X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FSolverGit.pm;h=12024dbcc6dfa143b444e0e24959e1b405991a0c;hb=37df9d6964061f437e37646d8858ade4b6f13a24;hp=bf806f65661a5d6637ef89401f34104d25c7ffde;hpb=ea30a1fcab1823062e5375f2670dadc80c712f74;p=public-inbox.git diff --git a/lib/PublicInbox/SolverGit.pm b/lib/PublicInbox/SolverGit.pm index bf806f65..12024dbc 100644 --- a/lib/PublicInbox/SolverGit.pm +++ b/lib/PublicInbox/SolverGit.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2019 all contributors +# Copyright (C) 2019-2020 all contributors # License: AGPL-3.0+ # "Solve" blobs which don't exist in git code repositories by @@ -9,12 +9,11 @@ # local filesystem layouts in the process. package PublicInbox::SolverGit; use strict; -use warnings; -use 5.010_001; +use v5.10.1; use File::Temp 0.19 (); # 0.19 for ->newdir use Fcntl qw(SEEK_SET); use PublicInbox::Git qw(git_unquote git_quote); -use PublicInbox::MsgIter qw(msg_iter msg_part_text); +use PublicInbox::MsgIter qw(msg_part_text); use PublicInbox::Qspawn; use PublicInbox::Tmpfile; use URI::Escape qw(uri_escape_utf8); @@ -34,6 +33,12 @@ my $OID_MIN = 7; # work fairly. Other PSGI servers may have trouble, though. my $MAX_PATCH = 9999; +my $LF = qr!\r?\n!; +my $ANY = qr![^\r\n]+!; +my $MODE = '100644|120000|100755'; +my $FN = qr!(?:("?[^/\n]+/[^\r\n]+)|/dev/null)!; +my %BAD_COMPONENT = ('' => 1, '.' => 1, '..' => 1); + # di = diff info / a hashref with information about a diff ($di): # { # oid_a => abbreviated pre-image oid, @@ -42,7 +47,7 @@ my $MAX_PATCH = 9999; # hdr_lines => string of various header lines for mode information # mode_a => original mode of oid_a (string, not integer), # ibx => PublicInbox::Inbox object containing the diff -# smsg => PublicInbox::SearchMsg object containing diff +# smsg => PublicInbox::Smsg object containing diff # path_a => pre-image path # path_b => post-image path # n => numeric path of the patch (relative to worktree) @@ -65,34 +70,38 @@ sub ERR ($$) { die $err; } -# look for existing objects already in git repos +# look for existing objects already in git repos, returns arrayref +# if found, number of remaining git coderepos to try if not. sub solve_existing ($$) { my ($self, $want) = @_; + my $try = $want->{try_gits} //= [ @{$self->{gits}} ]; # array copy + my $git = shift @$try or die 'BUG {try_gits} empty'; my $oid_b = $want->{oid_b}; - my $have_hints = scalar keys %$want > 1; - my @ambiguous; # Array of [ git, $oids] - foreach my $git (@{$self->{gits}}) { - my ($oid_full, $type, $size) = $git->check($oid_b); - if (defined($type) && (!$have_hints || $type eq 'blob')) { - return [ $git, $oid_full, $type, int($size) ]; - } + my ($oid_full, $type, $size) = $git->check($oid_b); - next if length($oid_b) == 40; + # other than {oid_b, try_gits, try_ibxs} + my $have_hints = scalar keys %$want > 3; + if (defined($type) && (!$have_hints || $type eq 'blob')) { + delete $want->{try_gits}; + return [ $git, $oid_full, $type, int($size) ]; # done, success + } - # parse stderr of "git cat-file --batch-check" - my $err = $git->last_check_err; - my (@oids) = ($err =~ /\b([a-f0-9]{40})\s+blob\b/g); - next unless scalar(@oids); + # TODO: deal with 40-char "abbreviations" with future SHA-256 git + return scalar(@$try) if length($oid_b) >= 40; - # TODO: do something with the ambiguous array? - # push @ambiguous, [ $git, @oids ]; + # parse stderr of "git cat-file --batch-check" + my $err = $git->last_check_err; + my (@oids) = ($err =~ /\b([a-f0-9]{40,})\s+blob\b/g); + return scalar(@$try) unless scalar(@oids); - dbg($self, "`$oid_b' ambiguous in " . - join("\n\t", $git->pub_urls($self->{psgi_env})) - . "\n" . - join('', map { "$_ blob\n" } @oids)); - } - scalar(@ambiguous) ? \@ambiguous : undef; + # TODO: do something with the ambiguous array? + # push @ambiguous, [ $git, @oids ]; + + dbg($self, "`$oid_b' ambiguous in " . + join("\n\t", $git->pub_urls($self->{psgi_env})) + . "\n" . + join('', map { "$_ blob\n" } @oids)); + scalar(@$try); } sub extract_diff ($$) { @@ -110,10 +119,6 @@ sub extract_diff ($$) { $s =~ s/\r\n/\n/sg; } - state $LF = qr!\r?\n!; - state $ANY = qr![^\r\n]+!; - state $MODE = '100644|120000|100755'; - state $FN = qr!(?:("?[^/\n]+/[^\r\n]+)|/dev/null)!; $s =~ m!( # $1 start header lines we save for debugging: @@ -174,8 +179,7 @@ sub extract_diff ($$) { # get rid of path-traversal attempts and junk patches: # it's junk at best, an attack attempt at worse: - state $bad_component = { map { $_ => 1 } ('', '.', '..') }; - foreach (@a, @b) { return if $bad_component->{$_} } + foreach (@a, @b) { return if $BAD_COMPONENT{$_} } $di->{path_a} = join('/', @a) if @a; $di->{path_b} = join('/', @b); @@ -228,12 +232,11 @@ sub find_extract_diffs ($$$) { } } - my $msgs = $srch->query($q, { relevance => 1 }); - + my $mset = $srch->mset($q, { relevance => 1 }); my $diffs = []; - foreach my $smsg (@$msgs) { - $ibx->smsg_mime($smsg) or next; - msg_iter(delete $smsg->{mime}, \&extract_diff, + for my $smsg (@{$srch->mset_to_smsg($ibx, $mset)}) { + my $eml = $ibx->smsg_eml($smsg) or next; + $eml->each_part(\&extract_diff, [$self, $diffs, $pre, $post, $ibx, $smsg], 1); } @$diffs ? $diffs : undef; @@ -270,7 +273,7 @@ sub prepare_index ($) { my $in = tmpfile("update-index.$oid_full") or die "tmpfile: $!"; print $in "$mode_a $oid_full\t$path_a\0" or die "print: $!"; $in->flush or die "flush: $!"; - sysseek($in, 0, 0) or die "seek: $!"; + sysseek($in, 0, SEEK_SET) or die "seek: $!"; dbg($self, 'preparing index'); my $rdr = { 0 => $in }; @@ -524,10 +527,12 @@ sub resolve_patch ($$) { # see if we can find the blob in an existing git repo: my $cur_want = $want->{oid_b}; - if ($self->{seen_oid}->{$cur_want}++) { + if (!$want->{try_ibxs} && $self->{seen_oid}->{$cur_want}++) { die "Loop detected solving $cur_want\n"; } - if (my $existing = solve_existing($self, $want)) { + $want->{try_ibxs} //= [ @{$self->{inboxes}} ]; # array copy + my $existing = solve_existing($self, $want); + if (ref $existing) { my ($found_git, undef, $type, undef) = @$existing; dbg($self, "found $cur_want in " . join(" ||\n\t", @@ -540,13 +545,17 @@ sub resolve_patch ($$) { } mark_found($self, $cur_want, $existing); return next_step($self); # onto patch application + } elsif ($existing > 0) { + push @{$self->{todo}}, $want; + return next_step($self); # retry solve_existing + } else { # $existing == 0: we may retry if inbox scan (below) fails + delete $want->{try_gits}; } # scan through inboxes to look for emails which results in # the oid we want: - foreach my $ibx (@{$self->{inboxes}}) { - my $diffs = find_extract_diffs($self, $ibx, $want) or next; - + my $ibx = shift(@{$want->{try_ibxs}}) or die 'BUG: {try_ibxs} empty'; + if (my $diffs = find_extract_diffs($self, $ibx, $want)) { unshift @{$self->{patches}}, @$diffs; dbg($self, "found $cur_want in ". join(" ||\n\t", map { di_url($self, $_) } @$diffs)); @@ -563,7 +572,14 @@ sub resolve_patch ($$) { } return next_step($self); # onto the next todo item } - if (length($cur_want) > $OID_MIN) { + + if (scalar @{$want->{try_ibxs}}) { # do we have more inboxes to try? + push @{$self->{todo}}, $want; + return next_step($self); + } + + if (length($cur_want) > $OID_MIN) { # maybe a shorter OID will work + delete $want->{try_ibxs}; # drop empty arrayref chop($cur_want); dbg($self, "retrying $want->{oid_b} as $cur_want"); $want->{oid_b} = $cur_want; @@ -593,7 +609,7 @@ sub new { } # recreate $oid_want using $hints -# hints keys: path_a, path_b, oid_a +# hints keys: path_a, path_b, oid_a (note: `oid_b' is NOT a hint) # Calls {user_cb} with: [ ::Git object, oid_full, type, size, di (diff_info) ] # with found object, or undef if nothing was found # Calls {user_cb} with a string error on fatal errors