-# Copyright (C) 2019 all contributors <meta@public-inbox.org>
+# Copyright (C) 2019-2020 all contributors <meta@public-inbox.org>
# License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
# "Solve" blobs which don't exist in git code repositories by
# searching inboxes for post-image blobs.
# this emits a lot of debugging/tracing information which may be
-# publically viewed over HTTP(S). Be careful not to expose
+# publicly viewed over HTTP(S). Be careful not to expose
# local filesystem layouts in the process.
package PublicInbox::SolverGit;
use strict;
-use warnings;
-use File::Temp 0.19 ();
+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);
# 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,
# oid_b => abbreviated post-image oid,
# tmp => anonymous file handle with the diff,
-# hdr_lines => arrayref of various header lines for mode information
+# 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)
# }
-# don't bother if somebody sends us a patch with these path components,
-# it's junk at best, an attack attempt at worse:
-my %bad_component = map { $_ => 1 } ('', '.', '..');
-
sub dbg ($$) {
print { $_[0]->{out} } $_[1], "\n" or ERR($_[0], "print(dbg): $!");
}
die $err;
}
-# look for existing objects already in git repos
+# look for existing objects already in git repos, returns arrayref
+# if found, number of remaining git coderepos to try if not.
sub solve_existing ($$) {
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 ($$) {
my ($p, $arg) = @_;
- my ($self, $diffs, $re, $ibx, $smsg) = @$arg;
+ my ($self, $want, $smsg) = @$arg;
my ($part) = @$p; # ignore $depth and @idx;
- my $hdr_lines; # diff --git a/... b/...
- my $tmp;
my $ct = $part->content_type || 'text/plain';
my ($s, undef) = msg_part_text($part, $ct);
defined $s or return;
- my $di = {};
+ my $post = $want->{oid_b};
+ my $pre = $want->{oid_a};
+ if (!defined($pre) || $pre !~ /\A[a-f0-9]+\z/) {
+ $pre = '[a-f0-9]{7}'; # for RE below
+ }
# Email::MIME::Encodings forces QP to be CRLF upon decoding,
# change it back to LF:
$s =~ s/\r\n/\n/sg;
}
- foreach my $l (split(/^/m, $s)) {
- if ($l =~ $re) {
- $di->{oid_a} = $1;
- $di->{oid_b} = $2;
- if (defined($3)) {
- my $mode_a = $3;
- if ($mode_a =~ /\A(?:100644|120000|100755)\z/) {
- $di->{mode_a} = $mode_a;
- }
- }
-
-
- # start writing the diff out to a tempfile
- my $path = ++$self->{tot};
- $di->{n} = $path;
- open($tmp, '>', $self->{tmp}->dirname . "/$path") or
- die "open(tmp): $!";
-
- push @$hdr_lines, $l;
- $di->{hdr_lines} = $hdr_lines;
- utf8::encode($_) for @$hdr_lines;
- print $tmp @$hdr_lines or die "print(tmp): $!";
-
- # for debugging/diagnostics:
- $di->{ibx} = $ibx;
- $di->{smsg} = $smsg;
- } elsif ($l =~ m!\Adiff --git ("?[^/]+/.+) ("?[^/]+/.+)$!) {
- last if $tmp; # got our blob, done!
-
- my ($path_a, $path_b) = ($1, $2);
-
- # diff header lines won't have \r because git
- # will quote them, but Email::MIME gives CRLF
- # for quoted-printable:
- $path_b =~ tr/\r//d;
- # don't care for leading 'a/' and 'b/'
- my (undef, @a) = split(m{/}, git_unquote($path_a));
- my (undef, @b) = split(m{/}, git_unquote($path_b));
+ $s =~ m!( # $1 start header lines we save for debugging:
+
+ # everything before ^index is optional, but we don't
+ # want to match ^(old|copy|rename|deleted|...) unless
+ # we match /^diff --git/ first:
+ (?: # begin optional stuff:
+
+ # try to get the pre-and-post filenames as $2 and $3
+ (?:^diff\x20--git\x20$FN\x20$FN$LF)
+
+ (?:^(?: # pass all this to git-apply:
+ # old mode $4
+ (?:old\x20mode\x20($MODE))
+ |
+ # new mode (possibly new file) ($5)
+ (?:new\x20(?:file\x20)?mode\x20($MODE))
+ |
+ (?:(?:copy|rename|deleted|
+ dissimilarity|similarity)$ANY)
+ )$LF)*
+
+ )? # end of optional stuff, everything below is required
+
+ # match the pre and post-image OIDs as $6 $7
+ ^index\x20(${pre}[a-f0-9]*)\.\.(${post}[a-f0-9]*)
+ # mode if unchanged $8
+ (?:\x20(100644|120000|100755))?$LF
+ ) # end of header lines ($1)
+ ( # $9 is the patch body
+ # "--- a/foo.c" sets pre-filename ($10) in case
+ # $2 is missing
+ (?:^---\x20$FN$LF)
+
+ # "+++ b/foo.c" sets post-filename ($11) in case
+ # $3 is missing
+ (?:^\+{3}\x20$FN$LF)
+
+ # the meat of the diff, including "^\\No newline ..."
+ # We also allow for totally blank lines w/o leading spaces,
+ # because git-apply(1) handles that case, too
+ (?:^(?:[\@\+\x20\-\\][^\n]*|)$LF)+
+ )!smx or return;
+
+ my $di = {
+ hdr_lines => $1,
+ oid_a => $6,
+ oid_b => $7,
+ mode_a => $5 // $8 // $4, # new (file) // unchanged // old
+ };
+ my $path_a = $2 // $10;
+ my $path_b = $3 // $11;
+ my $patch = $9;
+
+ # don't care for leading 'a/' and 'b/'
+ my (undef, @a) = split(m{/}, git_unquote($path_a)) if defined($path_a);
+ my (undef, @b) = split(m{/}, git_unquote($path_b));
+
+ # get rid of path-traversal attempts and junk patches:
+ # it's junk at best, an attack attempt at worse:
+ foreach (@a, @b) { return if $BAD_COMPONENT{$_} }
+
+ $di->{path_a} = join('/', @a) if @a;
+ $di->{path_b} = join('/', @b);
+
+ my $path = ++$self->{tot};
+ $di->{n} = $path;
+ open(my $tmp, '>:utf8', $self->{tmp}->dirname . "/$path") or
+ die "open(tmp): $!";
+ print $tmp $di->{hdr_lines}, $patch or die "print(tmp): $!";
+ close $tmp or die "close(tmp): $!";
- # get rid of path-traversal attempts and junk patches:
- foreach (@a, @b) {
- return if $bad_component{$_};
- }
+ # for debugging/diagnostics:
+ $di->{ibx} = $want->{cur_ibx};
+ $di->{smsg} = $smsg;
- $di->{path_a} = join('/', @a);
- $di->{path_b} = join('/', @b);
- $hdr_lines = [ $l ];
- } elsif ($tmp) {
- utf8::encode($l);
- print $tmp $l or die "print(tmp): $!";
- } elsif ($hdr_lines) {
- push @$hdr_lines, $l;
- if ($l =~ /\Anew file mode (100644|120000|100755)$/) {
- $di->{mode_a} = $1;
- }
- }
- }
- return undef unless $tmp;
- close $tmp or die "close(tmp): $!";
- push @$diffs, $di;
+ push @{$self->{tmp_diffs}}, $di;
}
sub path_searchable ($) { defined($_[0]) && $_[0] =~ m!\A[\w/\. \-]+\z! }
join('', map { qq( dfn:"$_") } split(/\.\./, $_[0]));
}
-sub find_extract_diffs ($$$) {
+sub find_smsgs ($$$) {
my ($self, $ibx, $want) = @_;
my $srch = $ibx->search or return;
my $pre = $want->{oid_a};
if (defined $pre && $pre =~ /\A[a-f0-9]+\z/) {
$q .= " dfpre:$pre";
- } else {
- $pre = '[a-f0-9]{7}'; # for $re below
}
my $path_b = $want->{path_b};
$q .= filename_query($path_a);
}
}
-
- my $msgs = $srch->query($q, { relevance => 1 });
- my $re = qr/\Aindex ($pre[a-f0-9]*)\.\.($post[a-f0-9]*)(?: ([0-9]+))?/;
- my $diffs = [];
- foreach my $smsg (@$msgs) {
- $ibx->smsg_mime($smsg) or next;
- my $mime = delete $smsg->{mime};
- msg_iter($mime, \&extract_diff,
- [$self, $diffs, $re, $ibx, $smsg]);
- }
- @$diffs ? $diffs : undef;
+ my $mset = $srch->mset($q, { relevance => 1 });
+ $mset->size ? $srch->mset_to_smsg($ibx, $mset) : undef;
}
sub update_index_result ($$) {
# no index creation for added files
$oid_a =~ /\A0+\z/ and return next_step($self);
- die "BUG: $oid_a not not found" unless $existing;
+ die "BUG: $oid_a not found" unless $existing;
my $oid_full = $existing->[1];
my $path_a = $di->{path_a} or die "BUG: path_a missing for $oid_full";
- my $mode_a = $di->{mode_a} || extract_old_mode($di);
+ my $mode_a = $di->{mode_a} // '100644';
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 => $in };
prepare_index($self);
}
-sub extract_old_mode ($) {
- my ($di) = @_;
- if (join('', @{$di->{hdr_lines}}) =~
- /^old mode (100644|100755|120000)\b/) {
- return $1;
- }
- '100644';
-}
-
sub do_finish ($) {
my ($self) = @_;
my ($found, $oid_want) = @$self{qw(found oid_want)};
if ($nxt && oids_same_ish($nxt->{oid_b}, $di->{oid_b})) {
dbg($self, $msg);
dbg($self, 'trying '.di_url($self, $nxt));
+ return do_git_apply($self);
} else {
ERR($self, $msg);
}
my $i = ++$self->{nr};
$di = shift @$patches;
dbg($self, "\napplying [$i/$total] " . di_url($self, $di) .
- "\n" . join('', @{$di->{hdr_lines}}));
+ "\n" . $di->{hdr_lines});
my $path = $di->{n};
$len += length($path) + 1;
push @cmd, $path;
} while (@$patches && $len < $ARG_SIZE_MAX &&
!oids_same_ish($patches->[0]->{oid_b}, $prv_oid_b));
- my $opt = { 2 => 1, -C => $dn };
+ my $opt = { 2 => 1, -C => $dn, quiet => 1 };
my $qsp = PublicInbox::Qspawn->new(\@cmd, $self->{git_env}, $opt);
$self->{-cur_di} = $di;
$self->{-qsp} = $qsp;
defined($url) ? "$url$mid/" : "<$mid>";
}
+sub retry_current {
+ # my ($self, $want) = @_;
+ push @{$_[0]->{todo}}, $_[1];
+ goto \&next_step # retry solve_existing
+}
+
+sub try_harder {
+ my ($self, $want) = @_;
+
+ # do we have more inboxes to try?
+ goto \&retry_current if scalar @{$want->{try_ibxs}};
+
+ my $cur_want = $want->{oid_b};
+ if (length($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} = $cur_want;
+ goto \&retry_current; # retry with shorter abbrev
+ }
+
+ dbg($self, "could not find $cur_want");
+ eval { done($self, undef) };
+ die "E: $@" if $@;
+}
+
sub resolve_patch ($$) {
my ($self, $want) = @_;
+ my $cur_want = $want->{oid_b};
if (scalar(@{$self->{patches}}) > $MAX_PATCH) {
die "Aborting, too many steps to $self->{oid_want}";
}
+ if (my $msgs = $want->{try_smsgs}) {
+ my $smsg = shift @$msgs;
+ if (my $eml = $want->{cur_ibx}->smsg_eml($smsg)) {
+ $eml->each_part(\&extract_diff,
+ [ $self, $want, $smsg ], 1);
+ }
+
+ # try the remaining smsgs later
+ goto \&retry_current if scalar @$msgs;
+
+ delete $want->{try_smsgs};
+ delete $want->{cur_ibx};
+
+ my $diffs = delete $self->{tmp_diffs};
+ if (scalar @$diffs) {
+ unshift @{$self->{patches}}, @$diffs;
+ dbg($self, "found $cur_want in " . join(" ||\n\t",
+ map { di_url($self, $_) } @$diffs));
+
+ # good, we can find a path to the oid we $want, now
+ # lets see if we need to apply more patches:
+ my $di = $diffs->[0];
+ my $src = $di->{oid_a};
+
+ unless ($src =~ /\A0+\z/) {
+ # we have to solve it using another oid, fine:
+ my $job = {
+ oid_b => $src,
+ path_b => $di->{path_a},
+ };
+ push @{$self->{todo}}, $job;
+ }
+ goto \&next_step; # onto the next todo item
+ }
+ goto \&try_harder;
+ }
+
# 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", $found_git->pub_urls($self->{psgi_env})));
+ join(" ||\n\t",
+ $found_git->pub_urls($self->{psgi_env})));
if ($cur_want eq $self->{oid_want} || $type ne 'blob') {
eval { done($self, $existing) };
return;
}
mark_found($self, $cur_want, $existing);
- return next_step($self); # onto patch application
+ goto \&next_step; # onto patch application
+ } elsif ($existing > 0) {
+ goto \&retry_current;
+ } 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, $want) or next;
-
- unshift @{$self->{patches}}, @$diffs;
- dbg($self, "found $cur_want in ".
- join("\n\t", map { di_url($self, $_) } @$diffs));
-
- # good, we can find a path to the oid we $want, now
- # lets see if we need to apply more patches:
- my $di = $diffs->[0];
- my $src = $di->{oid_a};
-
- unless ($src =~ /\A0+\z/) {
- # we have to solve it using another oid, fine:
- my $job = { oid_b => $src, path_b => $di->{path_a} };
- push @{$self->{todo}}, $job;
- }
- return next_step($self); # onto the next todo item
+ my $ibx = shift(@{$want->{try_ibxs}}) or die 'BUG: {try_ibxs} empty';
+ if (my $msgs = find_smsgs($self, $ibx, $want)) {
+ $want->{try_smsgs} = $msgs;
+ $want->{cur_ibx} = $ibx;
+ $self->{tmp_diffs} = [];
+ goto \&retry_current;
}
- if (length($cur_want) > $OID_MIN) {
- chop($cur_want);
- dbg($self, "retrying $want->{oid_b} as $cur_want");
- $want->{oid_b} = $cur_want;
- push @{$self->{todo}}, $want;
- return next_step($self); # retry with shorter abbrev
- }
-
- dbg($self, "could not find $cur_want");
- eval { done($self, undef) };
- die "E: $@" if $@;
+ goto \&try_harder;
}
# this API is designed to avoid creating self-referential structures;
}
# 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