#!/usr/bin/perl =head1 NAME mr - a Multiple Repository management tool =head1 SYNOPSIS B [options] checkout B [options] update B [options] status B [options] commit [-m "message"] B [options] record [-m "message"] B [options] push B [options] diff B [options] log B [options] run command [param ...] B [options] bootstrap url [directory] B [options] register [repository] B [options] config section ["parameter=[value]" ...] B [options] action [params ...] B [options] [online|offline] B [options] remember action [params ...] =head1 DESCRIPTION B is a Multiple Repository management tool. It can checkout, update, or perform other actions on a set of repositories as if they were one combined repository. It supports any combination of subversion, git, cvs, mercurial, bzr, darcs and fossil repositories, and support for other version control systems can easily be added. B cds into and operates on all registered repositories at or below your working directory. Or, if you are in a subdirectory of a repository that contains no other registered repositories, it will stay in that directory, and work on only that repository, B is configured by .mrconfig files, which list the repositories. It starts by reading the .mrconfig file in your home directory, and this can in turn chain load .mrconfig files from repositories. It also automatically looks for a .mrconfig file in the current directory, or in one of its parent directories. These predefined commands should be fairly familiar to users of any version control system: =over 4 =item checkout (or co) Checks out any repositories that are not already checked out. =item update Updates each repository from its configured remote repository. If a repository isn't checked out yet, it will first check it out. =item status Displays a status report for each repository, showing what uncommitted changes are present in the repository. =item commit (or ci) Commits changes to each repository. (By default, changes are pushed to the remote repository too, when using distributed systems like git. If you don't like this default, you can change it in your .mrconfig, or use record instead.) The optional -m parameter allows specifying a commit message. =item record Records changes to the local repository, but does not push them to the remote repository. Only supported for distributed version control systems. The optional -m parameter allows specifying a commit message. =item push Pushes committed local changes to the remote repository. A no-op for centralized version control systems. =item diff Show a diff of uncommitted changes. =item log Show the commit log. =item run command [param ...] Runs the specified command in each repository. =back These commands are also available: =over 4 =item bootstrap url [directory] Causes mr to download the url, and use it as a .mrconfig file to checkout the repositories listed in it, into the specified directory. The directory will be created if it does not exist. If no directory is specified, the current directory will be used. If the .mrconfig file includes a repository named ".", that is checked out into the top of the specified directory. =item list (or ls) List the repositories that mr will act on. =item register Register an existing repository in a mrconfig file. By default, the repository in the current directory is registered, or you can specify a directory to register. The mrconfig file that is modified is chosen by either the -c option, or by looking for the closest known one at or in a parent of the current directory. =item config Adds, modifies, removes, or prints a value from a mrconfig file. The next parameter is the name of the section the value is in. To add or modify values, use one or more instances of "parameter=value". Use "parameter=" to remove a parameter. Use just "parameter" to get the value of a parameter. For example, to add (or edit) a repository in src/foo: mr config src/foo checkout="svn co svn://example.com/foo/trunk foo" To show the command that mr uses to update the repository in src/foo: mr config src/foo update To see the built-in library of shell functions contained in mr: mr config DEFAULT lib The mrconfig file that is used is chosen by either the -c option, or by looking for the closest known one at or in a parent of the current directory. =item offline Advises mr that it is in offline mode. Any commands that fail in offline mode will be remembered, and retried when mr is told it's online. =item online Advices mr that it is in online mode again. Commands that failed while in offline mode will be re-run. =item remember Remember a command, to be run later when mr re-enters online mode. This implicitly puts mr into offline mode. The command can be any regular mr command. This is useful when you know that a command will fail due to being offline, and so don't want to run it right now at all, but just remember to run it when you go back online. =item help Displays this help. =back Actions can be abbreviated to any unambiguous substring, so "mr st" is equivalent to "mr status", and "mr up" is equivalent to "mr update" Additional parameters can be passed to most commands, and are passed on unchanged to the underlying version control system. This is mostly useful if the repositories mr will act on all use the same version control system. =head1 OPTIONS =over 4 =item -d directory =item --directory directory Specifies the topmost directory that B should work in. The default is the current working directory. =item -c mrconfig =item --config mrconfig Use the specified mrconfig file. The default is to use both B<~/.mrconfig> as well as look for a .mrconfig file in the current directory, or in one of its parent directories. =item -v =item --verbose Be verbose. =item -q =item --quiet Be quiet. This supresses mr's usual output, as well as any output from commands that are run (including stderr output). If a command fails, the output will be shown. =item -k =item --insecure Accept untrusted SSL certificates when bootstrapping. =item -s =item --stats Expand the statistics line displayed at the end to include information about exactly which repositories failed and were skipped, if any. =item -i =item --interactive Interactive mode. If a repository fails to be processed, a subshell will be started which you can use to resolve or investigate the problem. Exit the subshell to continue the mr run. =item -n [number] =item --no-recurse [number] If no number if specified, just operate on the repository for the current directory, do not recurse into deeper repositories. If a number is specified, will recurse into repositories at most that many subdirectories deep. For example, with -n 2 it would recurse into ./src/foo, but not ./src/packages/bar. =item -j [number] =item --jobs [number] Run the specified number of jobs in parallel, or an unlimited number of jobs with no number specified. This can greatly speed up operations such as updates. It is not recommended for interactive operations. Note that running more than 10 jobs at a time is likely to run afoul of ssh connection limits. Running between 3 and 5 jobs at a time will yield a good speedup in updates without loading the machine too much. =item -t =item --trust-all Trust all mrconfig files even if they are not listed in ~/.mrtrust. Use with caution. =item -p =item --path This obsolete flag is ignored. =back =head1 MRCONFIG FILES Here is an example .mrconfig file: [src] checkout = svn checkout svn://svn.example.com/src/trunk src chain = true [src/linux-2.6] checkout = git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git && cd linux-2.6 && git checkout -b mybranch origin/master The .mrconfig file uses a variant of the INI file format. Lines starting with "#" are comments. Values can be continued to the following line by indenting the line with whitespace. The "DEFAULT" section allows setting default values for the sections that come after it. The "ALIAS" section allows adding aliases for actions. Each parameter is an alias, and its value is the action to use. All other sections add repositories. The section header specifies the directory where the repository is located. This is relative to the directory that contains the mrconfig file, but you can also choose to use absolute paths. (Note that you can use environment variables in section names; they will be passed through the shell for expansion. For example, "[$HOSTNAME]", or "[${HOSTNAME}foo]") Within a section, each parameter defines a shell command to run to handle a given action. mr contains default handlers for "update", "status", "commit", and other standard actions. Normally you only need to specify what to do for "checkout". Here you specify the command to run in order to create a checkout of the repository. The command will be run in the parent directory, and must create the repository's directory. So use "git clone", "svn checkout", "bzr branch" or "bzr checkout" (for a bound branch), etc. Note that these shell commands are run in a "set -e" shell environment, where any additional parameters you pass are available in "$@". All commands other than "checkout" are run inside the repository, though not necessarily at the top of it. The "MR_REPO" environment variable is set to the path to the top of the repository. (For the "register" action, "MR_REPO" is instead set to the basename of the directory that should be created when checking the repository out.) The "MR_CONFIG" environment variable is set to the .mrconfig file that defines the repo being acted on, or, if the repo is not yet in a config file, the .mrconfig file that should be modified to register the repo. The "MR_ACTION" environment variable is set to the command being run (update, checkout, etc). A few parameters have special meanings: =over 4 =item skip If the "skip" parameter is set and its command returns true, then B will skip acting on that repository. The command is passed the action name in $1. Here are two examples. The first skips the repo unless mr is run by joey. The second uses the hours_since function (included in mr's built-in library) to skip updating the repo unless it's been at least 12 hours since the last update. [mystuff] checkout = ... skip = test `whoami` != joey [linux] checkout = ... skip = [ "$1" = update ] && ! hours_since "$1" 12 Another way to use skip is for a lazy checkout. This makes mr skip operating on a repo unless it already exists. To enable the repo, you have to explicitly check it out (using "mr -d foo checkout"). [foo] checkout = ... skip = lazy =item order The "order" parameter can be used to override the default ordering of repositories. The default order value is 10. Use smaller values to make repositories be processed earlier, and larger values to make repositories be processed later. Note that if a repository is located in a subdirectory of another repository, ordering it to be processed earlier is not recommended. =item chain If the "chain" parameter is set and its command returns true, then B will try to load a .mrconfig file from the root of the repository. =item include If the "include" parameter is set, its command is ran, and should output additional mrconfig file content. The content is included as if it were part of the including file. Unlike all other parameters, this parameter does not need to be placed within a section. B ships several libraries that can be included to add support for additional version control type things (unison, git-svn, vcsh, git-fake-bare, git-subtree). To include them all, you could use: include = cat /usr/share/mr/* See the individual files for details. =item deleted If the "deleted" parameter is set and its command returns true, then B will treat the repository as deleted. It won't ever actually delete the repository, but it will warn if it sees the repository's directory. This is useful when one mrconfig file is shared amoung multiple machines, to keep track of and remember to delete old repositories. =item lib The "lib" parameter can specify some shell code that will be run before each command, this can be a useful way to define shell functions for other commands to use. =item fixups If the "fixups" parameter is set, its command is run whenever a repository is checked out, or updated. This provides an easy way to do things like permissions fixups, or other tweaks to the repository content, whenever the repository is changed. =item pre_ and post_ If a "pre_action" parameter is set, its command is run before mr performs the specified action. Similarly, "post_action" parameters are run after mr successfully performs the specified action. For example, "pre_commit" is run before committing; "post_update" is run after updating. =back When looking for a command to run for a given action, mr first looks for a parameter with the same name as the action. If that is not found, it looks for a parameter named "VCS_action" (substituting in the name of the version control system and the action). The name of the version control system is itself determined by running each defined "VCS_test" action, until one succeeds. Internally, mr has settings for "git_update", "svn_update", etc. To change the action that is performed for a given version control system, you can override these VCS specific actions. To add a new version control system, you can just add VCS specific actions for it. =head1 UNTRUSTED MRCONFIG FILES Since mrconfig files can contain arbitrary shell commands, they can do anything. This flexibility is good, but it also allows a malicious mrconfig file to delete your whole home directory. Such a file might be contained inside a repository that your main ~/.mrconfig checks out. To avoid worries about evil commands in a mrconfig file, mr defaults to reading all mrconfig files other than the main ~/.mrconfig in untrusted mode. In untrusted mode, mrconfig files are limited to running only known safe commands (like "git clone") in a carefully checked manner. To configure mr to trust other mrconfig files, list them in ~/.mrtrust. One mrconfig file should be listed per line. Either the full pathname should be listed, or the pathname can start with "~/" to specify a file relative to your home directory. =head1 OFFLINE LOG FILE The ~/.mrlog file contains commands that mr has remembered to run later, due to being offline. You can delete or edit this file to remove commands, or even to add other commands for 'mr online' to run. If the file is present, mr assumes it is in offline mode. =head1 EXTENSIONS mr can be extended to support things such as unison and git-svn. Some files providing such extensions are available in /usr/share/mr/. See the documentation in the files for details about using them. =head1 EXIT STATUS mr returns nonzero if a command failed in any of the repositories. =head1 AUTHOR Copyright 2007-2011 Joey Hess Licensed under the GNU GPL version 2 or higher. http://kitenet.net/~joey/code/mr/ =cut use warnings; use strict; use Getopt::Long; use Cwd qw(getcwd abs_path); # things that can happen when mr runs a command use constant { OK => 0, FAILED => 1, SKIPPED => 2, ABORT => 3, }; # configurables my $config_overridden=0; my $verbose=0; my $quiet=0; my $stats=0; my $insecure=0; my $interactive=0; my $max_depth; my $no_chdir=0; my $jobs=1; my $trust_all=0; my $directory=getcwd(); $ENV{MR_CONFIG}=find_mrconfig(); # globals :-( my %config; my %configfiles; my %knownactions; my %alias; my (@ok, @failed, @skipped); main(); my %vcs; sub vcs_test { my ($action, $dir, $topdir, $subdir) = @_; if (exists $vcs{$dir}) { return $vcs{$dir}; } my $test="set -e\n"; foreach my $vcs_test ( sort { length $a <=> length $b || $a cmp $b } grep { /_test$/ } keys %{$config{$topdir}{$subdir}}) { my ($vcs)=$vcs_test=~/(.*)_test/; $test="my_$vcs_test() {\n$config{$topdir}{$subdir}{$vcs_test}\n}\n".$test; $test.="if my_$vcs_test; then echo $vcs; fi\n"; } $test=$config{$topdir}{$subdir}{lib}."\n".$test if exists $config{$topdir}{$subdir}{lib}; print "mr $action: running vcs test >>$test<<\n" if $verbose; my $vcs=`$test`; chomp $vcs; if ($vcs=~/\n/s) { $vcs=~s/\n/, /g; print STDERR "mr $action: found multiple possible repository types ($vcs) for ".fulldir($topdir, $subdir)."\n"; return undef; } if (! length $vcs) { return $vcs{$dir}=undef; } else { return $vcs{$dir}=$vcs; } } sub findcommand { my ($action, $dir, $topdir, $subdir, $is_checkout) = @_; if (exists $config{$topdir}{$subdir}{$action}) { return $config{$topdir}{$subdir}{$action}; } if ($is_checkout) { return undef; } my $vcs=vcs_test(@_); if (defined $vcs && exists $config{$topdir}{$subdir}{$vcs."_".$action}) { return $config{$topdir}{$subdir}{$vcs."_".$action}; } else { return undef; } } sub fulldir { my ($topdir, $subdir) = @_; return $subdir =~ /^\// ? $subdir : $topdir.$subdir; } sub action { my ($action, $dir, $topdir, $subdir, $force_checkout) = @_; my $fulldir=fulldir($topdir, $subdir); $ENV{MR_CONFIG}=$configfiles{$topdir}; my $lib=exists $config{$topdir}{$subdir}{lib} ? $config{$topdir}{$subdir}{lib}."\n" : ""; my $is_checkout=($action eq 'checkout'); my $is_update=($action =~ /update/); $ENV{MR_REPO}=$dir; $ENV{MR_ACTION}=$action; foreach my $testname ("skip", "deleted") { my $testcommand=findcommand($testname, $dir, $topdir, $subdir, $is_checkout); if (defined $testcommand) { my $test="set -e;".$lib. "my_action(){ $testcommand\n }; my_action '$action'"; print "mr $action: running $testname test >>$test<<\n" if $verbose; my $ret=system($test); if ($ret != 0) { if (($? & 127) == 2) { print STDERR "mr $action: interrupted\n"; return ABORT; } elsif ($? & 127) { print STDERR "mr $action: $testname test received signal ".($? & 127)."\n"; return ABORT; } } if ($ret >> 8 == 0) { if ($testname eq "deleted") { if (-d $dir) { print STDERR "mr error: $dir should be deleted yet still exists\n"; return FAILED; } } print "mr $action: skip $dir skipped\n" if $verbose; return SKIPPED; } } } if ($is_checkout) { if (! $force_checkout) { if (-d $dir) { print "mr $action: $dir already exists, skipping checkout\n" if $verbose; return SKIPPED; } $dir=~s/^(.*)\/[^\/]+\/?$/$1/; } } elsif ($is_update) { if (! -d $dir) { return action("checkout", $dir, $topdir, $subdir); } } my $command=findcommand($action, $dir, $topdir, $subdir, $is_checkout); if ($is_checkout && ! -d $dir) { print "mr $action: creating parent directory $dir\n" if $verbose; system("mkdir", "-p", $dir); } if (! $no_chdir && ! chdir($dir)) { print STDERR "mr $action: failed to chdir to $dir: $!\n"; return FAILED; } elsif (! defined $command) { my $vcs=vcs_test(@_); if (! defined $vcs) { print STDERR "mr $action: unknown repository type and no defined $action command for $fulldir\n"; return FAILED; } else { print STDERR "mr $action: no defined action for $vcs repository $fulldir, skipping\n"; return SKIPPED; } } else { my $actionmsg; if (! $no_chdir) { $actionmsg="mr $action: $fulldir"; } else { my $s=$directory; $s=~s/^\Q$fulldir\E\/?//; $actionmsg="mr $action: $fulldir (in subdir $s)"; } print "$actionmsg\n" unless $quiet; my $hookret=hook("pre_$action", $topdir, $subdir); return $hookret if $hookret != OK; $command="set -e; ".$lib. "my_action(){ $command\n }; my_action ". join(" ", map { s/\\/\\\\/g; s/"/\"/g; '"'.$_.'"' } @ARGV); print "mr $action: running >>$command<<\n" if $verbose; my $ret; if ($quiet) { my $output = qx/$command 2>&1/; $ret = $?; if ($ret != 0) { print "$actionmsg\n"; print STDERR $output; } } else { $ret=system($command); } if ($ret != 0) { if (($? & 127) == 2) { print STDERR "mr $action: interrupted\n"; return ABORT; } elsif ($? & 127) { print STDERR "mr $action: received signal ".($? & 127)."\n"; return ABORT; } print STDERR "mr $action: failed ($ret)\n" if $verbose; if ($ret >> 8 != 0) { print STDERR "mr $action: command failed\n"; if (-e "$ENV{HOME}/.mrlog" && $action ne 'remember') { # recreate original command line to # remember, and avoid recursing my @orig=@ARGV; @ARGV=('-n', $action, @orig); action("remember", $dir, $topdir, $subdir); @ARGV=@orig; } } elsif ($ret != 0) { print STDERR "mr $action: command died ($ret)\n"; } return FAILED; } else { if ($is_checkout && ! -d $dir) { print STDERR "mr $action: $dir missing after checkout\n";; return FAILED; } my $ret=hook("post_$action", $topdir, $subdir); return $ret if $ret != OK; if (($is_checkout || $is_update)) { my $ret=hook("fixups", $topdir, $subdir); return $ret if $ret != OK; } return OK; } } } sub hook { my ($hook, $topdir, $subdir) = @_; my $command=$config{$topdir}{$subdir}{$hook}; return OK unless defined $command; my $lib=exists $config{$topdir}{$subdir}{lib} ? $config{$topdir}{$subdir}{lib}."\n" : ""; my $shell="set -e;".$lib. "my_hook(){ $command\n }; my_hook"; print "mr $hook: running >>$shell<<\n" if $verbose; my $ret; if ($quiet) { my $output = qx/$shell 2>&1/; $ret = $?; if ($ret != 0) { print STDERR $output; } } else { $ret=system($shell); } if ($ret != 0) { if (($? & 127) == 2) { print STDERR "mr $hook: interrupted\n"; return ABORT; } elsif ($? & 127) { print STDERR "mr $hook: received signal ".($? & 127)."\n"; return ABORT; } else { return FAILED; } } return OK; } # run actions on multiple repos, in parallel sub mrs { my $action=shift; my @repos=@_; $| = 1; my @active; my @fhs; my @out; my $running=0; while (@fhs or @repos) { while ((!$jobs || $running < $jobs) && @repos) { $running++; my $repo = shift @repos; pipe(my $outfh, CHILD_STDOUT); pipe(my $errfh, CHILD_STDERR); my $pid; unless ($pid = fork) { die "mr $action: cannot fork: $!" unless defined $pid; open(STDOUT, ">&CHILD_STDOUT") || die "mr $action cannot reopen stdout: $!"; open(STDERR, ">&CHILD_STDERR") || die "mr $action cannot reopen stderr: $!"; close CHILD_STDOUT; close CHILD_STDERR; close $outfh; close $errfh; exit action($action, @$repo); } close CHILD_STDOUT; close CHILD_STDERR; push @active, [$pid, $repo]; push @fhs, [$outfh, $errfh]; push @out, ['', '']; } my ($rin, $rout) = ('',''); my $nfound; foreach my $fh (@fhs) { next unless defined $fh; vec($rin, fileno($fh->[0]), 1) = 1 if defined $fh->[0]; vec($rin, fileno($fh->[1]), 1) = 1 if defined $fh->[1]; } $nfound = select($rout=$rin, undef, undef, 1); foreach my $channel (0, 1) { foreach my $i (0..$#fhs) { next unless defined $fhs[$i]; my $fh = $fhs[$i][$channel]; next unless defined $fh; if (vec($rout, fileno($fh), 1) == 1) { my $r = ''; if (sysread($fh, $r, 1024) == 0) { close($fh); $fhs[$i][$channel] = undef; if (! defined $fhs[$i][0] && ! defined $fhs[$i][1]) { waitpid($active[$i][0], 0); print STDOUT $out[$i][0]; print STDERR $out[$i][1]; record($active[$i][1], $? >> 8); splice(@fhs, $i, 1); splice(@active, $i, 1); splice(@out, $i, 1); $running--; } } $out[$i][$channel] .= $r; } } } } } sub record { my $dir=shift()->[0]; my $ret=shift; if ($ret == OK) { push @ok, $dir; print "\n" unless $quiet; } elsif ($ret == FAILED) { if ($interactive) { chdir($dir) unless $no_chdir; print STDERR "mr: Starting interactive shell. Exit shell to continue.\n"; system((getpwuid($<))[8], "-i"); } push @failed, $dir; print "\n" unless $quiet; } elsif ($ret == SKIPPED) { push @skipped, $dir; } elsif ($ret == ABORT) { exit 1; } else { die "unknown exit status $ret"; } } sub showstats { my $action=shift; if (! @ok && ! @failed && ! @skipped) { die "mr $action: no repositories found to work on\n"; } print "mr $action: finished (".join("; ", showstat($#ok+1, "ok", "ok"), showstat($#failed+1, "failed", "failed"), showstat($#skipped+1, "skipped", "skipped"), ).")\n" unless $quiet; if ($stats) { if (@skipped) { print "mr $action: (skipped: ".join(" ", @skipped).")\n" unless $quiet; } if (@failed) { print STDERR "mr $action: (failed: ".join(" ", @failed).")\n"; } } } sub showstat { my $count=shift; my $singular=shift; my $plural=shift; if ($count) { return "$count ".($count > 1 ? $plural : $singular); } return; } # an ordered list of repos sub repolist { my @list; foreach my $topdir (sort keys %config) { foreach my $subdir (sort keys %{$config{$topdir}}) { push @list, { topdir => $topdir, subdir => $subdir, order => $config{$topdir}{$subdir}{order}, }; } } return sort { $a->{order} <=> $b->{order} || $a->{topdir} cmp $b->{topdir} || $a->{subdir} cmp $b->{subdir} } @list; } sub repodir { my $repo=shift; my $topdir=$repo->{topdir}; my $subdir=$repo->{subdir}; my $ret=($subdir =~/^\//) ? $subdir : $topdir.$subdir; $ret=~s/\/\.$//; return $ret; } # figure out which repos to act on sub selectrepos { my @repos; foreach my $repo (repolist()) { my $topdir=$repo->{topdir}; my $subdir=$repo->{subdir}; next if $subdir eq 'DEFAULT'; my $dir=repodir($repo); my $d=$directory; $dir.="/" unless $dir=~/\/$/; $d.="/" unless $d=~/\/$/; next if $dir ne $d && $dir !~ /^\Q$d\E/; if (defined $max_depth) { my @a=split('/', $dir); my @b=split('/', $d); do { } while (@a && @b && shift(@a) eq shift(@b)); next if @a > $max_depth || @b > $max_depth; } push @repos, [$dir, $topdir, $subdir]; } if (! @repos) { # fallback to find a leaf repo foreach my $repo (reverse repolist()) { my $topdir=$repo->{topdir}; my $subdir=$repo->{subdir}; next if $subdir eq 'DEFAULT'; my $dir=repodir($repo); my $d=$directory; $dir.="/" unless $dir=~/\/$/; $d.="/" unless $d=~/\/$/; if ($d=~/^\Q$dir\E/) { push @repos, [$dir, $topdir, $subdir]; last; } } $no_chdir=1; } return @repos; } sub expandenv { my $val=shift; if ($val=~/\$/) { $val=`echo "$val"`; chomp $val; } return $val; } my %trusted; sub is_trusted_config { my $config=shift; # must be abs_pathed already # We always trust ~/.mrconfig. return 1 if $config eq abs_path("$ENV{HOME}/.mrconfig"); return 1 if $trust_all; my $trustfile=$ENV{HOME}."/.mrtrust"; if (! %trusted) { $trusted{"$ENV{HOME}/.mrconfig"}=1; if (open (TRUST, "<", $trustfile)) { while () { chomp; s/^~\//$ENV{HOME}\//; $trusted{abs_path($_)}=1; } close TRUST; } } return $trusted{$config}; } sub is_trusted_repo { my $repo=shift; # Tightly limit what is allowed in a repo name. # No ../, no absolute paths, and no unusual filenames # that might try to escape to the shell. return $repo =~ /^[-_.+\/A-Za-z0-9]+$/ && $repo !~ /\.\./ && $repo !~ /^\//; } sub is_trusted_checkout { my $command=shift; # To determine if the command is safe, compare it with the # *_trusted_checkout config settings. Those settings are # templates for allowed commands, so make sure that each word # of the command matches the corresponding word of the template. my @words; foreach my $word (split(' ', $command)) { # strip quoting if ($word=~/^'(.*)'$/) { $word=$1; } elsif ($word=~/^"(.*)"$/) { $word=$1; } push @words, $word; } foreach my $key (grep { /_trusted_checkout$/ } keys %{$config{''}{DEFAULT}}) { my @twords=split(' ', $config{''}{DEFAULT}{$key}); next if @words > @twords; my $match=1; my $url; for (my $c=0; $c < @twords && $match; $c++) { if ($twords[$c] eq '$url') { # Match all the typical characters found in # urls, plus @ which svn can use. Note # that the "url" might also be a local # directory. $match=( defined $words[$c] && $words[$c] =~ /^[-_.+:@\/A-Za-z0-9]+$/ ); $url=$words[$c]; } elsif ($twords[$c] eq '$repo') { # If a repo is not specified, assume it # will be the last path component of the # url, or something derived from it, and # check that. if (! defined $words[$c] && defined $url) { ($words[$c])=$url=~/\/([^\/]+)\/?$/; } $match=( defined $words[$c] && is_trusted_repo($words[$c]) ); } elsif (defined $words[$c] && $words[$c]=~/^($twords[$c])$/) { $match=1; } else { $match=0; } } return 1 if $match; } return 0; } sub trusterror { my ($err, $file, $line, $url)=@_; if (defined $url) { die "$err in untrusted $url line $line\n". "(To trust this url, --trust-all can be used; but please use caution;\n". "this can allow arbitrary code execution!)\n"; } else { die "$err in untrusted $file line $line\n". "(To trust this file, list it in ~/.mrtrust.)\n"; } } my %loaded; sub loadconfig { my $f=shift; my $dir=shift; my $bootstrap_url=shift; my @toload; my $in; my $trusted; if (ref $f eq 'GLOB') { $dir=""; $in=$f; $trusted=1; } else { my $absf=abs_path($f); if ($loaded{$absf}) { return; } $loaded{$absf}=1; $trusted=is_trusted_config($absf); if (! defined $dir) { ($dir)=$f=~/^(.*\/)[^\/]+$/; if (! defined $dir) { $dir="."; } } $dir=abs_path($dir)."/"; if (! exists $configfiles{$dir}) { $configfiles{$dir}=$f; } # copy in defaults from first parent my $parent=$dir; while ($parent=~s/^(.*\/)[^\/]+\/?$/$1/) { if ($parent eq '/') { $parent=""; } if (exists $config{$parent} && exists $config{$parent}{DEFAULT}) { $config{$dir}{DEFAULT}={ %{$config{$parent}{DEFAULT}} }; last; } } if (! -e $f) { return; } print "mr: loading config $f\n" if $verbose; open($in, "<", $f) || die "mr: open $f: $!\n"; } my @lines=<$in>; close $in unless ref $f eq 'GLOB'; my $section; my $line=0; while (@lines) { $_=shift @lines; $line++; chomp; if (! $trusted && /[[:cntrl:]]/) { trusterror("mr: illegal control character", $f, $line, $bootstrap_url); } next if /^\s*\#/ || /^\s*$/; if (/^\[([^\]]*)\]\s*$/) { $section=$1; if (! $trusted) { if (! is_trusted_repo($section) || $section eq 'ALIAS' || $section eq 'DEFAULT') { trusterror("mr: illegal section \"[$section]\"", $f, $line, $bootstrap_url) } } $section=expandenv($section) if $trusted; if ($section ne 'ALIAS' && ! exists $config{$dir}{$section} && exists $config{$dir}{DEFAULT}) { # copy in defaults $config{$dir}{$section}={ %{$config{$dir}{DEFAULT}} }; } } elsif (/^(\w+)\s*=\s*(.*)/) { my $parameter=$1; my $value=$2; # continued value while (@lines && $lines[0]=~/^\s(.+)/) { shift(@lines); $line++; $value.="\n$1"; chomp $value; } if (! $trusted) { # Untrusted files can only contain a few # settings in specific known-safe formats. if ($parameter eq 'checkout') { if (! is_trusted_checkout($value)) { trusterror("mr: illegal checkout command \"$value\"", $f, $line, $bootstrap_url); } } elsif ($parameter eq 'order') { # not interpreted as a command, so # safe. } elsif ($value eq 'true' || $value eq 'false') { # skip=true , deleted=true etc are # safe. } else { trusterror("mr: illegal setting \"$parameter=$value\"", $f, $line, $bootstrap_url); } } if ($parameter eq "include") { print "mr: including output of \"$value\"\n" if $verbose; unshift @lines, `$value`; if ($?) { print STDERR "mr: include command exited nonzero ($?)\n"; } next; } if (! defined $section) { die "$f line $.: parameter ($parameter) not in section\n"; } if ($section eq 'ALIAS') { $alias{$parameter}=$value; } elsif ($parameter eq 'lib') { $config{$dir}{$section}{lib}.=$value."\n"; } else { $config{$dir}{$section}{$parameter}=$value; if ($parameter =~ /.*_(.*)/) { $knownactions{$1}=1; } else { $knownactions{$parameter}=1; } if ($parameter eq 'chain' && length $dir && $section ne "DEFAULT") { my $chaindir="$section"; if ($chaindir !~ m!/!) { $chaindir=$dir.$chaindir; } if (-e "$chaindir/.mrconfig") { my $ret=system($value); if ($ret != 0) { if (($? & 127) == 2) { print STDERR "mr: chain test interrupted\n"; exit 2; } elsif ($? & 127) { print STDERR "mr: chain test received signal ".($? & 127)."\n"; } } else { push @toload, ["$chaindir/.mrconfig", $chaindir]; } } } } } else { die "$f line $line: parse error\n"; } } foreach my $c (@toload) { loadconfig(@$c); } } sub startingconfig { %alias=%config=%configfiles=%knownactions=%loaded=(); my $datapos=tell(DATA); loadconfig(\*DATA); seek(DATA,$datapos,0); # rewind } sub modifyconfig { my $f=shift; # the section to modify or add my $targetsection=shift; # fields to change in the section # To remove a field, set its value to "". my %changefields=@_; my @lines; my @out; if (-e $f) { open(my $in, "<", $f) || die "mr: open $f: $!\n"; @lines=<$in>; close $in; } my $formatfield=sub { my $field=shift; my @value=split(/\n/, shift); return "$field = ".shift(@value)."\n". join("", map { "\t$_\n" } @value); }; my $addfields=sub { my @blanks; while ($out[$#out] =~ /^\s*$/) { unshift @blanks, pop @out; } foreach my $field (sort keys %changefields) { if (length $changefields{$field}) { push @out, "$field = $changefields{$field}\n"; delete $changefields{$field}; } } push @out, @blanks; }; my $section; while (@lines) { $_=shift(@lines); if (/^\s*\#/ || /^\s*$/) { push @out, $_; } elsif (/^\[([^\]]*)\]\s*$/) { if (defined $section && $section eq $targetsection) { $addfields->(); } $section=expandenv($1); push @out, $_; } elsif (/^(\w+)\s*=\s(.*)/) { my $parameter=$1; my $value=$2; # continued value while (@lines && $lines[0]=~/^\s(.+)/) { shift(@lines); $value.="\n$1"; chomp $value; } if ($section eq $targetsection) { if (exists $changefields{$parameter}) { if (length $changefields{$parameter}) { $value=$changefields{$parameter}; } delete $changefields{$parameter}; } } push @out, $formatfield->($parameter, $value); } } if (defined $section && $section eq $targetsection) { $addfields->(); } elsif (%changefields) { push @out, "\n[$targetsection]\n"; foreach my $field (sort keys %changefields) { if (length $changefields{$field}) { push @out, $formatfield->($field, $changefields{$field}); } } } open(my $out, ">", $f) || die "mr: write $f: $!\n"; print $out @out; close $out; } sub dispatch { my $action=shift; # actions that do not operate on all repos if ($action eq 'help') { help(@ARGV); } elsif ($action eq 'config') { config(@ARGV); } elsif ($action eq 'register') { register(@ARGV); } elsif ($action eq 'bootstrap') { bootstrap(); } elsif ($action eq 'remember' || $action eq 'offline' || $action eq 'online') { my @repos=selectrepos; action($action, @{$repos[0]}) if @repos; exit 0; } if (!$jobs || $jobs > 1) { mrs($action, selectrepos()); } else { foreach my $repo (selectrepos()) { record($repo, action($action, @$repo)); } } } sub help { exec($config{''}{DEFAULT}{help}) || die "exec: $!"; } sub config { if (@_ < 2) { die "mr config: not enough parameters\n"; } my $section=shift; if ($section=~/^\//) { # try to convert to a path relative to the config file my ($dir)=$ENV{MR_CONFIG}=~/^(.*\/)[^\/]+$/; $dir=abs_path($dir); $dir.="/" unless $dir=~/\/$/; if ($section=~/^\Q$dir\E(.*)/) { $section=$1; } } my %changefields; foreach (@_) { if (/^([^=]+)=(.*)$/) { $changefields{$1}=$2; } else { my $found=0; foreach my $topdir (sort keys %config) { if (exists $config{$topdir}{$section} && exists $config{$topdir}{$section}{$_}) { print $config{$topdir}{$section}{$_}."\n"; $found=1; last if $section eq 'DEFAULT'; } } if (! $found) { die "mr config: $section $_ not set\n"; } } } modifyconfig($ENV{MR_CONFIG}, $section, %changefields) if %changefields; exit 0; } sub register { if ($config_overridden) { # Find the directory that the specified config file is # located in. ($directory)=abs_path($ENV{MR_CONFIG})=~/^(.*\/)[^\/]+$/; } else { # Find the closest known mrconfig file to the current # directory. $directory.="/" unless $directory=~/\/$/; my $foundconfig=0; foreach my $topdir (reverse sort keys %config) { next unless length $topdir; if ($directory=~/^\Q$topdir\E/) { $ENV{MR_CONFIG}=$configfiles{$topdir}; $directory=$topdir; $foundconfig=1; last; } } if (! $foundconfig) { $directory=""; # no config file, use builtin } } if (@ARGV) { my $subdir=shift @ARGV; if (! chdir($subdir)) { print STDERR "mr register: failed to chdir to $subdir: $!\n"; } } $ENV{MR_REPO}=getcwd(); my $command=findcommand("register", $ENV{MR_REPO}, $directory, 'DEFAULT', 0); if (! defined $command) { die "mr register: unknown repository type\n"; } $ENV{MR_REPO}=~s/.*\/(.*)/$1/; $command="set -e; ".$config{$directory}{DEFAULT}{lib}."\n". "my_action(){ $command\n }; my_action ". join(" ", map { s/\\/\\\\/g; s/"/\"/g; '"'.$_.'"' } @ARGV); print "mr register: running >>$command<<\n" if $verbose; exec($command) || die "exec: $!"; } sub bootstrap { my $url=shift @ARGV; my $dir=shift @ARGV || "."; if (! defined $url || ! length $url) { die "mr: bootstrap requires url\n"; } # Download the config file to a temporary location. eval q{use File::Temp}; die $@ if $@; my $tmpconfig=File::Temp->new(); my @curlargs = ("curl", "-A", "mr", "-L", "-s", $url, "-o", $tmpconfig); push(@curlargs, "-k") if $insecure; my $curlstatus = system(@curlargs); die "mr bootstrap: invalid SSL certificate for $url (consider -k)\n" if $curlstatus >> 8 == 60; die "mr bootstrap: download of $url failed\n" if $curlstatus != 0; if (! -e $dir) { system("mkdir", "-p", $dir); } chdir($dir) || die "chdir $dir: $!"; # Special case to handle checkout of the "." repo, which # would normally be skipped. my $topdir=abs_path(".")."/"; my @repo=($topdir, $topdir, "."); loadconfig($tmpconfig, $topdir, $url); record(\@repo, action("checkout", @repo, 1)) if exists $config{$topdir}{"."}{"checkout"}; if (-e ".mrconfig") { print STDERR "mr bootstrap: .mrconfig file already exists, not overwriting with $url\n"; } else { eval q{use File::Copy}; die $@ if $@; move($tmpconfig, ".mrconfig") || die "rename: $!"; } # Reload the config file (in case we got a different version) # and checkout everything else. startingconfig(); loadconfig(".mrconfig"); dispatch("checkout"); @skipped=grep { abs_path($_) ne abs_path($topdir) } @skipped; showstats("bootstrap"); exitstats(); } # alias expansion and command stemming sub expandaction { my $action=shift; if (exists $alias{$action}) { $action=$alias{$action}; } if (! exists $knownactions{$action}) { my @matches = grep { /^\Q$action\E/ } keys %knownactions, keys %alias; if (@matches == 1) { $action=$matches[0]; } elsif (@matches == 0) { die "mr: unknown action \"$action\" (known actions: ". join(", ", sort keys %knownactions).")\n"; } else { die "mr: ambiguous action \"$action\" (matches: ". join(", ", @matches).")\n"; } } return $action; } sub find_mrconfig { my $dir=getcwd(); while (length $dir) { if (-e "$dir/.mrconfig") { return "$dir/.mrconfig"; } $dir=~s/\/[^\/]*$//; } return "$ENV{HOME}/.mrconfig"; } sub getopts { my @saved=@ARGV; Getopt::Long::Configure("bundling", "no_permute"); my $result=GetOptions( "d|directory=s" => sub { $directory=abs_path($_[1]) }, "c|config=s" => sub { $ENV{MR_CONFIG}=$_[1]; $config_overridden=1 }, "p|path" => sub { }, # now default, ignore "v|verbose" => \$verbose, "q|quiet" => \$quiet, "s|stats" => \$stats, "k|insecure" => \$insecure, "i|interactive" => \$interactive, "n|no-recurse:i" => \$max_depth, "j|jobs:i" => \$jobs, "t|trust-all" => \$trust_all, ); if (! $result || @ARGV < 1) { die("Usage: mr [options] action [params ...]\n". "(Use mr help for man page.)\n"); } $ENV{MR_SWITCHES}=""; foreach my $option (@saved) { last if $option eq $ARGV[0]; $ENV{MR_SWITCHES}.="$option "; } } sub init { $SIG{INT}=sub { print STDERR "mr: interrupted\n"; exit 2; }; # This can happen if it's run in a directory that was removed # or other strangeness. if (! defined $directory) { die("mr: failed to determine working directory\n"); } # Make sure MR_CONFIG is an absolute path, but don't use abs_path since # the config file might be a symlink to elsewhere, and the directory it's # in is significant. if ($ENV{MR_CONFIG} !~ /^\//) { $ENV{MR_CONFIG}=getcwd()."/".$ENV{MR_CONFIG}; } # Try to set MR_PATH to the path to the program. eval { use FindBin qw($Bin $Script); $ENV{MR_PATH}=$Bin."/".$Script; }; } sub exitstats { if (@failed) { exit 1; } else { exit 0; } } sub main { getopts(); init(); startingconfig(); loadconfig("$ENV{HOME}/.mrconfig"); loadconfig($ENV{MR_CONFIG}); #use Data::Dumper; print Dumper(\%config); my $action=expandaction(shift @ARGV); dispatch($action); showstats($action); exitstats(); } # Finally, some useful actions that mr knows about by default. # These can be overridden in ~/.mrconfig. __DATA__ [ALIAS] co = checkout ci = commit ls = list [DEFAULT] order = 10 lib = error() { echo "mr: $@" >&2 exit 1 } warning() { echo "mr (warning): $@" >&2 } info() { echo "mr: $@" >&2 } hours_since() { if [ -z "$1" ] || [ -z "$2" ]; then error "mr: usage: hours_since action num" fi for dir in .git .svn .bzr CVS .hg _darcs _FOSSIL_; do if [ -e "$MR_REPO/$dir" ]; then flagfile="$MR_REPO/$dir/.mr_last$1" break fi done if [ -z "$flagfile" ]; then error "cannot determine flag filename" fi delta=`perl -wle 'print -f shift() ? int((-M _) * 24) : 9999' "$flagfile"` if [ "$delta" -lt "$2" ]; then return 1 else touch "$flagfile" return 0 fi } is_bzr_checkout() { LANG=C bzr info | egrep -q '^Checkout' } lazy() { if [ "$MR_ACTION" = checkout ] || [ -d "$MR_REPO" ]; then return 1 else return 0 fi } svn_test = test -d "$MR_REPO"/.svn git_test = test -d "$MR_REPO"/.git bzr_test = test -d "$MR_REPO"/.bzr cvs_test = test -d "$MR_REPO"/CVS hg_test = test -d "$MR_REPO"/.hg darcs_test = test -d "$MR_REPO"/_darcs fossil_test = test -f "$MR_REPO"/_FOSSIL_ git_bare_test = test -d "$MR_REPO"/refs/heads && test -d "$MR_REPO"/refs/tags && test -d "$MR_REPO"/objects && test -f "$MR_REPO"/config && test "`GIT_CONFIG="$MR_REPO"/config git config --get core.bare`" = true svn_update = svn update "$@" git_update = git pull "$@" bzr_update = if is_bzr_checkout; then bzr update "$@" else bzr merge --pull "$@" fi cvs_update = cvs update "$@" hg_update = hg pull "$@" && hg update "$@" darcs_update = darcs pull -a "$@" fossil_update = fossil pull "$@" svn_status = svn status "$@" git_status = git status -s "$@" || true bzr_status = bzr status --short "$@" cvs_status = cvs status "$@" hg_status = hg status "$@" darcs_status = darcs whatsnew -ls "$@" || true fossil_status = fossil changes "$@" svn_commit = svn commit "$@" git_commit = git commit -a "$@" && git push --all bzr_commit = if is_bzr_checkout; then bzr commit "$@" else bzr commit "$@" && bzr push fi cvs_commit = cvs commit "$@" hg_commit = hg commit -m "$@" && hg push darcs_commit = darcs record -a -m "$@" && darcs push -a fossil_commit = fossil commit "$@" git_record = git commit -a "$@" bzr_record = if is_bzr_checkout; then bzr commit --local "$@" else bzr commit "$@" fi hg_record = hg commit -m "$@" darcs_record = darcs record -a -m "$@" fossil_record = fossil commit "$@" svn_push = : git_push = git push "$@" bzr_push = bzr push "$@" cvs_push = : hg_push = hg push "$@" darcs_push = darcs push -a "$@" fossil_push = fossil push "$@" svn_diff = svn diff "$@" git_diff = git diff "$@" bzr_diff = bzr diff "$@" cvs_diff = cvs diff "$@" hg_diff = hg diff "$@" darcs_diff = darcs diff -u "$@" fossil_diff = fossil diff "$@" svn_log = svn log "$@" git_log = git log "$@" bzr_log = bzr log "$@" cvs_log = cvs log "$@" hg_log = hg log "$@" darcs_log = darcs changes "$@" git_bare_log = git log "$@" fossil_log = fossil timeline "$@" run = "$@" svn_register = url=`LC_ALL=C svn info . | grep -i '^URL:' | cut -d ' ' -f 2` if [ -z "$url" ]; then error "cannot determine svn url" fi echo "Registering svn url: $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="svn co '$url' '$MR_REPO'" git_register = url="`LC_ALL=C git config --get remote.origin.url`" || true if [ -z "$url" ]; then error "cannot determine git url" fi echo "Registering git url: $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="git clone '$url' '$MR_REPO'" bzr_register = url="`LC_ALL=C bzr info . | egrep -i 'checkout of branch|parent branch' | awk '{print $NF}'`" if [ -z "$url" ]; then error "cannot determine bzr url" fi echo "Registering bzr url: $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="bzr branch '$url' '$MR_REPO'" cvs_register = repo=`cat CVS/Repository` root=`cat CVS/Root` if [ -z "$root" ]; then error "cannot determine cvs root" fi echo "Registering cvs repository $repo at root $root" mr -c "$MR_CONFIG" config "`pwd`" checkout="cvs -d '$root' co -d '$MR_REPO' '$repo'" hg_register = url=`hg showconfig paths.default` echo "Registering mercurial repo url: $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="hg clone '$url' '$MR_REPO'" darcs_register = url=`cat _darcs/prefs/defaultrepo` echo "Registering darcs repository $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="darcs get '$url' '$MR_REPO'" git_bare_register = url="`LC_ALL=C GIT_CONFIG=config git config --get remote.origin.url`" || true if [ -z "$url" ]; then error "cannot determine git url" fi echo "Registering git url: $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="git clone --bare '$url' '$MR_REPO'" fossil_register = url=`fossil remote-url` repo=`fossil info | grep repository | sed -e 's/repository:*.//g' -e 's/ //g'` echo "Registering fossil repository $url in $MR_CONFIG" mr -c "$MR_CONFIG" config "`pwd`" checkout="mkdir -p '$MR_REPO' && cd '$MR_REPO' && fossil open '$repo'" svn_trusted_checkout = svn co $url $repo svn_alt_trusted_checkout = svn checkout $url $repo git_trusted_checkout = git clone $url $repo bzr_trusted_checkout = bzr checkout|clone|branch|get $url $repo # cvs: too hard hg_trusted_checkout = hg clone $url $repo darcs_trusted_checkout = darcs get $url $repo git_bare_trusted_checkout = git clone --bare $url $repo # fossil: messy to do help = case `uname -s` in SunOS) SHOWMANFILE="man -f" ;; Darwin) SHOWMANFILE="man" ;; *) SHOWMANFILE="man -l" ;; esac if [ ! -e "$MR_PATH" ]; then error "cannot find program path" fi tmp=$(mktemp -t mr.XXXXXXXXXX) || error "mktemp failed" trap "rm -f $tmp" exit pod2man -c mr "$MR_PATH" > "$tmp" || error "pod2man failed" $SHOWMANFILE "$tmp" || error "man failed" list = true config = bootstrap = online = if [ -s ~/.mrlog ]; then info "running offline commands" mv -f ~/.mrlog ~/.mrlog.old if ! sh -e ~/.mrlog.old; then error "offline command failed; left in ~/.mrlog.old" fi rm -f ~/.mrlog.old else info "no offline commands to run" fi offline = umask 077 touch ~/.mrlog info "offline mode enabled" remember = info "remembering command: 'mr $@'" command="mr -d '$(pwd)' $MR_SWITCHES" for w in "$@"; do command="$command '$w'" done if [ ! -e ~/.mrlog ] || ! grep -q -F "$command" ~/.mrlog; then echo "$command" >> ~/.mrlog fi ed = echo "A horse is a horse, of course, of course.." T = echo "I pity the fool." right = echo "Not found." # vim:sw=8:sts=0:ts=8:noet