package VCP::Source::cvs ; =head1 NAME VCP::Source::cvs - A CVS repository source =head1 SYNOPSIS vcp cvs:module/... -d ">=2000-11-18 5:26:30" # All file revs newer than a date/time vcp cvs:module/... -r foo # all files in module and below labelled foo vcp cvs:module/... -r foo: # All revs of files labelled foo and newer, # including files not tagged with foo. vcp cvs:module/... -r 1.1:1.10 # revs 1.1..1.10 vcp cvs:module/... -r 1.1: # revs 1.1 and up on main trunk ## NOTE: Unlike cvs, vcp requires spaces after option letters. =head1 DESCRIPTION Source driver enabling L|vcp> to extract versions form a cvs repository. The source specification for CVS looks like: cvs:cvsroot:module/filespec [] or optionally, if the C environment variable is set: cvs:module/filespec [] The cvsroot is passed to C with cvs' C<-d> option. The filespec and EoptionsE determine what revisions to extract. C may contain trailing wildcards, like C to extract an entire directory tree. If the cvsroot looks like a local filesystem (if it doesn't start with ":" and if it points to an existing directory or file), this module will read the RCS files directly from the hard drive unless --use-cvs is passed. This is more accurate (due to poor design of the cvs log command) and much, much faster. =head1 OPTIONS =over =item --cd Used to set the CVS working directory. VCP::Source::cvs will cd to this directory before calling cvs, and won't initialize a CVS workspace of it's own (normally, VCP::Source::cvs does a "cvs checkout" in a temporary directory). This is an advanced option that allows you to use a CVS workspace you establish instead of letting vcp create one in a temporary directory somewhere. This is useful if you want to read from a CVS branch or if you want to delete some files or subdirectories in the workspace. If this option is a relative directory, then it is treated as relative to the current directory. =item -k -k sadf Pass the CVS -k options through to the underlying CVS command. =item -kb Pass the -kb option to cvs, forces a binary checkout. This is useful when you want a text file to be checked out with Unix linends, or if you know that some files in the repository are not flagged as binary files and should be. =item -r -r v_0_001:v_0_002 -r v_0_002: Passed to C as a C<-r> revision specification. This corresponds to the C<-r> option for the rlog command, not either of the C<-r> options for the cvs command. Yes, it's confusing, but "cvs log" calls "rlog" and passes the options through. =item --use-cvs Do not try to read local repositories directly; use the cvs command line interface. This is much slower than reading the files directly but is useful to see if there is a bug in the RCS file parser or possibly when dealing with corrupt RCS files that cvs will read. If you find that this option makes something work, then there is a discrepancy between the code that reads the RCS files directly (in the absence of this option) and cvs itself. Please let me know (barrie@slaysys.com). Thanks. =item -d -d "2000-11-18 5:26:30<=" Passed to 'cvs log' as a C<-d> date specification. WARNING: if this string doesn't contain a '>' or '<', you're probably doing something wrong, since you're not specifying a range. vcp may warn about this in the future. see "log" command in cvs(1) man page for syntax of the date specification. =back =head1 CVS Conversion issues =head2 Files that aren't tagged CVS has one peculiarity that this driver works around. If a file does not contain the tag(s) used to select the source files, C outputs the entire life history of that file. We don't want to capture the entire history of such files, so L ignores any revisions before and after the oldest and newest tagged file. =head2 Branches with multiple tags / "cloned" branches CVS allows branches to be tagged with multiple tags using a command like cvs admin second_branch_tag:branch_tag When VCP::Source::cvs notices this, it creates multiple branches with identical revisions. This is done by choosing the first branch tag for the branches to be the primary branch and applying all actual changes to it then "clone"ing each revision from that branch to all others. For instance, if file foo is branched once in to a branch tagged with "bar" and later a "goof" tag is aliased to the "bar" tag, then trunk branch bar branch goof ======= ================ ================== foo#1.1 | \ | \ | \ ... foo#1.1.1.1 | \ | \ | \ | foo#1.1.1.1 | foo#1.1.1.2 | \ | \ | \ | foo#1.1.1.2 | ... This is EXPERIMENTAL and it's likely to give VCP::Dest::cvs fits. It is tested with CVS->p4 transfers. If you only want the primary branch, you may use a Map: section in the .vcp file to discard non-primary branches: Map: ... <> Currently, there is no way to ignore the primary branch other than getting rid of that branch tag in the RCS files or hacking VCP::Source::cvs's source code to ignore it. =head1 FEATURES (EXPERIMENTAL) It's possible somehow (I've never done it) to set the state on edited revisions to "dead", which may result in a series of revisions all marked "dead". VCP::Source::cvs detects consecutive "dead" revisions and inserts concocted "edit" revisions between the "delete" revisions to force the destination to alter the file between deleted revisions. =head1 LIMITATIONS Stores all revisions for a file in RAM before sending so it can link all the revisions properly. Also stores all branch parents and the first revision on every branch for all files scanned so it can insert placeholders for branches with no revs. Except for these branch point revisions, all other revs for each file are sent before the next file is scanned. TODO: just send placeholders for all branches that match the filespec and revspec? Does not yet set the same time in all branch creation revisions. This may be necessary in order to help the changeset aggregator. It will probably take buffering all branch revs before sending them on. Also, it is not possible in the general case: not all files on a branch are actually branched from parents that are checked in before the first file on a branch is created. It also makes no sense to do this for untagged branches as there is no detectable semantic association between untagged branches. CVS does not try to protect itself from people checking in things that look like snippets of CVS log file: they come out exactly like they went in, confusing the log file parser. So, if a repository contains messages in the log file that look like the output from some other "cvs log" command, things will likely go awry when using remote repositories (local repositories are read directly and do not suffer this problem). The direct RCS file parser does not have this problem. CVS stores the -k keyword expansion setting per file, not per revision, so vcp will mark all revisions of a file with the current setting of the -k flag for a file. At least one cvs repository out there has multiple revisions of a single file with the same rev number. The second and later revisions with the same rev number are ignored with a warning like "Can't add same revision twice:...". The xfree86 repository has several files xc/programs/Xserver/hw/xfree86//vga256/drivers/s3/s3Bt485.h: 1.2 dead lines +1, -1 1.1 Exp lines 1.2.2.2 Exp lines +1, -1 1.2.2.1 Exp lines +1, -1 In this case, VCP::Source::cvs doesn't know how to retrieve rev 1.2 to create the branch 1.2.2.x, so it uses 1.1. If you know how to force it to get rev 1.2, please let me know (in the future, the RCS parser will allow this, but currently we always use cvs checkout to retrieve versions). I'd like to know how to use the cvs command to modify a revision and then force it to the dead state without upping the revision number, as that appears to have happened here. I suspect something other than the cvs command at play here, like the rcs command or RCS file editing by hand or by script =for test_script t/80rcs_parser.t t/91cvs2revml.t =cut $VERSION = 1.2 ; # Removed docs for -f, since I now think it's overcomplicating things... #Without a -f This will normally only replicate files which are tagged. This #means that files that have been added since, or which are missing the tag for #some reason, are ignored. # #Use the L option to force files that don't contain the tag to be #=item -f # #This option causes vcp to attempt to export files that don't contain a #particular tag but which occur in the date range spanned by the revisions #specified with -r. The typical use is to get all files from a certain #tag to now. # #It does this by exporting all revisions of files between the oldest and #newest files that the -r specified. Without C<-f>, these would #be ignored. # #It is an error to specify C<-f> without C<-r>. # #exported. =begin developerdocs CVS branching needs some help to allow VCP::Dest::* drivers to branch files appropriately because CVS branch creation (cvs rtag -b) does not actually create a branched revision, it marks the parent revision for a branch. However, we need to include information about all the files that should be branched when a branch is created. CVS does not reliably record the data of branch creation (though we might be able to find it in CVS history, I have not found that to be a reliable approach; different CVS versions seem to capture different things, at least by default). There's a dilemma here: we need to create branched revisions for created branches, but we don't know when to do it. So the first time we see a change on a branch, we create dummy revisions for all files that are also on that branch. These dummy revisions are VCP::Rev instances with a ".0" rev_id and with no , , , or elements. See the docs for VCP::Rev::is_placeholder_rev(). Detection of branches that have been initiated this transfer has to occur after unwanted revisions are thrown out (due to rev_id < last_rev_in_filebranch or due to culling of unwanted revs) so we don't think all branches have been created this transfer. =end developerdocs =cut use strict ; use Carp ; use VCP::Debug qw( :debug :profile ) ; use VCP::Logger qw( lg pr pr_doing pr_done pr_done_failed BUG ); use VCP::Rev qw( iso8601format ); use VCP::Source ; use VCP::Utils qw( empty is_win32 shell_quote start_dir_rel2abs xchdir ); use VCP::Utils::cvs ; use constant debug_parser => 0; use base qw( VCP::Source VCP::Utils::cvs ) ; use fields ( 'CVS_INFO', ## Results of the 'cvs --version' command and CVSROOT 'CVS_REV_SPEC', ## The revision spec to pass to `cvs log` 'CVS_DATE_SPEC', ## The date spec to pass to `cvs log` 'CVS_WORK_DIR', ## working directory set via --cd option 'CVS_USE_CVS', ## holds the value of the --use-cvs option 'CVS_K_OPTION', ## Which of the CVS/RCS "-k" options to use, if any 'CVS_PARSE_RCS_FILES', ## Read CVS files directly instead of through the ## cvs command. Used if the CVSROOT looks local. 'CVS_LOG_FILE_DATA', ## Data about all revs of a file from the log file 'CVS_LOG_STATE', ## Parser state machine state 'CVS_LOG_REV', ## The revision being parsed (a hash) 'CVS_NEEDS_BASE_REV', ## What base revisions are needed. Base revs are ## needed for incremental (ie non-bootstrap) updates, ## which is decided on a per-file basis by looking ## at VCP::Source::is_bootstrap_mode( $file ) and ## the file's rev number (ie does it end in .1). 'CVS_SAW_EQUALS', ## Set when we see the ==== lines in log file [1] ## The following are for parsing RCS files directly 'CVS_RCS_FILE_PATH', ## The file currently being scanned when reading ## RCS files directly. 'CVS_RCS_FILE_BUFFER', ## The file currently being scanned when reading 'CVS_RCS_FILE_LINES', ## How many lines have already been purged from ## CVS_RCS_FILE_BUFFER. 'CVS_RCS_FILE_EOF', ## Set if we've read the end of file. 'CVS_MIN_REV', ## The first desired rev_id or tag, if defined 'CVS_MAX_REV', ## The first desired rev_id or tag, if defined 'CVS_READ_SIZE', ## Used in test suite to probe the RCS parser by ## forcing a really tiny buffer size on it. 'CVS_PRINTING_HASHES', 'CVS_FIRST_BRANCHED_REVS', ## HASH of files that have been branched using ## given branch tags, so that we can put ## placholders in for any that *weren't* ## branched for a given tag. 'CVS_BRANCH_PARENTS', ## All revisions that are the parents of ## branches. Keyed on tag, each value is ## an array of [ $filename, $rev_id ]s for ## parent revs for that branch. 'CVS_ALIASED_BRANCH_TAGS', ## CVS allows filebranches to have more than ## one tag. In this case, we clone all entries ## on the oldest tag in to entries on the ## newer tags, rev by rev, almost as though ## each was branched on to the new dest. 'CVS_CLONE_COUNT', ## How many clones had to be made 'CVS_APPLIED_TAGS_COUNT', ## Statistics gathering 'CVS_HIGHEST_REV_TIME', ## When a branch with no revisions is created, ## we need to fake up a time. So we use the ## time of the youngest revision in this ## transfer plus one second. 'CVS_FOUNDING_REVS', ## Set for all revs that are "founding" ## revs. ) ; sub new { my $class = shift ; $class = ref $class || $class ; my VCP::Source::cvs $self = $class->SUPER::new( @_ ) ; $self->{CVS_READ_SIZE} = 100_000; ## Parse the options my ( $spec, $options ) = @_ ; $self->parse_cvs_repo_spec( $spec ) unless empty $spec; $self->parse_options( $options ); return $self ; } sub parse_options { my $self = shift; $self->SUPER::parse_options( @_ ); } sub options_spec { my VCP::Source::cvs $self = shift; return ( $self->SUPER::options_spec, "cd=s" => \$self->{CVS_WORK_DIR}, "d=s" => sub { shift; $self->date_spec( @_ ) }, "k=s" => sub { shift; $self->{CVS_K_OPTION} .= shift if @_; my $v = $self->{CVS_K_OPTION}; $v =~ s/b//g if defined $v; defined $v && length $v ? $v : undef; }, "kb" => sub { shift; $self->{CVS_K_OPTION} .= "b" if @_; return defined $self->{CVS_K_OPTION} && $self->{CVS_K_OPTION} =~ "b"; }, "r=s" => sub { shift; $self->rev_spec( @_ ) }, "use-cvs" => \$self->{CVS_USE_CVS}, ); } sub init { my VCP::Source::cvs $self= shift ; $self->SUPER::init; ## Set default repo_id. $self->repo_id( "cvs:" . $self->repo_server ) if empty $self->repo_id && ! empty $self->repo_server ; $self->{CVS_PARSE_RCS_FILES} = ! $self->{CVS_USE_CVS} && do { # If the CVSROOT does not start with a colon, it must be # a direct read. But check to see if it exists anyway, # because we'd prefer CVS give the error messages around here. my $root = $self->cvsroot; substr( $root, 0, 1 ) ne ":" && -d $root; }; my $files = $self->repo_filespec ; $self->deduce_rev_root( $files ) unless defined $self->rev_root; ## Don't normalize the filespec. $self->repo_filespec( $files ) ; ## Make sure the cvs command is available $self->command_stderr_filter( qr{^ (?:cvs\s (?: (?:server|add|remove):\suse\s'cvs\scommit'\sto.* |tag.*(?:waiting for.*lock|obtained_lock).* ) )\n }x ) ; if ( $self->{CVS_PARSE_RCS_FILES} ) { my $root = $self->cvsroot; $self->{CVS_INFO} = <rev_spec; if ( defined $rev_spec ) { for ( $rev_spec ) { if ( /^([^:]*):([^:]*)\z/ ) { @$self{qw( CVS_MIN_REV CVS_MAX_REV )} = ( $1, $2 ); } else { die "can't parse revision specification '$rev_spec'"; } } } } else { $self->cvs( ['--version' ], undef, \$self->{CVS_INFO} ) ; ## This does a checkout, so we'll blow up quickly if there's a problem. my $work_dir = $self->{CVS_WORK_DIR}; unless ( defined $work_dir ) { $self->create_cvs_workspace ; } else { $self->work_root( start_dir_rel2abs $work_dir ) ; $self->command_chdir( $self->work_path ) ; } } } sub ui_set_cvs_work_dir { my VCP::Source::cvs $self = shift ; my ($dir) = @_; $self->{CVS_WORK_DIR} = $dir; die "Warning: '$dir' not found!\n" unless -e $dir; die "Error: '$dir' exists, but is not a directory.\n" unless -d $dir; } sub is_incremental { my VCP::Source::cvs $self= shift ; my ( $r ) = @_; my $file = $r->source_name; my $bootstrap_mode = ( exists $self->{CVS_FOUNDING_REVS}->{$r->id} || $self->is_bootstrap_mode( $file ) ); return $bootstrap_mode ? 0 : "incremental" ; } sub rev_spec { my VCP::Source::cvs $self = shift ; $self->{CVS_REV_SPEC} = shift if @_ ; return $self->{CVS_REV_SPEC} ; } sub rev_spec_cvs_option { my VCP::Source::cvs $self = shift ; return defined $self->rev_spec? "-r" . $self->rev_spec : (), } sub date_spec { my VCP::Source::cvs $self = shift ; $self->{CVS_DATE_SPEC} = shift if @_ ; return $self->{CVS_DATE_SPEC} ; } sub date_spec_cvs_option { my VCP::Source::cvs $self = shift ; return defined $self->date_spec ? "-d" . $self->date_spec : (), } sub denormalize_name { my VCP::Source::cvs $self = shift ; ( my $n = '/' . $self->SUPER::denormalize_name( @_ ) ) =~ s{/+}{/}g; return $n; } sub handle_header { my VCP::Source::cvs $self = shift; my ( $header ) = @_; $header->{rep_type} = 'cvs'; $header->{rep_desc} = $self->{CVS_INFO}; $header->{rev_root} = $self->rev_root; $self->{CVS_FOUNDING_REVS} = {}; $self->dest->handle_header( $header ); } sub get_source_file { my VCP::Source::cvs $self = shift ; my VCP::Rev $r ; ( $r ) = @_ ; BUG "can't check out ", $r->as_string, "\n" unless $r->is_base_rev || $r->action eq "add" || $r->action eq "edit"; my $wp = $self->work_path( "revs", $r->source_name, $r->source_rev_id ) ; $self->mkpdir( $wp ) ; my $cvs_name = $self->denormalize_name( $r->source_name ); $self->cvs( [ "checkout", "-r" . $r->source_rev_id, "-p", !empty( $self->{CVS_K_OPTION} ) ? "-k" . $self->{CVS_K_OPTION} : (), $cvs_name ], undef, $wp, ) ; return $wp; } sub _concoct_base_rev { my VCP::Source::cvs $self = shift; my ( $r ) = @_; Carp::confess "undefined \$r" unless defined $r; ## TODO: test for when the base revision is on a different branch? if ( empty( $r->previous_id ) && $self->is_incremental( $r ) ) { my $rev_id = $r->rev_id ; $rev_id =~ s{(\d+)$}{$1-1}e ; Carp::confess "illegal rev concocted: $rev_id (from ", $r->as_string, ")" if $rev_id =~ /\.[^1-9]\d*\z/; my $br = VCP::Rev->new( id => $self->denormalize_name( $r->name )."#$rev_id", name => $r->name, source_name => $r->source_name, source_filebranch_id => $r->source_filebranch_id, source_repo_id => $r->source_repo_id, branch_id => $r->branch_id, source_branch_id => $r->source_branch_id, rev_id => $rev_id, source_rev_id => $rev_id, type => $r->type, ); $r->previous_id( $br->id ); if ( debugging ) { debug "concocted base rev: ", $br->as_string; debug " for: ", $r->as_string; } $self->set_last_rev_in_filebranch_previous_id( $br ); $self->queue_rev( $br ); } } sub _concoct_edit_rev_for_dead_rev { my VCP::Source::cvs $self = shift; my ( $nr ) = @_; ## When two revisions in a row are in state "dead", we need to emit ## an edit between the two delete actions. Carp::confess "undefined \$r" unless defined $nr; my $er = VCP::Rev->new( id => $nr->id . "(dead)", action => "edit", name => $nr->name, source_name => $nr->source_name, source_filebranch_id => $nr->source_filebranch_id, source_repo_id => $nr->source_repo_id, branch_id => $nr->source_branch_id, source_branch_id => $nr->source_branch_id, rev_id => $nr->source_rev_id, source_rev_id => $nr->source_rev_id, type => $nr->type, comment => "edit of dead file", ); if ( debugging ) { debug "concocted edit rev: ", $er->as_string; debug " for: ", $nr->as_string; } $nr->previous_id( $er->id ); $self->queue_rev( $er ); } sub _concoct_branch_rev { my VCP::Source::cvs $self = shift; my ( $r, $branch_tag, $magic_branch_number, $time ) = @_; ## $r is either the parent of the branch or the first rev on the branch $magic_branch_number =~ /\A((?:\d+\.\d+\.)+)(?:0\.)?(\d+)\z/ or BUG "can't parse magic branch number '$magic_branch_number'"; my ( $branch_number, $rev_id ) = ( "$1$2", "$1$2.0" ); ## .0 'cuz it's a placeholder ( my $filebranch_id = $r->source_filebranch_id ) =~ s/<.*>\z/<$branch_number>/; ## use source_...() variants in case the rev has passed through a ## modifying filter. my $source_name = $r->source_name; my $pr = VCP::Rev->new( action => "branch", id => $self->denormalize_name( $source_name ) . "#$rev_id", name => $source_name, source_name => $source_name, source_filebranch_id => $filebranch_id, source_repo_id => $r->source_repo_id, time => $time, branch_id => $branch_tag, source_branch_id => $branch_tag, rev_id => $rev_id, source_rev_id => $rev_id, user_id => $r->user_id, comment => "[vcp] create branch '$branch_tag'", ); if ( debugging ) { debug "concocted: ", $pr->as_string; debug " from: ", $r->as_string; } $self->queue_rev( $pr ); return $pr; } sub _concoct_cloned_rev { my VCP::Source::cvs $self = shift; my ( $r, $branch_tag ) = @_; ( my $filebranch_id = $r->source_filebranch_id ) =~ s/<.*>\z/<$branch_tag>/; ## The comment field is for end users and to help the changeset ## aggregator group things properly when "comment equal" is ## specified. Really, the changeset aggregator needs to detect ## both the cloned-from parent and the preceding version, but ## we set previous_id to the cloned-from parent, not the ## preceding version. my $pr = VCP::Rev->new( action => "clone", id => $r->id . "<$branch_tag clone>", name => $r->source_name, source_name => $r->source_name, source_filebranch_id => $filebranch_id, source_repo_id => $r->source_repo_id, time => $r->time, branch_id => $branch_tag, source_branch_id => $branch_tag, rev_id => $r->source_rev_id, source_rev_id => $r->source_rev_id, user_id => $r->user_id, previous_id => $r->id, comment => $r->is_branch_rev ? join( "", "[vcp] create branch '", $branch_tag, "' by cloning '", $r->source_branch_id, "'" ) : $r->comment, ); if ( debugging ) { debug "cloned: ", $pr->as_string; debug " from: ", $r->as_string; } $self->queue_rev( $pr ); return $pr; } sub get_revs_from_log_file { my VCP::Source::cvs $self = shift; $self->{CVS_LOG_STATE} = '' ; $self->{CVS_LOG_FILE_DATA} = {} ; $self->{CVS_LOG_REV} = {} ; $self->{CVS_SAW_EQUALS} = 0 ; # The log command must be run in the directory above the work root, # since we pass in the name of the workroot dir as the first dir in # the filespec. my $tmpdir = $self->tmp_dir( "co" ) ; my $spec = $self->repo_filespec; $spec =~ s{/+(\.\.\.)?\z}{}; ## hack, since cvs always recurses. my @log_parms = ( "log", $self->rev_spec_cvs_option, $self->date_spec_cvs_option, length $spec ? $spec : (), ); pr "running ", shell_quote "cvs", @log_parms; $self->cvs( \@log_parms, undef, sub { $self->parse_cvs_log_output( @_ ) }, { in_dir => $tmpdir, stderr_filter => sub { my ( $err_text_ref ) = @_ ; $$err_text_ref =~ s{ ## This regexp needs to gobble newlines. ^cvs(?:\.exe)?\slog:\swarning:\sno\srevision\s.*?\sin\s[`"'](.*)[`"']\r?\n\r? }{}gxmi ; }, }, ) ; for my $r ( $self->queued_revs ) { ## Because of the order of the log file, the last rev set is always ## the first rev in the range. if ( $r->rev_id =~ /(.*)\.(\d+)\.1\z/ ) { my ( $parent_rev_id, $branch_number ) = ( $1, $2 ); my $magic_branch_number = "$parent_rev_id.0.$branch_number"; my $pr = $self->_concoct_branch_rev( $r, $r->branch_id, $magic_branch_number, $r->time ); $pr->previous_id( $r->previous_id ); $r->previous_id( $pr->id ); push @{$self->{CVS_FIRST_BRANCHED_REVS}->{$r->branch_id}}, $pr } } $self->_concoct_base_rev( $_ ) for $self->last_revs_for_all_filebranches; } sub parse_rcs_files { my VCP::Source::cvs $self = shift; require File::Find; require Cwd; my $root = $self->cvsroot; my $spec = $self->repo_filespec; # my $cwd = Cwd::cwd; xchdir $root; ## Be compatible with cvs and recurse by default. $spec =~ s{/*\z}{/...} if $spec !~ m{(^|\/)...\z} && -d $spec; $spec =~ s{^/+}{}; local $| = 1; my $hash_count = 0; my $hash_time = 0; $File::Find::prune = 0; ## Suppress used only once warning. my %seen; # Jump as far down the directory hierarchy as we can. # Figure out if this is a specific file by adding ,v # and checking for it (here and in the Attic), but that's # not worth the hassle right now. It would save us some # work when pulling a file out of the top of a big dir tree, # though. ( my $start = $spec ) =~ s{(^|/+)[^/]*(\*|\?|\.\.\.).*}{}; { my $where = "$root/"; $where .= "$start/" if length $start; $where .= "..."; pr_doing "scanning '$where': "; } debug "start: ", $start if debugging; debug "spec: ", $spec if debugging; my $files_count = 0; if ( -f "$start,v" ) { ++$files_count; $self->parse_rcs_file( $start ); goto SKIP_FILE_FIND; } ( my $attic_start = $start ) =~ s{((?:[\\/]|\A))([^\\/]+)\z}{${1}Attic/$2}; if ( -f "$attic_start,v" ) { ++$files_count; $self->parse_rcs_file( $attic_start ); goto SKIP_FILE_FIND; } while ( length $start && ! -d $start ) { last unless $start =~ s{/+[^/]*\z}{}; } # $spec = substr( $spec, length $start ); ## TODO: fix this for /foo/bar.../baz $spec =~ s{^[\\/]+}{}g; my $pat = $self->compile_path_re( $spec ); debug "pattern: ", $pat if debugging; $start = "." unless length $start && -d $start; my $ok = eval { File::Find::find( { no_chdir => 1, wanted => sub { if ( /CVSROOT\z/ ) { $File::Find::prune = 1; return; } return if -d; s/^\.\///; return unless s/,v\z//; ( my $undeleted_path = $_ ) =~ s/(\/)Attic\//$1/; if ( -f _ && $undeleted_path =~ $pat ) { if ( $seen{$undeleted_path}++ ) { pr "already scanned '$undeleted_path,v',", " ignoring '$_,v'"; return; } eval { lg "parseing '", $_, "'"; $self->parse_rcs_file( $_ ); 1; } or do { pr_done_failed; die "$@ for $_\n"; }; ++$files_count; } pr_doing; }, }, $start ); 1; }; my $x = $@; die $x unless $ok; SKIP_FILE_FIND: pr_done "found $files_count files"; } ## Used to detect symbols for branch tags and vendor branch tags. sub _is_branch_or_vendor_tag($) { return $_[0] =~ /\.0\.\d+\z/ || ! ( $_[0] =~ tr/.// % 2 ); } { my $special = "\$,.:;\@"; my $idchar = "[^\\s$special\\d\\.]"; # Differs from man rcsfile(1) my $num_re = "[0-9.]+"; my $id_re = "(?:(?:$num_re)?$idchar(?:$idchar|$num_re)*)"; my %id_map = ( # RCS file => "cvs log" (& its parser) field name changes "log" => "comment", "expand" => "keyword", ); sub _xdie { my VCP::Source::cvs $self = shift; my $buffer = $self->{CVS_RCS_FILE_BUFFER}; my $pos = pos( $$buffer ) || 0; my $line = $self->{CVS_RCS_FILE_LINES} + ( substr( $$buffer, 0, $pos ) =~ tr/\n// ); my $near = substr( $$buffer, $pos, 100 ); $near .= "..." if $pos + 100 > length $$buffer; $near =~ s/\n/\\n/g; $near =~ s/\r/\\r/g; die @_, " in RCS file $self->{CVS_RCS_FILE_PATH}, near line $line: '$near'\n"; } sub _read_rcs_goodness { my VCP::Source::cvs $self = shift; my ( $fh ) = @_; $self->_xdie( "read beyond end of file" ) if $self->{CVS_RCS_FILE_EOF}; my $buffer = $self->{CVS_RCS_FILE_BUFFER}; my $pos = pos( $$buffer ) || 0; ## || 0 in case no matches yet. $self->{CVS_RCS_FILE_LINES} += substr( $$buffer, 0, $pos ) =~ tr/\n//; substr( $$buffer, 0, $pos ) = ""; my $c = 0; { my $little_buffer; $c = read $fh, $little_buffer, $self->{CVS_READ_SIZE}; ## Hmmm, sometimes $c comes bak undefined at end of file, ## with $! not TRUE. most odd. Tested with 5.6.1 and 5.8.0 $self->_xdie( "$! reading rcs file" ) if ! defined $c && $!; $$buffer .= $little_buffer if $c; }; pos( $$buffer ) = 0; ## Prevent undefs from tripping up code later $self->{CVS_RCS_FILE_EOF} ||= ! $c; 1; } sub parse_rcs_file { my VCP::Source::cvs $self = shift; profile_start ref( $self ) . " parse_rcs_file()" if profiling; my ( $file ) = @_; require File::Spec::Unix; my $path = $self->{CVS_RCS_FILE_PATH} = File::Spec::Unix->canonpath( join "", $self->cvsroot, "/", $file, ",v" ); debug "going to read $path" if debugging; open F, "<$path" or die "$!: $path\n"; binmode F; my $rev_id; $file =~ s{\A(.*?)[\\/]+Attic}{$1}; my $file_data = { rcs => $path, working => $file, revs => {}, }; $self->{CVS_RCS_FILE_EOF} = 0; $self->{CVS_RCS_FILE_LINES} = 0; $self->{CVS_RCS_FILE_BUFFER} = \(my $b = ""); local $_; *_ = $self->{CVS_RCS_FILE_BUFFER}; pos = 0; my $h; # which hash to stick the data in. As the parsing progresses, # this is pointed at the per-file metadata # hash or a per-revision hash so that the low level # key/value parsing just parses things and stuffs them # in $h and it'll be stuffing them in the right place. my $id; # the name of the element to assign the next value to START: $self->_read_rcs_goodness( \*F ); if ( /\A($id_re)\s+(?=\S)/gc ) { $h = $file_data; $id = $1; $id = $id_map{$id} if exists $id_map{$id}; # had a buggy RE once... $self->_xdie( "$id should not have been parsed as an identifier" ) if $id =~ /\A$num_re\z/o; debug "parsing field ", $id if debug_parser && debugging; goto VALUE; } else { ## ASSume first identifier < 100 chars if ( ! $self->{CVS_RCS_FILE_EOF} && length() < 100 ) { debug "reading more for START parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto START; } $self->_xdie( "RCS file should begin with an identifier" ); } PARAGRAPH_START: if ( /\G($num_re)\r?\n/gc ) { $rev_id = $1; if ( debug_parser && debugging ) { my $is_new = ! exists $file_data->{revs}->{$rev_id}; debug "parsing", $is_new ? () : " MORE", " ", $rev_id, " fields"; } ## Throw away unwanted revs ASAP to save space and so the part of ## the culling algorithm that estimates limits can find the ## oldest / newest wanted revs easily. my $keep = 1; $keep &&= VCP::Rev->cmp_id( $rev_id, $file_data->{min_rev_id} ) >= 0 if defined $file_data->{min_rev_id}; $keep &&= VCP::Rev->cmp_id( $rev_id, $file_data->{max_rev_id} ) <= 0 if defined $file_data->{max_rev_id}; if ( $keep ) { ## Reuse the existing hash if this is a second pass $h = $file_data->{revs}->{$rev_id} ||= {}; } else { ## create a throw-away hash to keep the logic simpler ## in the parser (this way it doesn't have to test ## $h for definedness each time before writing it). $h = {}; } $h->{rev_id} = $rev_id; $id = undef; goto ID; } elsif ( /\Gdesc\s+(?=\@)/gc ) { ## We're at the end of the first set of per-rev sections of the ## RCS file, switch back to the per-file metadata hash to capture ## the "desc" field. $h = $file_data; $id = "desc"; $id = $id_map{$id} if exists $id_map{$id}; debug "parsing field ", $id if debug_parser && debugging; goto VALUE; } else { ## ASSume no identifier or rev number is > approx 1000 chars long if ( ! $self->{CVS_RCS_FILE_EOF} && length() - pos() < 1000 ) { debug "reading more for PARAGRAPH_START parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto PARAGRAPH_START; } $self->_xdie( "expected an identifier or version string" ); } ID: if ( /\G($id_re)(?:\s+(?=\S)|\s*(?=;))/gc ) { # No ^, unlike PARAGRAPH_START's first RE $id = exists $id_map{$1} ? $id_map{$1} : $1; # had a buggy RE once... $self->_xdie( "$id should not have been parsed as an identifier" ) if debug_parser && $id =~ /\A$num_re\z/o; debug "parsing field ", $id if debug_parser && debugging; # goto VALUE; } else { ## ASSume no identifier > approx 1000 chars long if ( ! $self->{CVS_RCS_FILE_EOF} && length() - pos() < 1000 ) { debug "reading more for ID parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto ID; } $self->_xdie( "expected an identifier or version string" ); } VALUE: $self->_xdie( "already assigned to '$h->{$id}'" ) if debug_parser && exists $h->{$id}; VALUE_DATA: if ( substr( $_, pos, 1 ) eq ";" ) { #/\G(?=;)/gc ) { $h->{$id} = ""; # goto VALUE_END; } elsif ( /\G\@/gcs ) { # It's an RCS string (@...@) goto STRING unless $id eq "text"; TEXT: # Ignore the often veeeerry long text field for now. # # This line causes segfaults in perl5.6.1 and perl5.8.1 # under linux: # # if ( /\G(?:[^\@]|\@\@)*/gc ) { # # So, instead, we look bite off the chunks by looking for # all text up to an @ and eating it. if ( /\G[^\@]*/gc ) { goto TEXT if /\G\@\@/gc; unless ( /\G\@(?=[^\@])/gc ) { # NOTE: RCS files must end in a newline, so it's safe # to assume a non-@ after the @. debug "reading more for TEXT parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto TEXT; } } ## TODO: save this, or the location of the text, so that ## we can get it ourselves later instead of executing cvs. $h->{text} = "FILE TEXT NOT EXTRACTED FROM RCS FILE $path\n"; goto VALUE_END; STRING: ## The 1000 limits perl's internal regex recursion limit to ## well below the 32766 limit. That's ok here because we keep ## looping back for more (originally the {0,1000} was a *). if ( /\G((?:[^\@]+|(?:\@\@)+){0,1000})/gc ) { $h->{$id} .= $1 unless $id eq "text"; unless ( /\G\@(?=[^\@])/gc ) { # NOTE: RCS files must end in a newline, so it's safe # to assume a non-@ after the @. debug "reading more for STRING parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto STRING; } } $self->_xdie( "odd number of '\@'s in RCS string for field '$id'" ) if ( $h->{$id} =~ tr/\@// ) % 2; $h->{$id} =~ s/\@\@/\@/g; # goto VALUE_END; } elsif ( /\G(?!\@)/gc ) { # Not a string, so it's a semicolon delimited value NOT_STRING: if ( /\G([^;]+)/gc ) { $h->{$id} .= $1; unless ( /\G(?=;)/gc ) { debug "reading more for NOT_STRING parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto NOT_STRING; } } if ( $id eq "date" ) { ## The below seems to monkey with $_, so protect pos(). my $p = pos; $h->{time} = $self->parse_time( $h->{date} ); pos = $p; } # goto VALUE_END; } else { # We only need one char. if ( ! $self->{CVS_RCS_FILE_EOF} && length() - pos() < 1 ) { debug "reading more for VALUE_DATA parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto VALUE_DATA; } $self->_xdie( "unable to parse value for $id" ); } VALUE_END: debug "$id='", substr( $h->{$id}, 0, 100 ), length $h->{$id} > 100 ? "..." : (), "'" if debug_parser && debugging; if ( $id eq "symbols" ) { my %tags; for ( split /\s+/, $h->{symbols} ) { my ( $tag, $rev_id ) = split /:/, $_, 2; $tags{$tag} = $rev_id; push @{$h->{RTAGS}->{$rev_id}}, $tag; } delete $h->{symbols}; ## Convert the passed-in min and max revs from symbolic tags ## to dotted rev numbers. The "symbols" $id only occurs ## once in a file, so this gets executed once and is setting ## fields in the file metadata (not in a rev's metadata). if ( ! empty $self->{CVS_MIN_REV} ) { my $min_rev_id = $self->{CVS_MIN_REV}; if ( $min_rev_id =~ /[^\d.]/ ) { $min_rev_id = exists $tags{$min_rev_id} ? $tags{$min_rev_id} : undef; if ( empty $min_rev_id ) { # $min_rev_id was a tag that is not found. Emulate # cvs -r and skip this file entirely. lg "-r tag $self->{CVS_MIN_REV} not found in $h->{rcs}"; return; } } $h->{min_rev_id} = [ VCP::Rev->split_id( $min_rev_id ) ]; } if ( ! empty $self->{CVS_MAX_REV} ) { my $max_rev_id = $self->{CVS_MAX_REV}; if ( $max_rev_id =~ /[^\d.]/ ) { $max_rev_id = exists $tags{$max_rev_id} ? $tags{$max_rev_id} : undef; if ( empty $max_rev_id ) { # $max_rev_id was a tag that is not found. Emulate # cvs -r and skip this file entirely. lg "-r tag $self->{CVS_MAX_REV} not found in $h->{rcs}"; return; } } $h->{max_rev_id} = [ VCP::Rev->split_id( $max_rev_id ) ]; } } $id = undef; VALUE_END_DELIMETER: if ( /\G[ \t]*(?:\r?\n|;[ \t]*(?:\r?\n|(?=[^ \t;]))|(?=[^ \t;]))/gc ) { VALUE_END_WS: if ( /\G(?=\S)/gc ) { goto ID; } if ( /\G[ \t\r\n]*(\r?\n)+(?=\S)/gc ) { goto PARAGRAPH_START; } ## ASSume no runs of \v or \r\n of mroe than 1000 chars. if ( ! $self->{CVS_RCS_FILE_EOF} && length() - pos() < 1000 ) { debug "reading more for VALUE_END_WS parsing" if debug_parser && debugging; $self->_read_rcs_goodness( \*F ); goto VALUE_END_WS; } goto FINISHED unless length; if ( ! /\G(\r?\n)/gc ) { $self->_xdie( "expected newline" ); } } # ASSume semi + whitespace + 1 more char is less than 1000 bytes if ( length() - pos() < 1000 ) { debug "reading more for VALUE_END_DELIMETER parsing" if debug_parser && debugging; eval { $self->_read_rcs_goodness( \*F ); goto FINISHED if /\G(\r?\n)*\z/gc; goto VALUE_END_DELIMETER; }; if ( 0 == index $@, "read beyond end of file" ) { goto FINISHED if /\G(\r?\n)*\z/gc; } else { die $@; } } $self->_xdie( "expected optional semicolon and tabs or spaces" ); FINISHED: close F; ## TODO: take out this debug code my $min_rev_id = $file_data->{min_rev_id}; my $max_rev_id = $file_data->{max_rev_id}; BUG "Unexpected revision: $_" for grep( ( defined $min_rev_id && VCP::Rev->cmp_id( $_, $min_rev_id ) < 0 ) || ( defined $max_rev_id && VCP::Rev->cmp_id( $_, $max_rev_id ) > 0 ), keys %{$file_data->{revs}} ); ## Convert the hashes in to revs to let VCP::Rev fold comments, ## especially, and names in to the global hashes, saving memory. my %nexts; for my $rev_id ( keys %{$file_data->{revs}} ) { my $rev_data = $file_data->{revs}->{$rev_id}; my $r = $self->_create_rev( $file_data, $rev_data, ( $rev_id =~ tr/.// ) == 1 && empty $rev_data->{next}, ); if ( defined $r ) { $nexts{$rev_data->{next}} = $r unless empty $rev_data->{next}; $file_data->{revs}->{$rev_id} = $r; } else { delete $file_data->{revs}->{$rev_id}; } } $self->_note_duplicate_branch_tags( $file_data->{RTAGS} ); delete $file_data->{RTAGS}; # conserve memory. $self->{CVS_RCS_FILE_BUFFER} = undef; profile_end ref( $self ) . " parse_rcs_file()" if profiling; $self->queue_parsed_revs( $file_data ); } } ## this ends a scope, not a sub {} sub queue_parsed_revs { my VCP::Source::cvs $self = shift; my ( $file_data ) = @_; my @parsed_revs = values %{delete $file_data->{revs}}; return unless @parsed_revs; my %oldest_revs; my @file_revs; for my $r ( @parsed_revs ) { BUG "undefined rev" unless defined $r; if ( $r->rev_id =~ /\A(.*)\.(\d+)\.1\z/ ) { my ( $parent_rev_id, $branch_number ) = ( $1, $2 ); my $magic_branch_number = "$parent_rev_id.0.$branch_number"; my $pr = $self->_concoct_branch_rev( $r, $r->branch_id, $magic_branch_number, $r->time ); if ( $pr ) { push @{$self->{CVS_FIRST_BRANCHED_REVS}->{ defined $r->branch_id ? $r->branch_id : "" }}, $pr; $pr->previous_id( $r->previous_id ); $r->previous_id( $pr->id ); push @file_revs, $pr; } } $self->queue_rev( $r ); push @file_revs, $r; $self->{CVS_HIGHEST_REV_TIME} = $r->time if ! defined $self->{CVS_HIGHEST_REV_TIME} || ( defined $r->time && $self->{CVS_HIGHEST_REV_TIME} < $r->time ); ## RCS files are not in newest-first order for branches, ## so we can't rely on $self->last_revs_for_all_filebranches. my $or = $oldest_revs{$r->source_filebranch_id}; $oldest_revs{$r->source_filebranch_id} = $r if ! $or || VCP::Rev->cmp_id( $r->rev_id, $or->rev_id ) < 0; } $self->_concoct_base_rev( $_ ) for values %oldest_revs; # Link all the revs to their predecessors on each branch # Could do this while parsing, but would need to disentangle # the CVS next and branch fields. my %revs_by_filebranch = (); for my $r ( @file_revs ) { push @{$revs_by_filebranch{$r->source_filebranch_id}}, $r; } for my $filebranch ( values %revs_by_filebranch ) { if ( @$filebranch > 1 ) { @$filebranch = sort { VCP::Rev->cmp_id( $a->rev_id, $b->rev_id ) } @$filebranch; my $pr = $filebranch->[0]; for my $r ( @{$filebranch}[1..$#$filebranch] ) { debug "previous id for ", $r->id, ": ", $pr->id if debugging; if ( $r->action eq "delete" && $pr && $pr->action eq "delete" ) { $self->_concoct_edit_rev_for_dead_rev( $r ); $self->last_rev->previous_id( $pr->id ); } else { $r->previous_id( $pr->id ); } $pr = $r; } } } ## TODO: clean this loop out of here once we get beyond some real world ## tests. for my $r ( @file_revs ) { next if $r->is_base_rev || ( grep( $_ == $r, values %oldest_revs ) && ! $self->is_incremental( $r ) ) || ! empty $r->previous_id; die "no previous rev for ", $_->as_string; } $self->_clone_aliased_branch_tags; $self->send_revs; } sub copy_revs { my VCP::Source::cvs $self = shift ; my $empty_branch_placeholders = 0; $self->{CVS_HIGHEST_REV_TIME} = undef; $self->{CVS_ALIASED_BRANCH_TAGS} = undef; $self->{CVS_CLONE_COUNT} = undef; if ( $self->{CVS_PARSE_RCS_FILES} ) { $self->parse_rcs_files; } else { $self->get_revs_from_log_file; $self->_clone_aliased_branch_tags; } ## Figure out what branches were initiated this transfer and create ## branch placeholders for files that are branched but don't have ## revisions on them. my $last_rev_time; for my $branch_tag ( keys %{$self->{CVS_BRANCH_PARENTS}} ) { my $min_time; my %branches_with_revs = map { ## Make the parallel branches occur at the same time as the ## first file revision appeared on the branch. ## TODO: either don't store revs in CVS_FIRST_BRANCHED_REVS ## or don't send those revs downstream before this point. $min_time = $_->time if ! defined $min_time || $_->time < $min_time; ( $_->source_name => undef ); } @{$self->{CVS_FIRST_BRANCHED_REVS }->{$branch_tag}}; my @parents_to_be_branched = grep ! exists $branches_with_revs{$_->[0]->source_name}, @{$self->{CVS_BRANCH_PARENTS}->{$branch_tag}}; if ( $self->continue && $self->dest ) { ## Don't add placeholders for branches that already exist ## in destination. @parents_to_be_branched = grep { my ( $r, $magic_branch_number ) = @$_; $magic_branch_number =~ /\A(.*)\.0\.(\d+)/ or BUG "couldn't parse '$magic_branch_number'"; my $branch_number = "$1.$2"; ( my $ph_filebranch_id = $r->source_filebranch_id ) =~ s/<[^<>]*>\z/<$branch_number>/; ! defined $self->dest->last_rev_in_filebranch( $self->repo_id, $ph_filebranch_id ); } @parents_to_be_branched; } ## Use the time of the youngest rev in the transfer as the ## timestamp for branch creation. $min_time = $self->{CVS_HIGHEST_REV_TIME} + 1 unless defined $min_time; for ( @parents_to_be_branched ) { my ( $r, $magic_branch_number ) = @$_; ## NOTE: @parents_to_be_branched contains revs that ## have already been sent. Ideally, _concoct_branch_rev ## would not access any fields a filter might modify. ## That's not possible yet, there's no equivalent for user_id, ## time, etc. ## Luckily, these are rarely modified in our testing. Hopefully ## the users will also choose not to modify these. ## TODO: Don't send revs that could end up in @parents_to_send. ## Same for CVS_FIRST_BRANCHED_REVS. Or capture the needed ## data and send the revs on. my $pr = $self->_concoct_branch_rev( $r, $branch_tag, $magic_branch_number, $min_time ); if ( defined $pr ) { ++$empty_branch_placeholders; $pr->previous_id( $r->id ); } } } ## conserve memory $self->{CVS_FIRST_BRANCHED_REVS} = undef; $self->{CVS_BRANCH_PARENTS} = undef; $self->{CVS_ALIASED_BRANCH_TAGS} = undef; $self->{CVS_FOUNDING_REVS} = undef; $self->send_revs; pr "found ", $self->sent_rev_count, " rev(s)", $self->{CVS_CLONE_COUNT} ? " ($self->{CVS_CLONE_COUNT} cloned)" : (), $empty_branch_placeholders ? " ($empty_branch_placeholders for empty branches)" : (), defined $self->{CVS_APPLIED_TAGS_COUNT} ? " with $self->{CVS_APPLIED_TAGS_COUNT} tag applications" : (), "\n"; } # Here's a typical file log entry. # ############################################################################### # #RCS file: /var/cvs/cvsroot/src/Eesh/Changes,v #Working file: src/Eesh/Changes #head: 1.3 #branch: #locks: strict #access list: #symbolic names: # Eesh_003_000: 1.3 # Eesh_002_000: 1.2 # Eesh_000_002: 1.1 #keyword substitution: kv #total revisions: 3; selected revisions: 3 #description: #---------------------------- #revision 1.3 #date: 2000/04/22 05:35:27; author: barries; state: Exp; lines: +5 -0 #*** empty log message *** #---------------------------- #revision 1.2 #date: 2000/04/21 17:32:14; author: barries; state: Exp; lines: +22 -0 #Moved a bunch of code from eesh, then deleted most of it. #---------------------------- #revision 1.1 #date: 2000/03/24 14:54:10; author: barries; state: Exp; #*** empty log message *** #============================================================================= ############################################################################### ## CVS allows multiple branch tags. The last one is the master (oldest) ## tag, we need to clone each rev on the master branch in to all aliased ## tags. sub _note_duplicate_branch_tags { my VCP::Source::cvs $self = shift; my ( $rtags ) = @_; for my $rev_id ( keys %$rtags ) { next unless _is_branch_or_vendor_tag $rev_id; my $tags = $rtags->{$rev_id}; my $master_tag = pop @$tags; next unless @$tags; $self->{CVS_ALIASED_BRANCH_TAGS}->{$master_tag} = $tags; } } sub _clone_aliased_branch_tags { my VCP::Source::cvs $self = shift; if ( $self->{CVS_ALIASED_BRANCH_TAGS} ) { my $t = $self->{CVS_ALIASED_BRANCH_TAGS}; for my $r ( $self->queued_revs ) { next if empty $r->branch_id || ! exists $t->{$r->branch_id}; for ( @{$t->{$r->branch_id}} ) { $self->_concoct_cloned_rev( $r, $_ ); ++$self->{CVS_CLONE_COUNT}; } } } $self->{CVS_ALIASED_BRANCH_TAGS} = undef; } sub _create_rev_from_cvs_log_output { my VCP::Source::cvs $self = shift; return unless keys %{$self->{CVS_LOG_REV}} ; $self->{CVS_LOG_REV}->{comment} = '' if $self->{CVS_LOG_REV}->{comment} eq '*** empty log message ***' ; $self->{CVS_LOG_REV}->{comment} =~ s/\r\n|\n\r/\n/g ; my $r = $self->_create_rev( $self->{CVS_LOG_FILE_DATA}, $self->{CVS_LOG_REV}, empty $self->{CVS_LOG_REV}->{lines} ) ; $self->{CVS_LOG_REV} = {} ; return unless defined $r; if ( $r->action eq "delete" ) { my $nr = $self->last_rev_for_filebranch( $r->source_filebranch_id ); $self->_concoct_edit_rev_for_dead_rev( $nr ) if $nr && $nr->action eq "delete"; } $self->set_last_rev_in_filebranch_previous_id( $r ); if ( $self->id_seen( $r->id ) ) { pr "can't send rev a second time (corrupt RCS file?): ", $r->as_string; } else { $self->queue_rev( $r ) ; $self->{CVS_HIGHEST_REV_TIME} = $r->time if ! defined $self->{CVS_HIGHEST_REV_TIME} || ( defined $r->time && $self->{CVS_HIGHEST_REV_TIME} < $r->time ); 1 ; } } sub parse_cvs_log_output { my ( $self, $fh ) = @_ ; profile_start ref( $self ) . " parse_cvs_log_output()" if profiling; local $_ ; ## DOS, Unix, Mac lineends spoken here. while ( <$fh> ) { s/\r//g if is_win32; ## [1] See bottom of file for a footnote explaining this delaying of ## clearing CVS_LOG_FILE_DATA and CVS_LOG_STATE until we see ## a ========= line followed by something other than a ----------- ## line. ## TODO: Move to a state machine design, hoping that all versions ## of CVS emit similar enough output to not trip it up. ## TODO: BUG: Turns out that some CVS-philes like to put text ## snippets in their revision messages that mimic the equals lines ## and dash lines that CVS uses for delimiters!! PLEASE_TRY_AGAIN: if ( /^===========================================================*$/ ) { $self->_create_rev_from_cvs_log_output; $self->{CVS_SAW_EQUALS} = 1 ; next ; } if ( /^----------------------------*$/ ) { $self->_create_rev_from_cvs_log_output unless $self->{CVS_SAW_EQUALS} ; $self->{CVS_SAW_EQUALS} = 0 ; $self->{CVS_LOG_STATE} = 'rev' ; next ; } if ( $self->{CVS_SAW_EQUALS} ) { $self->_note_duplicate_branch_tags( $self->{CVS_LOG_FILE_DATA}->{RTAGS} ); $self->{CVS_LOG_FILE_DATA} = {} ; $self->{CVS_LOG_STATE} = '' ; $self->{CVS_SAW_EQUALS} = 0 ; } unless ( $self->{CVS_LOG_STATE} ) { if ( /^(RCS file|Working file|head|branch|locks|access list|keyword substitution):\s*(.*)/i ) { $self->{CVS_LOG_FILE_DATA}->{lc( (split /\s+/, $1 )[0] )} = $2 ; } elsif ( /^total revisions:\s*([^;]*)/i ) { } elsif ( /^symbolic names:/i ) { $self->{CVS_LOG_STATE} = 'tags' ; next ; } elsif ( /^description:/i ) { $self->{CVS_LOG_STATE} = 'desc' ; next ; } else { carp "Unhandled CVS log line '$_'" if /\S/ ; } } elsif ( $self->{CVS_LOG_STATE} eq 'tags' ) { if ( /^\S/ ) { $self->{CVS_LOG_STATE} = '' ; goto PLEASE_TRY_AGAIN ; } my ( $tag, $rev_id ) = m{(\S+):\s+(\S+)} ; unless ( defined $tag ) { carp "Can't parse tag from CVS log line '$_'" ; $self->{CVS_LOG_STATE} = '' ; next ; } push( @{$self->{CVS_LOG_FILE_DATA}->{RTAGS}->{$rev_id}}, $tag ) ; } elsif ( $self->{CVS_LOG_STATE} eq 'rev' ) { ( $self->{CVS_LOG_REV}->{rev_id} ) = m/([\d.]+)/ ; $self->{CVS_LOG_STATE} = 'rev_meta' ; next ; } elsif ( $self->{CVS_LOG_STATE} eq 'rev_meta' ) { for ( split /;\s*/ ) { my ( $key, $value ) = m/(\S+):\s+(.*?)\s*$/ ; $self->{CVS_LOG_REV}->{lc($key)} = $value ; } $self->{CVS_LOG_STATE} = 'rev_message' ; next ; } elsif ( $self->{CVS_LOG_STATE} eq 'rev_message' ) { $self->{CVS_LOG_REV}->{comment} .= $_ unless /\Abranches: .*;$/; } } ## Never, ever forget the last rev. "Wait for me! Wait for me!" ## Most of the time, this should not be a problem: cvs log puts a ## line of "=" at the end. But just in case I don't know of a ## funcky condition where that might not happen... $self->_create_rev_from_cvs_log_output ; $self->_note_duplicate_branch_tags( $self->{CVS_LOG_FILE_DATA}->{RTAGS} ); $self->{CVS_LOG_REV} = undef ; $self->{CVS_LOG_FILE_DATA} = undef ; profile_end ref( $self ) . " parse_cvs_log_output()" if profiling; } # Here's a (probably out-of-date by the time you read this) dump of the args # for _create_rev: # ############################################################################### #$file = { # 'WORKING' => 'src/Eesh/eg/synopsis', ## 'SELECTED' => '2', # 'LOCKS' => 'strict', ## 'TOTAL' => '2', # 'ACCESS' => '', # 'RCS' => '/var/cvs/cvsroot/src/Eesh/eg/synopsis,v', # 'KEYWORD' => 'kv', # 'RTAGS' => { # '1.1' => [ # 'Eesh_003_000', # 'Eesh_002_000' # ] # }, # 'HEAD' => '1.2', ### 'TAGS' => { <== not used, so commented out. ### 'Eesh_002_000' => '1.1', ### 'Eesh_003_000' => '1.1' ### }, # 'BRANCH' => '' #}; #$rev = { # 'DATE' => '2000/04/21 17:32:16', # 'comment' => 'Moved a bunch of code from eesh, then deleted most of it. #', # 'STATE' => 'Exp', # 'AUTHOR' => 'barries', # 'REV' => '1.1' #}; ############################################################################### sub _create_rev { my VCP::Source::cvs $self = shift ; my ( $file_data, $rev_data, $is_founding_rev ) = @_ ; $file_data->{working} =~ s{([\\/])[\\/]+}{$1}g; my $norm_name = $self->normalize_name( $file_data->{working} ) ; my $action = $rev_data->{state} eq "dead" ? "delete" : "edit" ; my $type = defined $file_data->{keyword} && $file_data->{keyword} =~ /[o|b]/ ? "binary" : "text" ; #debug map "$_ => $rev_data->{$_}, ", sort keys %{$rev_data} if debugging; my $rev_id = $rev_data->{rev_id}; my $branch_id; my $previous_id; my $denorm_name = $self->denormalize_name( $norm_name ); my $id = "$denorm_name#$rev_id"; ( my $branch_number = $rev_id ) =~ s/\.\d+\z//; # 1.x, 2.x, etc are all on the main branch in CVS. $branch_number = "" unless $branch_number =~ tr/.//; ## Using branch number here instead of the $branch_id (which may be ## a CVS branch tag so that altering the CVS source by moving or ## changing or removing the branch's branch tag does not result in ## different filebranch_ids. my $filebranch_id = "$denorm_name<$branch_number>"; if ( $rev_id =~ /\A(\d+(?:\.\d+)+)\.(\d+)\.(\d+)\z/ ) { ## It's on a branch, set branch_id equal to the tag my ( $parent_rev_id, $branch_number, $rev_number ) = ( $1, $2, $3 ); $previous_id = $self->denormalize_name( $norm_name ) . "#$parent_rev_id" if $rev_number eq "1"; my $magic_branch_number = "$parent_rev_id.0.$branch_number"; my $vendor_branch_number = "$parent_rev_id.$branch_number"; my $branch_ids = exists $file_data->{RTAGS}->{$magic_branch_number} ? $file_data->{RTAGS}->{$magic_branch_number} : exists $file_data->{RTAGS}->{$vendor_branch_number} ? $file_data->{RTAGS}->{$vendor_branch_number} : do { my $invented_tag = "_branch_$magic_branch_number"; # TODO: Consider what happens if two files brancehd # at the same revision number but aren't really in the # same branch. # # Also: consider doing invented branch consolidation # for files that branched around the same time. [ $invented_tag ]; }; ## CVS stores symbols in newest first order. CVS allows multiple ## tags for the same branch. So we take the oldest branch tag ## and use that as the branch_id. Other code will note and ## perform the cloning. $branch_id = $branch_ids->[-1]; } elsif ( ( $rev_id =~ tr/.// ) > 1 ) { die "Did not parse ${rev_id}'s branch number"; } my VCP::Rev $r = VCP::Rev->new( id => $id, name => $norm_name, source_name => $norm_name, rev_id => $rev_id, source_rev_id => $rev_id, type => $type, action => $action, time => $self->parse_time( $rev_data->{date} ), user_id => $rev_data->{author}, comment => $rev_data->{comment}, # state => $rev_data->{state}, labels => $file_data->{RTAGS}->{$rev_id}, branch_id => $branch_id, source_branch_id => $branch_id, source_filebranch_id => $filebranch_id, source_repo_id => $self->repo_id, previous_id => $previous_id, ) ; $self->{CVS_FOUNDING_REVS}->{$id} = 1 if $is_founding_rev; $self->{CVS_APPLIED_TAGS_COUNT} += @{$file_data->{RTAGS}->{$rev_id}} if $file_data->{RTAGS}->{$rev_id}; my $magic_id_prefix = "$rev_id.0"; ## Note all branches from this rev (not including vendor branches, they ## should work without this logic). for my $tag_rev_id ( grep( ( 0 == index $_, $rev_id ) && ( ( 0 == index $_, $magic_id_prefix ) || ! ( tr/.// % 2 ) ## vendor tags have even # of dots ), keys %{$file_data->{RTAGS}} ) ) { ## CVS allows multiple tags per same revision; note only the ## master; cloning will take care of the slaves. push @{ $self->{CVS_BRANCH_PARENTS}->{ $file_data->{RTAGS}->{$tag_rev_id}->[-1]} }, [ $r, $tag_rev_id ]; } if ( $self->continue && $self->dest ) { my $previous_rev_id = $self->dest->last_rev_in_filebranch( $self->repo_id, $filebranch_id ); if ( defined $previous_rev_id && VCP::Rev->cmp_id( $previous_rev_id, $rev_id ) >= 0 ) { return undef; } } return $r; } ## FOOTNOTES: # [1] :pserver:guest@cvs.tigris.org:/cvs hass some goofiness like: #---------------------------- #revision 1.12 #date: 2000/09/05 22:37:42; author: thom; state: Exp; lines: +8 -4 # #merge revision history for cvspatches/root/log_accum.in #---------------------------- #revision 1.11 #date: 2000/08/30 01:29:38; author: kfogel; state: Exp; lines: +8 -4 #(derive_subject_from_changes_file): use \t to represent tab #characters, not the incorrect \i. #============================================================================= #---------------------------- #revision 1.11 #date: 2000/09/05 22:37:32; author: thom; state: Exp; lines: +3 -3 # #merge revision history for cvspatches/root/log_accum.in #---------------------------- #revision 1.10 #date: 2000/07/29 01:44:06; author: kfogel; state: Exp; lines: +3 -3 #Change all "Tigris" ==> "Helm" and "tigris" ==> helm", as per Daniel #Rall's email about how the tigris path is probably obsolete. #============================================================================= #---------------------------- #revision 1.10 #date: 2000/09/05 22:37:23; author: thom; state: Exp; lines: +22 -19 # #merge revision history for cvspatches/root/log_accum.in #---------------------------- #revision 1.9 #date: 2000/07/29 01:12:26; author: kfogel; state: Exp; lines: +22 -19 #tweak derive_subject_from_changes_file() #============================================================================= #---------------------------- #revision 1.9 #date: 2000/09/05 22:37:13; author: thom; state: Exp; lines: +33 -3 # #merge revision history for cvspatches/root/log_accum.in # =head1 SEE ALSO L, L, L. =head1 AUTHOR Barrie Slaymaker =head1 COPYRIGHT Copyright (c) 2000, 2001, 2002 Perforce Software, Inc. All rights reserved. See L (C) for the terms of use. =cut 1