eval '(exit $?0)' && eval 'exec perl -S $0 ${1+"$@"}' & eval 'exec perl -S $0 $argv:q' if 0; # THE PRECEEDING STUFF EXECS perl via $PATH # -*-Fundamental-*- # $Id: //guest/richard_geiger/utils/cvs2p4/bin/genmetadata#45 $ # # Richard Geiger # require 5.000; require "timelocal.pl"; #use bytes; sub dirname { local($dir) = @_; $dir =~ s%^$%.%; $dir = "$dir/"; if ($dir =~ m%^/[^/]*//*$%) { return "/"; } if ($dir =~ m%^.*[^/]//*[^/][^/]*//*$%) { $dir =~ s%^(.*[^/])//*[^/][^/]*//*$%$1%; { return $dir; } } return "."; } use Carp; # ...or flounder. (This will fail unless 'perl' is a perl5!) $| = 1; ($Myname = $0) =~ s%^.*/%%; $Mydir = &dirname($0); $Here = `/bin/pwd`; chop $Here; if ($Mydir ne ".") { chdir "$Mydir" || die "$Myname: can't chdir \"$Mydir\": $!"; } chdir ".." || die "$Myname: can't chdir \"..\": $!"; $Mydir = `/bin/pwd`; chop $Mydir; chdir $Here || die "$Myname: can't chdir \"$Here\": $!"; require "$Mydir/lib/util.pl"; $Usage = <) { chomp $_; if (/^\s*#/ || /^\s*$/) { next; } $_ =~ s/^\s*//; $_ =~ s/\s*#.*$//; $Exclude_tags{$_} = 1; $ntags++; } close EXCLTAGS; print "$Myname: loaded $ntags tag exclusions.\n"; } if (-e "$Convdir/exclude_branches") { if (! open(EXCLTAGS, "<$Convdir/exclude_branches")) { printf STDERR "open(\"$Convdir/exclude_branches\" failed: $!\n"; exit 1; } $ntags = 0; while () { chomp $_; if (/^\s*#/ || /^\s*$/) { next; } $_ =~ s/^\s*//; $_ =~ s/\s*#.*$//; $Exclude_branches{$_} = 1; $ntags++; } close EXCLTAGS; print "$Myname: loaded $ntags branch exclusions.\n"; } } ###### # # Perlstuff for parsing RCS repository files # # Some globals used by these routines... # $Rcs_Inquote = 0; # remembers when we're in a '@' quoted string $Rcs_Eofatal = 1; # die if we hit the end of the file $Rcs_File = "?"; # caller should set this for the error message sub lead { if (defined($Myname)) { return "$Myname: "; } else { return ""; } } my $rcsline_buf; # the read buffer my $readsize = 10 * 1024 * 1024; # read chunk size 10Mb seems good. # The approach we now use bypasses stdio, in order to help handle # large binary archive files. rcsline() knows when it's parsing # a delta text token, and makes use of the fact that it can toss # the delta text away. # # readbuf() is called to refill the input buffer. sub readbuf { my $nread = sysread(RCS, $rcsline_buf, $readsize); if ($nread == 0) { $readbuf_eof = 1; if ($Rcs_Eofatal) { $m = "unexpected eof on \"$Rcs_File\"."; printf STDERR "%s$m\n", &lead(); exit 1; } else { return undef; } } return $nread; } # rcsline() returns the next line of text from the read buffer, except # when parsing a delta text token; since these may be huge, we # leverage knowing that we don't care about the content on them, and # throw them out, always just returning "@text@". This seems to get # acceptable performance with large binary files. # sub rcsline { my ($in_text) = @_; # Flag says when the last token seen was "text", # i.e., when we're parsing the actual delta text my $linesave = ""; # Used to accumulate text longer than rcsline_buf's # read size. Well, in theory, at least. Should only happen # if there are log messages > 10Mb! while (1) { # If the buffer is empty, we read: # if (! $rcsline_buf) { if (&readbuf() == undef) { return undef; } } # ASSUMPTION: RCS will always put the '@' starting a string # at the beginning of a new line... thus we'll # KNOW at this point to expect a delta text. # if ($in_text && ($rcsline_buf =~ /^\@/)) { $rcsline_buf =~ s/^\@//; # Handle the special case where the revision delta text is empty: # if ($rcsline_buf =~ /^\@[^\@]/s) { $rcsline_buf =~ s/^\@//; return "@@\n"; } # Aha! We're in a (possibly long, binary!) string, that we # don't care about the value of; in a "text" token # value... we're just looking for the end! "Return, or bust" # from here! # while (1) { # Can we see the end of the string? # if ($rcsline_buf =~ /^(.*?[^\@])\@([^\@].*)$/s) { $rcsline_buf = $2; return "\@text\@\n"; } elsif ($rcsline_buf =~ /[^\@]\@$/s) { # If the last char in the buffer is an "@", we can't know if # it is an escaped @ or the closing delimiter until we see the # next character... # if (&readbuf() == undef) { return "\@text\@"; } if ($rcsline_buf =~ /^\@/) { # we hit the seam! Eat the @ and continue... $rcsline_buf =~ s/^\@//; } else { # Well I'll be! That -was- a closing string delim! return "\@text\@"; # (We leave rcsline_buf unmolested for future processing) } } else { # We KNOW the entire remainder of the buffer is part of the string value. # and can just chuck it, grabbing a new buffer'sworth. # if (&readbuf() == undef) { return undef; } } } } else { my ($line, $remainder) = ($rcsline_buf =~ m/^([^\n]*\n)(.*)$/s); if ($line) { $rcsline_buf = $remainder; if ($line =~ /\r{0,1}\n$/) { $line =~ s/\r{0,1}\n$//; } return $linesave.$line; } else { $linesave .= $rcsline_buf; $rcsline_buf = ""; } } } } $Rcstok_Buf = ""; $Rcstok_pushed = undef; # Return the next token from the RCS repository file. # Caller should open the file on descriptor RCS. # (Caller should also empty $Rcstok_Buf!) my $lasttok; sub rcstok { my $rcsstr; my $m; my $strpart; if (defined($Rcstok_pushed)) { my $ret = $Rcstok_pushed; $Rcstok_pushed = undef; return ($lasttok = $ret); } $Rcstok_Buf =~ s/^\s+//; if ($Rcstok_Buf eq "") { while (1) { $Rcstok_Buf = &rcsline($lasttok eq "text"); if (! defined ($Rcstok_Buf)) { return undef; } if ($Rcstok_Buf ne "") { last; } } $Rcstok_Buf =~ s/^\s+//; } # : ; id # # Note: the character class for "idchar" assumes all characters # are printable ascii! May break with binary RCS files. # (Actually, I've now convinced myself that there is no # concern here). # if ($Rcstok_Buf =~ /^(:|;|[0-9.]*[^ \t\r\n$,:;@]+)(.*)$/) { $Rcstok_Buf = $2; return ($lasttok = $1); } # num # if ($Rcstok_Buf =~ /^([0-9][0-9.]*)(.*)$/) { $Rcstok_Buf = $2; return ($lasttok = $1); } # string # if ($Rcstok_Buf =~ /^@(.*)$/) { $Rcstok_Buf = $1; $rcsstr = ""; while (1) { if ($Rcstok_Buf eq "") { if ($lasttok ne "text") { $rcsstr .= "\n"; } $Rcstok_Buf = &rcsline($lasttok eq "text"); if (! defined ($Rcstok_Buf)) { return undef; } } if ($Rcstok_Buf =~ /^(.*[^@])@([^@].*)$/) { my $tmp = $1; $Rcstok_Buf = $2; $tmp =~ s/\@\@/\@/g; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } else { $rcsstr .= $tmp; } return ($lasttok = $rcsstr); } if ($Rcstok_Buf =~ /^.*\@\@$/) { # check for multiple @s at the end - have to look for odd number $Rcstok_Buf =~ /^(.*[^@])(\@+)$/; my $tmp = $1; $Rcstok_Buf = $2; $tmp =~ s/\@\@/\@/g; if ($lasttok eq "text") { $rcsstr .= $tmp; } my $odd = length($Rcstok_Buf) % 2; if ($odd) { $Rcstok_Buf = substr($Rcstok_Buf, 1); } $Rcstok_Buf =~ s/\@\@/\@/g; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } else { $rcsstr .= $Rcstok_Buf; } $Rcstok_Buf = ""; if ($odd) { return ($lasttok = $rcsstr); } } if ($Rcstok_Buf =~ /^(.*)\@$/) { my $tmp = $1; $Rcstok_Buf = ""; $tmp =~ s/\@\@/\@/g; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } else { $rcsstr .= $tmp; } return ($lasttok = $rcsstr); } if ($Rcstok_Buf =~ /^\@([^@].*)$/) { $Rcstok_Buf = $1; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } return ($lasttok = $rcsstr); } if ($Rcstok_Buf =~ /^\@$/) { $Rcstok_Buf = ""; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } return ($lasttok = $rcsstr); } $Rcstok_Buf =~ s/\@\@/\@/g; if ($lasttok eq "text") { $rcsstr = "[deltatext]"; } else { $rcsstr .= $Rcstok_Buf; } $Rcstok_Buf = ""; } } $m = "rcstok(): internal error: \$Rcstok_Buf <$Rcstok_Buf>"; printf STDERR "%s$m\n", &lead(); exit 1; } sub dirname { my ($dir) = @_; $dir =~ s%^$%.%; $dir = "$dir/"; if ($dir =~ m%^/[^/]*//*$%) { return "/"; } if ($dir =~ m%^.*[^/]//*[^/][^/]*//*$%) { $dir =~ s%^(.*[^/])//*[^/][^/]*//*$%$1%; { return $dir; } } return "."; } sub skip_to_rcstok { my ($this) = @_; my $tok; while (($tok = &rcstok()) ne $this) { }; } sub skip_to_deltas { my $tok; # Called after we encounter ";" for symbols. We must now skip: # # locks {id : num}*; {strict ;} # { comment {string}; } # { expand {string}; } # { newphrase }* while (1) { my $fatalsave = $Rcs_eofatal; $tok = &rcstok(); if ($tok eq "expand") { $RCS_expand = &rcstok(); if (&rcstok() ne ";") { die "$Myname: skip_to_deltas(): expected ';' after expand."; } next; } if ($tok =~ /^[0-9]/) { $Rcstok_pushed = $tok; last; } while (1) { # Turning this off here because it looks like cvs problems can create # delta-less ,v files; this allows us to handle this. # $Rcs_Eofatal = 0; $tok = &rcstok(); $Rcs_Eofatal = $fatalsave; if (! defined($tok)) { return undef; } if ($tok eq ";") { last; } } } return 1; } sub sdump { my ($s) = $1; my @s = split(//, $s); my $ret = ""; foreach my $c (@s) { $ret .= sprintf(" %02x", ord($c)); } return $ret; } sub test_rcstoks { my $tok; print "open $ARGV[1]\n"; open(RCS, "<$ARGV[1]") || die; $Rcstok_Buf = ""; $Rcs_Eofatal = 0; while (defined($tok = &rcstok)) { print "<$tok>\n"; } exit 0; } sub setrevs { my($d_rev, $d_next, $d_branches, $d_date, $d_author, $d_state) = @_; my($b_rev); $RCS_Revs{$d_rev} = "$d_next:$d_branches"; $d_date = "19$d_date" if length(( split( /\./, $d_date ))[0]) < 4; $RCS_Dates{$d_rev} = "$d_date"; $RCS_Authors{$d_rev} = "$d_author"; $RCS_States{$d_rev} = "$d_state"; if ($d_rev =~ /^1\.1\.1(\.[0-9]+)?$/) { # We have a "vendor" branch - spoof a branch tag for it. # my $import_branch = "import"; if ($PureRCS) { $import_branch = "1.1.1"; } $RCS_Branchtags{$import_branch} = "1.1.0.1"; $RCS_Tags{$import_branch} = "1.1.0.1"; } if ($d_rev =~ /^[0-9]+\.[0-9]+$/) { $RCS_Prevs{$d_rev} = $d_next; if ($d_next) { $RCS_Nexts{$d_next} = $d_rev; } } else { if ($d_next) { $RCS_Prevs{$d_next} = $d_rev; } $RCS_Nexts{$d_rev} = $d_next; } foreach $b_rev (split(/ /, $d_branches)) { $RCS_Prevs{$b_rev} = $d_rev; } } # initialize RCS_Tags, RCS_Revs, (etc.) from an RCS ,v file. # sub set_RCS_revs { my ($path, $do_texts) = @_; my $repdir; my $repfile; my $tag; my $rev; my $tok; my ($d_havedelta, $d_branches, $d_next, $d_rev); my ($ext, $format); my $rcspath; my $msg; my $admaci = 0; my $File; undef $RCS_File; undef $RCS_Valid; undef $RCS_expand; undef $RCS_exec; undef %RCS_Tags; # both plain and branch (and special "head") undef %RCS_Branchtags; # branch, only undef %RCS_rev_brtags; # inverse of the above - keyed by branch tag value undef %RCS_Revs; undef %RCS_States; undef %RCS_Authors; undef %RCS_Dates; undef %RCS_Prevs; undef %RCS_Nexts; undef $RCS_Branch; undef %RCS_Texts; undef $RCS_import_is_main; ($Rcs_File = $path) =~ s%^.*/%%; ($File = $Rcs_File) =~ s/,v$//; $repdir = &dirname($path); if (-r "$repdir/.adamci,v") { $adamci = 1; } $repfile = $path; $rcspath = "<$repfile"; if (! open(RCS, $rcspath)) { printf STDERR "%scan't open \"$rcspath\": $!\n", &lead(); return 0; } $RCS_File = $rcspath; if (-x $repfile) { $RCS_exec = "x"; } $Rcstok_Buf = ""; &skip_to_rcstok("head"); $RCS_Tags{"head"} = &rcstok(); $tok = &rcstok(); $tok = &rcstok(); if ($tok eq "branch") { $RCS_Branch = &rcstok(); } if ($RCS_Branch eq "1.1.1") { $RCS_import_is_main = 1; } &skip_to_rcstok("symbols"); while (1) { $tok = &rcstok(); if ($tok eq ";") { last; } $tag = $tok; &rcstok(); $rev = &rcstok(); if (($cnt = $rev =~ tr/\./\./) % 2 == 0) { if ($rev ne "1.1.1") # 'cause we call it "import"! { # Handle "RCS" branch tags: # my @nums = split(/\./, $rev); splice @nums, $#nums, 0, (0); $rev = join(".", @nums); } } if ((! $PureRCS) && $rev =~ /\.0\.[0-9]+$/) { if (! $adamci) { $RCS_Tags{$tag} = $rev; $RCS_Branchtags{$tag} = $rev; if (defined($BRANCH_FLASH)) { $tag =~ s/$BRANCH_FLASH$//; } $Brtags{$tag} = 1; if ($RCS_rev_brtags{$rev}) { print "WARNING: file: $RCS_File: dup CVS branch tags on rev <$rev> (tag <$tag>)(ignored)\n"; } else { $RCS_rev_brtags{$rev} = $tag; } } } elsif (($cnt = $rev =~ tr/\./\./) % 2 == 1) { $RCS_Tags{$tag} = $rev; my $rcspath; ($rcspath = $repfile) =~ s/,v$//; if ($IMPORTTAGSPOOF && $rev eq "1.1.1.1") { $rev = "1.1"; } if (! defined($Exclude_tags{$tag})) { print LABELS "$tag$S$rcspath$S$rev\n"; # if (! defined($RCS_Revtags{$tag})) # { $RCS_Revtags{$tag} = [ ] ; } # if ($IMPORTTAGSPOOF && $rev eq "1.1.1.1") { $rev = "1.1"; } # push(@{$RCS_Revtags{$tag}}, "$reppath/$rev"); # don't want to wipe out previously observed one, as it might # already have found the "real" mapping! # if (! defined($Tags{$tag})) { $Tags{$tag} = ""; } } } } # OK, we get here having seen every tag in the file. See whether, # for any of the tags we saw in this file, we can determine the # branch for that tag, and remember the mapping globally. try_tag: foreach my $tag (sort(keys(%RCS_Tags))) { # Only want to consider rev tags: # if (defined($RCS_Branchtags{$tag})) { next; } my $tagrev = $RCS_Tags{$tag}; # So: do any of the known branch tags in this file select the # tagged revision? # my $tagrev_brnum; ($tagrev_brnum = $tagrev) =~ s/\.\d+$//; # Figure out which branch this revision is present in by virtue # of *other* than being the branch point: # my $sel_br; if ($tagrev_brnum =~ /^\d+$/) { $sel_br = "main"; } else { my ($sel_brbase, $sel_brnum) = ($tagrev_brnum =~ /^(.*)\.(\d+)$/); my $sel_brrev = "$sel_brbase.0.$sel_brnum"; if ($RCS_rev_brtags{$sel_brrev}) { $sel_br = $RCS_rev_brtags{$sel_brrev} } } if (! $sel_br) { # We can't make a determination :-( # (Might want to track these TBD; shouldn't happen much if at all, right? # One could argue that this would represent CVS "corruption" - a tagless branch! # next try_tag; } # OK, so now, is our revision also selected by virtue of being a branch point? # (If so, it's ambiguous for this file) # my $try_limit = 5; my $g = 0; my $try_tag; for ($i = 1; $g <= $try_limit; $i++) { my $try_rev = "$RCS_Tags{$tag}.0.$i"; if ($RCS_rev_brtags{$try_rev}) { next try_tag; } # If we see that the branch number exists, we should keep looking. # We want to give up when we either A) find the branch or # B) have found $try_limit unused branch bumbers in a row. # Since CVS is supposed to dole them out sequentially, this # -should- be sufficient to decide that there are no more. # This should be better for efficiency, compared to checking # every key in $RCS_rev_brtags. # if (defined($RCS_rev_brtags{$try_tag})) { $g = 0; next; } $g++; } # so... if we get here, in theory, the heuristic was applicable! # $sel_br has the presumptive mapping. # Now, make sure we didn't get a different answer than from some previous # file: if ($Tags{$tag} && ($Tags{$tag} ne $sel_br)) { if ($Tags{$tag} eq "UNMAPPED") { next try_tag; } # Probably want to log and proceed... TBD print "\nERROR: file: $RCS_File: conflicting branch determination for tag \"$tag\": ". " <$Tags{$tag}> vs <$sel_br>; "; print " (previous file:\"$Tagfiles{$tag}\").\n"; # The idea is to not trust the result for this tag! # $Tags{$tag} = "UNMAPPED"; } else { # Ahhhhhhhhhh that feels so good.... # $Tags{$tag} = $sel_br; # And remember where we saw it!: # $Tagfiles{$tag} = $RCS_File; } } if ($Prescan) { return 1; } if ($adamci) { undef %metadata; eval `$CO -q -p $repdir/.adamci,v`; # TBD: Optimize this to only reread if new dir? foreach my $p (keys(%{$metadata{'context_rules'}})) { my $k; ($k) = keys( %{${${$metadata{'context_rules'}}{$p}}{$File}} ); if (defined($k) && $k ne "TBB") { my $rev = ${${${${$metadata{'context_rules'}}{$p}}{$File}}{$k}}[1]; $rev =~ s/^(.*\.\d+)\.(\d+)$/$1.0.$2/; $RCS_Tags{$p} = $rev; $RCS_Branchtags{$p} = $rev; } } } if (&skip_to_deltas() == undef) { return undef; } $d_rev = ""; $d_havedelta = 0; $d_branches = ""; $d_next = ""; while (1) { $tok = &rcstok(); if ($tok =~ /^[0-9.]+$/) { if ($d_havedelta) { &setrevs($d_rev, $d_next, $d_branches, $d_date, $d_author, $d_state); } $d_rev = $tok; $d_havedelta = 1; $d_branches = ""; $d_next = ""; } elsif ($tok eq "branches") { while (1) { if (($tok = &rcstok()) eq ";") { last; } if ($PureRCS) { my ($l, $b, $r) = ($tok =~ /(.*)\.(\d+)\.(\d+)$/); my $tag = "$l.$b"; $RCS_Branchtags{$tag} = $rev; } if ($d_branches ne "") { $d_branches .= " "; } $d_branches .= $tok; } } elsif ($tok eq "date") { $tok = &rcstok(); if ($tok ne ";") { $d_date = $tok; &rcstok(); } } elsif ($tok eq "author") { $tok = &rcstok(); if ($tok ne ";") { $d_author = $tok; &rcstok(); } } elsif ($tok eq "state") { $tok = &rcstok(); if ($tok ne ";") { $d_state = $tok; &rcstok(); } } elsif ($tok eq "next") { $tok = &rcstok(); if ($tok ne ";") { $d_next = $tok; &rcstok(); } } # Note: "format" and "ext" seem to be extentions used by MKS # Source Integrity. They don't show up in my rcsfile(5) man # page! elsif ($tok eq "format") { $tok = &rcstok(); if ($tok ne ";") { $format = $tok; &rcstok(); } } elsif ($tok eq "ext") { $tok = &rcstok(); $ext = $tok; } elsif ($tok eq "desc") { last; } else { &skip_to_rcstok(";"); } } if ($d_havedelta) { &setrevs($d_rev, $d_next, $d_branches, $d_date, $d_author, $d_state); } if (! defined($do_texts)) { close RCS; $RCS_Valid = 1; return 1; } $Rcs_Eofatal = 0; while (1) { $tok = &rcstok(); if (! defined($tok)) { last; } if ($tok =~ /[0-9.]+/) { $d_rev = $tok; } elsif ($tok eq "text") { $msg = &rcstok(); if ($msg eq "") { $RCS_Texts{$d_rev} = 0; } else { $RCS_Texts{$d_rev} = 1; } } elsif ($tok eq "log") { $log = &rcstok(); $RCS_Logs{$d_rev} = $log; } } $Rcs_Eofatal = 1; close RCS; $RCS_Valid = 1; return 1; } sub rcs_tip { my ($rev) = @_; my $next; # Find the tip of the branch... # while (1) { if (! defined($RCS_Revs{$rev})) { return "???"; } ($next) = split(/:/, $RCS_Revs{$rev}); if ($next eq "") { return $rev; } $rev = $next; } } # given a "CVS line spec" (revision #, "head", or a tag) # sub rev_on_line { my($line) = @_; if ($line eq $TRUNKLINE) { $line = "head"; } # else # { $line = "${line}_BRANCH"; } if (defined($RCS_Tags{$line})) { $line = $RCS_Tags{$line}; } elsif ($line !~ /^[0-9.]+$/) { return "none"; } if ($line =~ /\.0\.([0-9]+)$/) { # It's a CVS branch revision number... demunge it: # $line =~ s/\.0(\.[0-9]+)$/$1/; # OK, see whether the branch actually exists: # (We have an assumption here that first rev is always ".1") # $line = "$line.1"; if (! defined($RCS_Revs{$line})) { # Nope, so fall back to the root, which we know to be an # existing revision... $line =~ s/\.[0-9]+\.[0-9]+$//; return $line; } # Yep, the branch exists; so it *is* a branch; so, we go out to # the tip. (Right?) # return &rcs_tip($line); } # OK, do we have an RCS branch or an RCS revision number? (count # the dots) # if (($line =~ tr/\././) % 2) { # An odd number of dots... it's a revision number # if (defined($RCS_Revs{$line})) { return $line; } return "none"; # Or should we assert? } else { # An even number of dots... it's a branch number # (We have an assumption here that first rev is always ".1") # return &rcs_tip("$line.1"); } } # Is rev "$this" < rev "$that"? # Note: "" is considered infinitely high # revs must be of the same order (I.e., same # of "."s) # sub rev_lt { my($this, $that) = @_; my(@this, @that); if (! $that) { return 1; } @this = split(/\./, $this); @that = split(/\./, $that); while (1) { $this_n = shift(@this); $that_n = shift(@that); if ($this_n < $that_n) { return 1; } if ($this_n > $that_n) { return 0; } if ($#this < 0) { return 0; } } } # Note: "" is considered infinitely high # sub linerev_gt { my($this, $that) = @_; my $ret; if (! $that) { $ret = 1; } else { my $thisord, $thatord; $thisord = ($this =~ tr/\././); $thatord = ($that =~ tr/\././); if ($thisord < $thatord) { $ret = 1; } elsif ($thisord > $thatord) { $ret = 0; } else { $ret = &rev_lt($that, $this); } } return $ret; } # Maximum size for a log message we'll keep. # Messages beyond this get truncated, to accomodate a limitation # on the key/value pair size in ndbm. That's life. # $MAXSZ = 256*3; # Generate the metadata for a single file # sub dofile { local($dir, $file) = @_; if ($file !~ /,v$/) { return; } if ($IGNOREFILES && $file =~ /$IGNOREFILES/) { return; } if ($file =~ /[\000-\037\177-\377]/) { print "$Myname: RCS filename with non-printable characters (skipped): "; $l = length($file); for ($i = 0; $i <= $l; $i++) { $c = substr($file, $i, 1); if ($c =~ /[\000-\037\177-\377]/) { printf "\\%03o", ord($c); } else { print "$c"; } } print "\n"; return; } elsif ($file =~ /\.\.\./) { print "$Myname: RCS filename with illegal Perforce characters (skipped): $file\n"; return; } #if ($file ne "rm.c,v") { return; } undef %RCS_lines; undef %RCS_Branches; undef $Firstusedrev; print "$dir/$file"; # This parses the RCS information from the ,v file, filling # in various data structures that we use, below. # if (&set_RCS_revs("$dir/$file", 0) == undef) { print " (empty)\n"; return; } # empty ,v # What RCS keyword expansion options are in effect? # (We use this to detect binary files) # $options = "${RCS_expand}$RCS_exec"; if (! $options) { $options = "-"; } @path = split(/\//, "$dir/$file"); $file = pop(@path); $file =~ s/,v$//; if ($path[$#path] eq "Attic") { pop @path; } $na_dir = join("/", @path); $path = sprintf("%s%s%s", $dir, $dir ? "/" : "", $file); $na_path = sprintf("%s%s%s", $na_dir, $na_dir ? "/" : "", $file); print " -> $path\n"; # The "defined($Filesseen{$path})" saves lots of stat()s! # if (defined($Filesseen{$path})) { my @p = split(/\//, "$na_path"); splice(@p, $#p, 0, "Attic"); my $a_path = join("/", @p); # Users have seen this, which previous caused mysterious death # in the sort phase... let's be a little more informative: # if (-f "$na_path,v" && -f "$a_path,v") { die "assert: CVS repository has both\n $na_path\nand\n $a_path"; } else { die "assert: dofile(): duplicate path: $na_path"; } } $Filesseen{$path} = 1; if ($Prescan) { return; } # For all of the branches we see, store the tip revision in # $HAVELINES{$line}; this is also where we weed out # codelines we are not interested in. # foreach $line ((keys %RCS_Branchtags), $TRUNKLINE) { # Note: lines added to the exclude_branches file should # give the actual, complete branch tag name, not the # "de-flashed" (if any) rendition. # if (defined($Exclude_branches{$line})) { next; } $no_flash_line = $line; if (defined($BRANCH_FLASH)) { $no_flash_line =~ s/$BRANCH_FLASH$//; } if ($WANTLINES && ! (defined($WANTLINES{$no_flash_line}))) { next; } if (($tiprev = &rev_on_line($line)) eq "none") { next; } $HAVELINES{$line} = $tiprev; } # Now we go through each line, to build a list of the RCS revs that # need to be exported into the metadata stream. # while (1) # We have more lines to deal with... { (@k) = (keys %HAVELINES); if ($#k < 0) { last; } my $theline; # Choose the highest numbered line of the lowest "order" for # the next one to export... this will always pickup lines on # branches nearer the trunk first, so the subsequent branches # will have a place to branch from! # foreach $k (@k) { if ($k eq $TRUNKLINE) { $theline = $k; last; } # if both lines select the *same* revision... # if ($HAVELINES{$k} eq $HAVELINES{$theline}) { # ...take the one with the lower branch tag order first # if (&linerev_gt($RCS_Branchtags{$k}, $RCS_Branchtags{$theline})) { $theline = $k; next; } } if (&linerev_gt($HAVELINES{$k}, $HAVELINES{$theline})) { $theline = $k; } } $rev = $tiprev = &rev_on_line($theline); $t = $theline; if (defined($BRANCH_FLASH)) { $t =~ s/$BRANCH_FLASH$//; } # This is where we build the list of codelines we've encountered. # $All_lines{$t} = 1; if ( (defined($RCS_lines{$rev})) && ($theline ne $TRUNKLINE) && (($rev =~ tr/\././) < ($RCS_Branchtags{$theline} =~ tr/\././))) { if ($RCS_Branches{$rev}) { $RCS_Branches{$rev} .= ":"; } $RCS_Branches{$rev} .= $t; } else { while ($rev && ! defined($RCS_lines{$rev})) { $RCS_lines{$rev} = $theline; $rev = $RCS_Prevs{$rev}; } if ($rev) { if ($RCS_Branches{$rev}) { $RCS_Branches{$rev} .= ":"; } $RCS_Branches{$rev} .= $t; } } # We test for "if $rev" here cause it may have gone null if the while loop # above ran off the end... # if ($rev && (($rev =~ tr/\././) == 1) && &rev_lt($rev, $Firstusedrev)) { $Firstusedrev = $rev } delete $HAVELINES{$theline}; } # OK, we have the set of revisions to export - write them to the # metadata stream. # foreach $rev (keys %RCS_Revs) { $revkey = "$path/$rev"; $state = $RCS_States{$rev}; $author = $RCS_Authors{$rev}; $date = $RCS_Dates{$rev}; my ($yr, $mo, $da, $hr, $mi, $se) = split (/\./, $date); $date = timegm($se,$mi,$hr,$da,$mo - 1,$yr); $line = $RCS_lines{$rev}; if (defined($BRANCH_FLASH)) { $line =~ s/$BRANCH_FLASH$//; } $branches = $RCS_Branches{$rev}; # Detect revisions before the first branch point, and # omit them if we're not doing ALLTHEWAYBACK. # if ( (! $ALLTHEWAYBACK) && ($line eq $TRUNKLINE) && ($rev ne $RCS_Tags{"head"}) && $Firstusedrev && &rev_lt($rev, $Firstusedrev)) { next; } if (! $line) { next; } if (! $branches) { $branches = "-"; } if ((! $ALLTHEWAYBACK) && $rev eq $Firstusedrev) { $prevrev = "-"; } elsif ($RCS_Prevs{$rev}) { $prevrev = $RCS_Prevs{$rev}; } else { $prevrev = "-"; } $All_lines{$line} = 1; if ($RCS_import_is_main) { if ($line eq "import") { $line = $TRUNKLINE; } my @btmp = split(/ /, $branches); my $newb = ""; foreach my $b (@btmp) { if ($b eq "import") { next; } if ($newb) { $newb .= " "; } $newb .= $b; } $branches = $newb; } print METATMP "$revkey$S$date$S$author$S$state$S$line$S$branches$S$prevrev$S$options\n"; # MAXSZ derives from a ndbm limitation on the size of a key/entry pair. # at (256*3) it allows for a $revkey up to 250 chars or so. # $logmsg = substr($RCS_Logs{$rev}, 0, $MAXSZ); if ($logmsg !~ /\n$/) { $logmsg .= "\n"; } if (length($logmsg)+length($revkey) > 1010) { print "$Myname: revkey + log too long for <$revkey>\n"; exit 1; } $MSGS{$revkey} = $logmsg; } } # option switch variables get defaults here... $Convdir = ""; $Boolopt = 0; $Valopt = 0; $Prescan = 0; while ($#ARGV >= 0) { if ($ARGV[0] eq "-testtoks") { &test_rcstoks; } if ($ARGV[0] eq "-prescan") { $Prescan = 1; shift; next; } elsif ($ARGV[0] eq "-valopt") { shift; if ($ARGV[0] < 0) { &usage; } $Valopt = $ARGV[0]; shift; next; } elsif ($ARGV[0] eq "-help") { &help; } elsif ($ARGV[0] =~ /^-/) { &usage; } if ($Args ne "") { $Args .= " "; } push(@Args, $ARGV[0]); shift; } if ($#Args ne 0) { &usage; } $Convdir = $Args[0]; $Metatmp = "$Convdir/metatmp"; $Metadata = "$Convdir/metadata"; $Labels = "$Convdir/labels"; $Tags = "$Convdir/tags"; $Tagfiles = "$Convdir/tagfiles"; $Brtags = "$Convdir/brtags"; $Logmsgs = "$Convdir/logmsgs"; $Filesseen = "$Convdir/filesseen"; $Changes = "$Convdir/changes"; $Revmap = "$Convdir/revmap"; $Clientdir = "$Convdir/p4"; require "$Convdir/config"; &load_excludes(); # (Handle either f or f.db or f.pag, f,dir style dbs): # if (&s("/bin/rm -rf $Logmsgs $Logmsgs.db $Logmsgs.pag $Logmsgs.dir ". "$Changes $Clientdir ". "$Tags.txt $Tags $Tags.db $Tags.pag $Tags.dir ". "$Tagfiles.txt $Tagfiles $Tagfiles.db $Tagfiles.pag $Tagfiles.dir ". "$Brtags.txt $Brtags.db $Brtags.pag $Brtags.dir ". "$Revmap $Revmap.db $Revmap.pag $Revmap.dir $Labels")) { die "/bin/rm -rf $Logmsgs ..."; } use DB_File; $DBMCLASS="DB_File"; #$myhashinfo = new DB_File::HASHINFO; #$myhashinfo->{bsize} = 4096; $myhashinfo = new DB_File::BTREEINFO; if (! tie(%Files_seen, $DBMCLASS, $Filesseen, O_CREAT|O_RDWR, 0666, $myhashinfo)) { print "$Myname: can't tie \"$Filesseen\": $!\n"; exit 1; } if (! tie(%MSGS, $DBMCLASS, $Logmsgs, O_CREAT|O_RDWR, 0666, $myhashinfo)) { print "$Myname: can't tie \"$Logmsgs\": $!\n"; exit 1; } if (! open(LABELS, ">$Labels")) { print "$Myname: can't open \">$Labels\": $!\n"; exit 1; } # The $Tags hash is keyed by the tag name. It's value is set to the # branch tag of the branch it belongs to, iff the mapping can be # detemermined by observing that a tagged revision is present in # exactly one branch, i.e., has moved beyond the branch's branch # point, AND is not selected as the base of some other branch. # if (! tie(%Tags, $DBMCLASS, $Tags, O_CREAT|O_RDWR, 0666, $myhashinfo)) { print "$Myname: can't tie \"$Tags\": $!\n"; exit 1; } # The Tagfiles hash remembers for each tag mapped by the heuristic, # the file where the mapping was established. This is useful in cases # where there are conflicts. # if (! tie(%Tagfiles, $DBMCLASS, $Tagfiles, O_CREAT|O_RDWR, 0666, $myhashinfo)) { print "$Myname: can't tie \"$Tagfiles\": $!\n"; exit 1; } # The $Brtags hash is keyed by branch tag name, and the value is the # actual branch number. # if (! tie(%Brtags, $DBMCLASS, $Brtags, O_CREAT|O_RDWR, 0666, $myhashinfo)) { print "$Myname: can't tie \"$Brtags\": $!\n"; exit 1; } if (! open(METATMP, ">$Metatmp")) { print "$Myname: can't open \">$Metatmp\": $!\n"; exit 1; } #chdir $CVS_MODULE || die "$Myname: can't chdir \"$CVS_MODULE\": $!"; #$CVS_MODULE = `/bin/pwd`; chop $CVS_MODULE; #chdir $Here || die "$Myname: can't chdir \"$Here\": $!"; &traverse($CVS_MODULE, 0, "dofile"); close METATMP; #close REVTAGS; untie %MSGS; if (! open(TAGS, ">$Tags.txt")) { print "$Myname: can't open \">$Tags.txt\": $!\n"; exit 1; } foreach $tag (sort(keys(%Tags))) { print TAGS "$tag\t"; if ($Tags{$tag}) { print TAGS "$Tags{$tag}\n"; } else { print TAGS "UNMAPPED\n"; } } close TAGS; print "Wrote $Tags.txt\n"; untie %Tags; untie %Tagfiles; &s("rm -f $Tags $Tags.db $Tags.pag $Tags.dir"); &s("rm -f $Tagfiles $Tagfiles.db $Tagfiles.pag $Tagfiles.dir"); if (! open(BRTAGS, ">$Brtags.txt")) { print "$Myname: can't open \">$Brtags.txt\": $!\n"; exit 1; } foreach my $brtag (sort(keys(%Brtags))) { print BRTAGS "$brtag\n"; } close BRTAGS; print "Wrote $Brtags.txt\n"; untie %Brtags; &s("rm -f $Brtags $Brtags.db $Brtags.pag $Brtags.dir"); if ($Prescan) { exit 0; } sub metasort { my @a = split(/$S/, $a); my @b = split(/$S/, $b); # The revision time is the primary sort key # - But this is now handled by the external sort; we still # do secondary and tertiary sort keys, below # # if ($a[1] != $b[1]) { return $a[1] <=> $b[1]; } $a[0] =~ s/^(.*)\///; my $apath = $1; $b[0] =~ s/^(.*)\///; my $bpath = $1; # Next is the pathname # if ($apath ne $bpath) { return $apath cmp $bpath; } # If we're still tied, it goes to the revision number! # @aa = split(/\./, $a[0]); @bb = split(/\./, $b[0]); for (my $i = 0; $i <= $#aa; $i=$i+2) { if (! defined($bb[$i])) { return 1; } # a has more positions, thus greater if ($aa[$i] < $bb[$i]) { return -1; } # a is less than b, thus less if ($aa[$i] > $bb[$i]) { return 1; } # and vice-versa # if they are equal, we look to the next position: # if (! defined($aa[$i+1])) { die "impossible sort key (RCS rev) \"$a[0]\"?\n"; } if (! defined($bb[$i+1])) { die "impossible sort key (RCS rev) \"$b[0]\"?\n"; } if ($aa[$i+1] < $bb[$i+1]) { return -1; } # a is less than b, thus less if ($aa[$i+1] > $bb[$i+1]) { return 1; } # and vice-versa # Otherwise, we go on to the next level... } if ($#bb > $#aa) { return -1; } die "impossible equal sort keys:\n <$a>\n <$b>\n"; } my $cmd = "sort -n -t $S -k 2 < $Metatmp |"; if (! open(METASORT, $cmd)) { print "$Myname: can't open \"$cmd\": $!\n"; exit 1; } if (! open(META, ">$Metadata")) { print "$Myname: can't open \">$Metadata\": $!\n"; exit 1; } # Do the sorting in chunks, per primary sort key. (We're going through # all of this, BTW, in order to constrain genmetadata's memory # footprint, which was getting huge when we held all of the tags and # metadata in-core) # my $t = 0; my @Meta; while () { chomp; my @r = split(/$S/, $_); if ($r[1] ne $t) { if ($#Meta >= 0) { my @Metasorted = sort metasort @Meta; foreach my $m (@Metasorted) { print META "$m\n"; } } $t = $r[1]; @Meta = (); } push (@Meta, $_); } if ($#Meta >= 0) { my @Metasorted = sort metasort @Meta; foreach my $m (@Metasorted) { print META "$m\n"; } } close METASORT; close META; #if (! open(LABELS, ">$Labels")) # { print "$Myname: can't open \">$Labels\": $!\n"; exit 1; } # #if (! open(REVTAGS, "sort $Revtags |")) # { print "$Myname: can't open \"sort $Revtags |\": $!\n"; exit 1; } # #my $curlabel; #while () # { # my ($label, $pathrev) = ($_ =~ m/^([^ ]+) (.*)/); # if ($label ne $curlabel) # { # $sublabel = $label; # if ($label =~ m/[\#\@]/) # { # $sublabel =~ s/\#/_hash_/g; # $sublabel =~ s/\@/_atsym_/g; # } # print LABELS "$sublabel\n"; # $curlabel = $label; # } # print LABELS " $pathrev\n"; # } #close REVTAGS; #close LABELS; #foreach $label (sort(keys(%RCS_Revtags))) # { # $sublabel = $label; # if ($label =~ m/[\#\@]/) # { # $sublabel =~ s/\#/_hash_/g; # $sublabel =~ s/\@/_atsym_/g; # } # print LABELS "$sublabel\n"; # foreach $rev (@{$RCS_Revtags{$label}}) # { print LABELS " $rev\n"; } # } #close LABELS; #&s("rm $Revtags"); $Lines = "$Convdir/lines"; if (! open(LINES, ">$Lines")) { print "$Myname: can't open \">$Lines\": $!\n"; } else { print "===== Lines referenced:\n"; print LINES "===== Lines referenced:\n"; foreach $line (sort keys %All_lines) { print "$line\n"; print LINES "$line\n"; } close LINES; } exit 0;