">>$filename") || die("Cannot open file $filename.\n");
print FILE $dir, "\n";
print FILE join("\n", @files), "\n";
close(FILE);
}
}

sub read_line {
local($line);
local($filename) = @_;

open(FILE, "<$filename") || die("Cannot open file $filename.\n");
$line = <FILE>;
close(FILE);
chop($line);
$line;
}

sub read_logfile {
local(@text);
local($filename, $leader) = @_;

open(FILE, "<$filename");
while (<FILE>) {
chop;
push(@text, $leader.$_);
}
close(FILE);
@text;
}

sub build_header {
local($header);
local($sec,$min,$hour,$mday,$mon,$year) = localtime(time);
$header = sprintf("CVSROOT:\t%s\nModule name:\t%s\n",
$cvsroot,
$modulename);
if (defined($branch)) {
$header .= sprintf("Branch: \t%s\n",
$branch);
}
$header .= sprintf("Changes by:\t%s(at)%s\t%02d/%02d/%02d %02d:%02d:%02d",
$login, $hostdomain,
$year%100, $mon+1, $mday,
$hour, $min, $sec);
}

sub mail_notification {
local($name, $subject, @text) = @_;
open(MAIL, "| $MAILER -s \"$subject\" $name");
print MAIL join("\n", @text), "\n";
close(MAIL);
}

sub write_commitlog {
local($logfile, @text) = @_;

open(FILE, ">>$logfile");
print FILE join("\n", @text), "\n\n";
close(FILE);
}

#
# Main Body
#

# Initialize basic variables
#
$debug = 0;
$id = getpgrp(); # note, you *must* use a shell which does setpgrp()
$state = $STATE_NONE;
$login = $ENV{'USER'} || (getpwuid($<))[0] || "nobody";
chop($hostname = `hostname`);
if ($hostname !~ /\./) {
chop($domainname = `domainname`);
$hostdomain = $hostname . "." . $domainname;
} else {
$hostdomain = $hostname;
}
$cvsroot = $ENV{'CVSROOT'};
$do_status = 1;
$modulename = "";

# parse command line arguments (file list is seen as one arg)
#
while (@ARGV) {
$arg = shift @ARGV;

if ($arg eq '-d') {
$debug = 1;
print STDERR "Debug turned on...\n";
} elsif ($arg eq '-m') {
$mailto = "$mailto " . shift @ARGV;
} elsif ($arg eq '-M') {
$modulename = shift @ARGV;
} elsif ($arg eq '-s') {
$do_status = 0;
} elsif ($arg eq '-f') {
($commitlog) && die("Too many '-f' args\n");
$commitlog = shift @ARGV;
} else {
($donefiles) && die("Too many arguments! Check usage.\n");
$donefiles = 1;
@files = split(/ /, $arg);
}
}
($mailto) || die("No -m mail recipient specified\n");

# for now, the first "file" is the repository directory being committed,
# relative to the $CVSROOT location
#
@path = split('/', $files[0]);

# XXX there are some ugly assumptions in here about module names and
# XXX directories relative to the $CVSROOT location -- really should
# XXX read $CVSROOT/CVSROOT/modules, but that's not so easy to do, since
# XXX we have to parse it backwards.
#
if ($modulename eq "") {
$modulename = $path[0]; # I.e. the module name == top-level dir
}
if ($commitlog ne "") {
$commitlog = $cvsroot . "/" . $modulename . "/" . $commitlog unless ($commitlog =~ /^\//);
}
if ($#path == 0) {
$dir = ".";
} else {
$dir = join('/', @path[1..$#path]);
}
$dir = $dir . "/";

if ($debug) {
print STDERR "module - ", $modulename, "\n";
print STDERR "dir - ", $dir, "\n";
print STDERR "path - ", join(":", @path), "\n";
print STDERR "files - ", join(":", @files), "\n";
print STDERR "id - ", $id, "\n";
}

# Check for a new directory first. This appears with files set as follows:
#
# files[0] - "path/name/newdir"
# files[1] - "-"
# files[2] - "New"
# files[3] - "directory"
#
if ($files[2] =~ /New/ && $files[3] =~ /directory/) {
local(@text);

@text = ();
push(@text, &build_header());
push(@text, "");
push(@text, $files[0]);
push(@text, "");

while (<STDIN>) {
chop; # Drop the newline
push(@text, $_);
}

&mail_notification($mailto, $files[0], @text);

if ($commitlog) {
&write_commitlog($commitlog, @text);
}

exit 0;
}

# Iterate over the body of the message collecting information.
#
while (<STDIN>) {
chop; # Drop the newline

if (/^Modified Files/) { $state = $STATE_CHANGED; next; }
if (/^Added Files/) { $state = $STATE_ADDED; next; }
if (/^Removed Files/) { $state = $STATE_REMOVED; next; }
if (/^Log Message/) { $state = $STATE_LOG; next; }
if (/^Revision\/Branch/) { /^[^:]+:\s*(.*)/; $branch = $+; next; }

s/^[ \t\n]+//; # delete leading whitespace
s/[ \t\n]+$//; # delete trailing whitespace

if ($state == $STATE_CHANGED) { push(@changed_files, split); }
if ($state == $STATE_ADDED) { push(@added_files, split); }
if ($state == $STATE_REMOVED) { push(@removed_files, split); }
if ($state == $STATE_LOG) { push(@log_lines, $_); }
}

# Strip leading and trailing blank lines from the log message. Also
# compress multiple blank lines in the body of the message down to a
# single blank line.
#
while ($#log_lines > -1) {
last if ($log_lines[0] ne "");
shift(@log_lines);
}
while ($#log_lines > -1) {
last if ($log_lines[$#log_lines] ne "");
pop(@log_lines);
}
for ($i = $#log_lines; $i > 0; $i--) {
if (($log_lines[$i - 1] eq "") && ($log_lines[$i] eq "")) {
splice(@log_lines, $i, 1);
}
}

# Check for an import command. This appears with files set as follows:
#
# files[0] - "path/name"
# files[1] - "-"
# files[2] - "Imported"
# files[3] - "sources"
#
if ($files[2] =~ /Imported/ && $files[3] =~ /sources/) {
local(@text);

@text = ();
push(@text, &build_header());
push(@text, "");

push(@text, "Log message:");
while ($#log_lines > -1) {
push (@text, " " . $log_lines[0]);
shift(@log_lines);
}

&mail_notification($mailto, "Import $file[0]", @text);

if ($commitlog) {
&write_commitlog($commitlog, @text);
}

exit 0;
}

if ($debug) {
print STDERR "Searching for log file index...";
}
# Find an index to a log file that matches this log message
#
for ($i = 0; ; $i++) {
local(@text);

last if (! -e "$LOG_FILE.$i.$id"); # the next available one
@text = &read_logfile("$LOG_FILE.$i.$id", "");
last if ($#text == -1); # nothing in this file, use it
last if (join(" ", @log_lines) eq join(" ", @text)); # it's the same log message as another
}
if ($debug) {
print STDERR " found log file at $i.$id, now writing tmp files.\n";
}

# Spit out the information gathered in this pass.
#
&append_names_to_file("$CHANGED_FILE.$i.$id", $dir, @changed_files);
&append_names_to_file("$ADDED_FILE.$i.$id", $dir, @added_files);
&append_names_to_file("$REMOVED_FILE.$i.$id", $dir, @removed_files);
&write_logfile("$LOG_FILE.$i.$id", @log_lines);

# Check whether this is the last directory. If not, quit.
#
if ($debug) {
print STDERR "Checking current dir against last dir.\n";
}
$_ = &read_line("$LAST_FILE.$id");

if ($_ ne $cvsroot . "/" . $files[0]) {
if ($debug) {
print STDERR sprintf("Current directory %s is not last directory %s.\n", $cvsroot . "/" .$files[0], $_);
}
exit 0;
}
if ($debug) {
print STDERR sprintf("Current directory %s is last directory %s -- all commits done.\n", $files[0], $_);
}

#
# End Of Commits!
#

# This is it. The commits are all finished. Lump everything together
# into a single message, fire a copy off to the mailing list, and drop
# it on the end of the Changes file.
#

#
# Produce the final compilation of the log messages
#
@text = ();
@status_txt = ();
@subject_files = ();
push(@text, &build_header());
push(@text, "");

for ($i = 0; ; $i++) {
last if (! -e "$LOG_FILE.$i.$id"); # we're done them all!
@lines = &read_logfile("$CHANGED_FILE.$i.$id", "");
if ($#lines >= 0) {
push(@text, "Modified files:");
push(@text, &format_lists(@lines));
push(@subject_files, &accum_subject(@lines));
}
@lines = &read_logfile("$ADDED_FILE.$i.$id", "");
if ($#lines >= 0) {
push(@text, "Added files:");
push(@text, &format_lists(@lines));
push(@subject_files, &accum_subject(@lines));
}
@lines = &read_logfile("$REMOVED_FILE.$i.$id", "");
if ($#lines >= 0) {
push(@text, "Removed files:");
push(@text, &format_lists(@lines));
push(@subject_files, &accum_subject(@lines));
}
if ($#text >= 0) {
push(@text, "");
}
@lines = &read_logfile("$LOG_FILE.$i.$id", "\t");
if ($#lines >= 0) {
push(@text, "Log message:");
push(@text, @lines);
push(@text, "");
}
if ($do_status) {
local(@changed_files);

@changed_files = ();
push(@changed_files, &read_logfile("$CHANGED_FILE.$i.$id", ""));
push(@changed_files, &read_logfile("$ADDED_FILE.$i.$id", ""));
push(@changed_files, &read_logfile("$REMOVED_FILE.$i.$id", ""));

if ($debug) {
print STDERR "main: pre-sort changed_files = ", join(":", @changed_files), ".\n";
}
sort(@changed_files);
if ($debug) {
print STDERR "main: post-sort changed_files = ", join(":", @changed_files), ".\n";
}

foreach $dofile (@changed_files) {
if ($dofile =~ /\/$/) {
next; # ignore the silly "dir" entries
}
if ($debug) {
print STDERR "main(): doing status on $dofile\n";
}
open(STATUS, "-|") || exec 'cvs', '-n', 'status', '-Qqv', $dofile;
while (<STATUS>) {
chop;
push(@status_txt, $_);
}
}
}
}

$subject_txt = &compile_subject(@subject_files);

# Write to the commitlog file
#
if ($commitlog) {
&write_commitlog($commitlog, @text);
}

if ($#status_txt >= 0) {
push(@text, @status_txt);
}

# Mailout the notification.
#
&mail_notification($mailto, $subject_txt, @text);

# cleanup
#
if (! $debug) {
&cleanup_tmpfiles();
}

exit 0;
==================================================