1 # Copyright (C) 2019-2020 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # "Solve" blobs which don't exist in git code repositories by
5 # searching inboxes for post-image blobs.
7 # this emits a lot of debugging/tracing information which may be
8 # publicly viewed over HTTP(S). Be careful not to expose
9 # local filesystem layouts in the process.
10 package PublicInbox::SolverGit;
13 use File::Temp 0.19 (); # 0.19 for ->newdir
14 use Fcntl qw(SEEK_SET);
15 use PublicInbox::Git qw(git_unquote git_quote);
16 use PublicInbox::MsgIter qw(msg_part_text);
17 use PublicInbox::Qspawn;
18 use PublicInbox::Tmpfile;
19 use URI::Escape qw(uri_escape_utf8);
21 # POSIX requires _POSIX_ARG_MAX >= 4096, and xargs is required to
22 # subtract 2048 bytes. We also don't factor in environment variable
24 use POSIX qw(sysconf _SC_ARG_MAX);
25 my $ARG_SIZE_MAX = (sysconf(_SC_ARG_MAX) || 4096) - 2048;
28 # By default, "git format-patch" generates filenames with a four-digit
29 # prefix, so that means 9999 patch series are OK, right? :>
30 # Maybe we can make this configurable, main concern is disk space overhead
31 # for uncompressed patch fragments. Aside from space, public-inbox-httpd
32 # is otherwise unaffected by having many patches, here, as it can share
33 # work fairly. Other PSGI servers may have trouble, though.
37 my $ANY = qr![^\r\n]+!;
38 my $MODE = '100644|120000|100755';
39 my $FN = qr!(?:("?[^/\n]+/[^\r\n]+)|/dev/null)!;
40 my %BAD_COMPONENT = ('' => 1, '.' => 1, '..' => 1);
42 # di = diff info / a hashref with information about a diff ($di):
44 # oid_a => abbreviated pre-image oid,
45 # oid_b => abbreviated post-image oid,
46 # tmp => anonymous file handle with the diff,
47 # hdr_lines => string of various header lines for mode information
48 # mode_a => original mode of oid_a (string, not integer),
49 # ibx => PublicInbox::Inbox object containing the diff
50 # smsg => PublicInbox::Smsg object containing diff
51 # path_a => pre-image path
52 # path_b => post-image path
53 # n => numeric path of the patch (relative to worktree)
57 print { $_[0]->{out} } $_[1], "\n" or ERR($_[0], "print(dbg): $!");
61 my ($self, $res) = @_;
62 my $ucb = delete($self->{user_cb}) or return;
63 $ucb->($res, $self->{uarg});
67 my ($self, $err) = @_;
68 print { $self->{out} } $err, "\n";
69 eval { done($self, $err) };
73 # look for existing objects already in git repos, returns arrayref
74 # if found, number of remaining git coderepos to try if not.
75 sub solve_existing ($$) {
76 my ($self, $want) = @_;
77 my $try = $want->{try_gits} //= [ @{$self->{gits}} ]; # array copy
78 my $git = shift @$try or die 'BUG {try_gits} empty';
79 my $oid_b = $want->{oid_b};
80 my ($oid_full, $type, $size) = $git->check($oid_b);
82 # other than {oid_b, try_gits, try_ibxs}
83 my $have_hints = scalar keys %$want > 3;
84 if (defined($type) && (!$have_hints || $type eq 'blob')) {
85 delete $want->{try_gits};
86 return [ $git, $oid_full, $type, int($size) ]; # done, success
89 # TODO: deal with 40-char "abbreviations" with future SHA-256 git
90 return scalar(@$try) if length($oid_b) >= 40;
92 # parse stderr of "git cat-file --batch-check"
93 my $err = $git->last_check_err;
94 my (@oids) = ($err =~ /\b([a-f0-9]{40,})\s+blob\b/g);
95 return scalar(@$try) unless scalar(@oids);
97 # TODO: do something with the ambiguous array?
98 # push @ambiguous, [ $git, @oids ];
100 dbg($self, "`$oid_b' ambiguous in " .
101 join("\n\t", $git->pub_urls($self->{psgi_env}))
103 join('', map { "$_ blob\n" } @oids));
107 sub extract_diff ($$) {
109 my ($self, $diffs, $pre, $post, $ibx, $smsg) = @$arg;
110 my ($part) = @$p; # ignore $depth and @idx;
111 my $ct = $part->content_type || 'text/plain';
112 my ($s, undef) = msg_part_text($part, $ct);
113 defined $s or return;
115 # Email::MIME::Encodings forces QP to be CRLF upon decoding,
116 # change it back to LF:
117 my $cte = $part->header('Content-Transfer-Encoding') || '';
118 if ($cte =~ /\bquoted-printable\b/i && $part->crlf eq "\n") {
123 $s =~ m!( # $1 start header lines we save for debugging:
125 # everything before ^index is optional, but we don't
126 # want to match ^(old|copy|rename|deleted|...) unless
127 # we match /^diff --git/ first:
128 (?: # begin optional stuff:
130 # try to get the pre-and-post filenames as $2 and $3
131 (?:^diff\x20--git\x20$FN\x20$FN$LF)
133 (?:^(?: # pass all this to git-apply:
135 (?:old\x20mode\x20($MODE))
137 # new mode (possibly new file) ($5)
138 (?:new\x20(?:file\x20)?mode\x20($MODE))
140 (?:(?:copy|rename|deleted|
141 dissimilarity|similarity)$ANY)
144 )? # end of optional stuff, everything below is required
146 # match the pre and post-image OIDs as $6 $7
147 ^index\x20(${pre}[a-f0-9]*)\.\.(${post}[a-f0-9]*)
148 # mode if unchanged $8
149 (?:\x20(100644|120000|100755))?$LF
150 ) # end of header lines ($1)
151 ( # $9 is the patch body
152 # "--- a/foo.c" sets pre-filename ($10) in case
156 # "+++ b/foo.c" sets post-filename ($11) in case
160 # the meat of the diff, including "^\\No newline ..."
161 # We also allow for totally blank lines w/o leading spaces,
162 # because git-apply(1) handles that case, too
163 (?:^(?:[\@\+\x20\-\\][^\n]*|)$LF)+
170 mode_a => $5 // $8 // $4, # new (file) // unchanged // old
172 my $path_a = $2 // $10;
173 my $path_b = $3 // $11;
176 # don't care for leading 'a/' and 'b/'
177 my (undef, @a) = split(m{/}, git_unquote($path_a)) if defined($path_a);
178 my (undef, @b) = split(m{/}, git_unquote($path_b));
180 # get rid of path-traversal attempts and junk patches:
181 # it's junk at best, an attack attempt at worse:
182 foreach (@a, @b) { return if $BAD_COMPONENT{$_} }
184 $di->{path_a} = join('/', @a) if @a;
185 $di->{path_b} = join('/', @b);
187 my $path = ++$self->{tot};
189 open(my $tmp, '>:utf8', $self->{tmp}->dirname . "/$path") or
191 print $tmp $di->{hdr_lines}, $patch or die "print(tmp): $!";
192 close $tmp or die "close(tmp): $!";
194 # for debugging/diagnostics:
201 sub path_searchable ($) { defined($_[0]) && $_[0] =~ m!\A[\w/\. \-]+\z! }
203 # ".." appears in path names, which confuses Xapian into treating
204 # it as a range query. So we split on ".." since Xapian breaks
205 # on punctuation anyways:
206 sub filename_query ($) {
207 join('', map { qq( dfn:"$_") } split(/\.\./, $_[0]));
210 sub find_extract_diffs ($$$) {
211 my ($self, $ibx, $want) = @_;
212 my $srch = $ibx->search or return;
214 my $post = $want->{oid_b} or die 'BUG: no {oid_b}';
215 $post =~ /\A[a-f0-9]+\z/ or die "BUG: oid_b not hex: $post";
217 my $q = "dfpost:$post";
218 my $pre = $want->{oid_a};
219 if (defined $pre && $pre =~ /\A[a-f0-9]+\z/) {
222 $pre = '[a-f0-9]{7}'; # for $re below
225 my $path_b = $want->{path_b};
226 if (path_searchable($path_b)) {
227 $q .= filename_query($path_b);
229 my $path_a = $want->{path_a};
230 if (path_searchable($path_a) && $path_a ne $path_b) {
231 $q .= filename_query($path_a);
235 my $mset = $srch->mset($q, { relevance => 1 });
237 for my $smsg (@{$srch->mset_to_smsg($ibx, $mset)}) {
238 my $eml = $ibx->smsg_eml($smsg) or next;
239 $eml->each_part(\&extract_diff,
240 [$self, $diffs, $pre, $post, $ibx, $smsg], 1);
242 @$diffs ? $diffs : undef;
245 sub update_index_result ($$) {
246 my ($bref, $self) = @_;
247 my ($qsp, $msg) = delete @$self{qw(-qsp -msg)};
248 if (my $err = $qsp->{err}) {
249 ERR($self, "git update-index error: $err");
252 next_step($self); # onto do_git_apply
255 sub prepare_index ($) {
257 my $patches = $self->{patches};
260 my $di = $patches->[0] or die 'no patches';
261 my $oid_a = $di->{oid_a} or die '{oid_a} unset';
262 my $existing = $self->{found}->{$oid_a};
264 # no index creation for added files
265 $oid_a =~ /\A0+\z/ and return next_step($self);
267 die "BUG: $oid_a not not found" unless $existing;
269 my $oid_full = $existing->[1];
270 my $path_a = $di->{path_a} or die "BUG: path_a missing for $oid_full";
271 my $mode_a = $di->{mode_a} // '100644';
273 my $in = tmpfile("update-index.$oid_full") or die "tmpfile: $!";
274 print $in "$mode_a $oid_full\t$path_a\0" or die "print: $!";
275 $in->flush or die "flush: $!";
276 sysseek($in, 0, SEEK_SET) or die "seek: $!";
278 dbg($self, 'preparing index');
279 my $rdr = { 0 => $in };
280 my $cmd = [ qw(git update-index -z --index-info) ];
281 my $qsp = PublicInbox::Qspawn->new($cmd, $self->{git_env}, $rdr);
282 $path_a = git_quote($path_a);
283 $self->{-qsp} = $qsp;
284 $self->{-msg} = "index prepared:\n$mode_a $oid_full\t$path_a";
285 $qsp->psgi_qx($self->{psgi_env}, undef, \&update_index_result, $self);
288 # pure Perl "git init"
289 sub do_git_init ($) {
291 my $dir = $self->{tmp}->dirname;
292 my $git_dir = "$dir/git";
294 foreach ('', qw(objects refs objects/info refs/heads)) {
295 mkdir("$git_dir/$_") or die "mkdir $_: $!";
297 open my $fh, '>', "$git_dir/config" or die "open git/config: $!";
298 print $fh <<'EOF' or die "print git/config $!";
300 repositoryFormatVersion = 0
303 fsyncObjectfiles = false
304 logAllRefUpdates = false
306 close $fh or die "close git/config: $!";
308 open $fh, '>', "$git_dir/HEAD" or die "open git/HEAD: $!";
309 print $fh "ref: refs/heads/master\n" or die "print git/HEAD: $!";
310 close $fh or die "close git/HEAD: $!";
312 my $f = 'objects/info/alternates';
313 open $fh, '>', "$git_dir/$f" or die "open: $f: $!";
314 foreach my $git (@{$self->{gits}}) {
315 print $fh $git->git_path('objects'),"\n" or die "print $f: $!";
317 close $fh or die "close: $f: $!";
318 my $tmp_git = $self->{tmp_git} = PublicInbox::Git->new($git_dir);
319 $tmp_git->{-tmp} = $self->{tmp};
322 GIT_INDEX_FILE => "$git_dir/index",
324 prepare_index($self);
329 my ($found, $oid_want) = @$self{qw(found oid_want)};
330 if (my $exists = $found->{$oid_want}) {
331 return done($self, $exists);
334 # let git disambiguate if oid_want was too short,
335 # but long enough to be unambiguous:
336 my $tmp_git = $self->{tmp_git};
337 if (my @res = $tmp_git->check($oid_want)) {
338 return done($self, $found->{$res[0]});
340 if (my $err = $tmp_git->last_check_err) {
349 # step 1: resolve blobs to patches in the todo queue
350 if (my $want = pop @{$self->{todo}}) {
351 # this populates {patches} and {todo}
352 resolve_patch($self, $want);
354 # step 2: then we instantiate a working tree once
355 # the todo queue is finally empty:
356 } elsif (!defined($self->{tmp_git})) {
359 # step 3: apply each patch in the stack
360 } elsif (scalar @{$self->{patches}}) {
363 # step 4: execute the user-supplied callback with
364 # our result: (which may be undef)
365 # Other steps may call user_cb to terminate prematurely
367 } elsif (exists $self->{user_cb}) {
370 die 'about to call user_cb twice'; # Oops :x
375 $err =~ s/^\s*Exception:\s*//; # bad word to show users :P
376 dbg($self, "E: $err");
377 eval { done($self, $err) };
383 # if outside of public-inbox-httpd, caller is expected to be
384 # looping event_step, anyways
385 my $async = $self->{psgi_env}->{'pi-httpd.async'} or return;
386 # PublicInbox::HTTPD::Async->new
387 $async->(undef, undef, $self);
390 sub mark_found ($$$) {
391 my ($self, $oid, $found_info) = @_;
392 my $found = $self->{found};
393 $found->{$oid} = $found_info;
394 my $oid_cur = $found_info->[1];
395 while ($oid_cur ne $oid && length($oid_cur) > $OID_MIN) {
396 $found->{$oid_cur} = $found_info;
401 sub parse_ls_files ($$) {
402 my ($self, $bref) = @_;
403 my ($qsp, $di) = delete @$self{qw(-qsp -cur_di)};
404 if (my $err = $qsp->{err}) {
405 die "git ls-files error: $err";
408 my ($line, @extra) = split(/\0/, $$bref);
409 scalar(@extra) and die "BUG: extra files in index: <",
410 join('> <', @extra), ">";
412 my ($info, $file) = split(/\t/, $line, 2);
413 my ($mode_b, $oid_b_full, $stage) = split(/ /, $info);
414 if ($file ne $di->{path_b}) {
416 "BUG: index mismatch: file=$file != path_b=$di->{path_b}";
419 my $tmp_git = $self->{tmp_git} or die 'no git working tree';
420 my (undef, undef, $size) = $tmp_git->check($oid_b_full);
421 defined($size) or die "check $oid_b_full failed";
423 dbg($self, "index at:\n$mode_b $oid_b_full\t$file");
424 my $created = [ $tmp_git, $oid_b_full, 'blob', $size, $di ];
425 mark_found($self, $di->{oid_b}, $created);
426 next_step($self); # onto the next patch
429 sub ls_files_result {
430 my ($bref, $self) = @_;
431 eval { parse_ls_files($self, $bref) };
432 ERR($self, $@) if $@;
435 sub oids_same_ish ($$) {
436 (index($_[0], $_[1]) == 0) || (index($_[1], $_[0]) == 0);
439 sub skip_identical ($$$) {
440 my ($self, $patches, $cur_oid_b) = @_;
441 while (my $nxt = $patches->[0]) {
442 if (oids_same_ish($cur_oid_b, $nxt->{oid_b})) {
443 dbg($self, 'skipping '.di_url($self, $nxt).
452 sub apply_result ($$) {
453 my ($bref, $self) = @_;
454 my ($qsp, $di) = delete @$self{qw(-qsp -cur_di)};
456 my $patches = $self->{patches};
457 if (my $err = $qsp->{err}) {
458 my $msg = "git apply error: $err";
459 my $nxt = $patches->[0];
460 if ($nxt && oids_same_ish($nxt->{oid_b}, $di->{oid_b})) {
462 dbg($self, 'trying '.di_url($self, $nxt));
463 return do_git_apply($self);
468 skip_identical($self, $patches, $di->{oid_b});
471 my @cmd = qw(git ls-files -s -z);
472 $qsp = PublicInbox::Qspawn->new(\@cmd, $self->{git_env});
473 $self->{-cur_di} = $di;
474 $self->{-qsp} = $qsp;
475 $qsp->psgi_qx($self->{psgi_env}, undef, \&ls_files_result, $self);
478 sub do_git_apply ($) {
480 my $dn = $self->{tmp}->dirname;
481 my $patches = $self->{patches};
483 # we need --ignore-whitespace because some patches are CRLF
484 my @cmd = (qw(git apply --cached --ignore-whitespace
485 --unidiff-zero --whitespace=warn --verbose));
486 my $len = length(join(' ', @cmd));
487 my $total = $self->{tot};
488 my $di; # keep track of the last one for "git ls-files"
492 my $i = ++$self->{nr};
493 $di = shift @$patches;
494 dbg($self, "\napplying [$i/$total] " . di_url($self, $di) .
495 "\n" . $di->{hdr_lines});
497 $len += length($path) + 1;
499 $prv_oid_b = $di->{oid_b};
500 } while (@$patches && $len < $ARG_SIZE_MAX &&
501 !oids_same_ish($patches->[0]->{oid_b}, $prv_oid_b));
503 my $opt = { 2 => 1, -C => $dn, quiet => 1 };
504 my $qsp = PublicInbox::Qspawn->new(\@cmd, $self->{git_env}, $opt);
505 $self->{-cur_di} = $di;
506 $self->{-qsp} = $qsp;
507 $qsp->psgi_qx($self->{psgi_env}, undef, \&apply_result, $self);
511 my ($self, $di) = @_;
512 # note: we don't pass the PSGI env unconditionally, here,
513 # different inboxes can have different HTTP_HOST on the same instance.
514 my $ibx = $di->{ibx};
515 my $env = $self->{psgi_env} if $ibx eq $self->{inboxes}->[0];
516 my $url = $ibx->base_url($env);
517 my $mid = $di->{smsg}->{mid};
518 defined($url) ? "$url$mid/" : "<$mid>";
521 sub resolve_patch ($$) {
522 my ($self, $want) = @_;
524 if (scalar(@{$self->{patches}}) > $MAX_PATCH) {
525 die "Aborting, too many steps to $self->{oid_want}";
528 # see if we can find the blob in an existing git repo:
529 my $cur_want = $want->{oid_b};
530 if (!$want->{try_ibxs} && $self->{seen_oid}->{$cur_want}++) {
531 die "Loop detected solving $cur_want\n";
533 $want->{try_ibxs} //= [ @{$self->{inboxes}} ]; # array copy
534 my $existing = solve_existing($self, $want);
536 my ($found_git, undef, $type, undef) = @$existing;
537 dbg($self, "found $cur_want in " .
539 $found_git->pub_urls($self->{psgi_env})));
541 if ($cur_want eq $self->{oid_want} || $type ne 'blob') {
542 eval { done($self, $existing) };
546 mark_found($self, $cur_want, $existing);
547 return next_step($self); # onto patch application
548 } elsif ($existing > 0) {
549 push @{$self->{todo}}, $want;
550 return next_step($self); # retry solve_existing
551 } else { # $existing == 0: we may retry if inbox scan (below) fails
552 delete $want->{try_gits};
555 # scan through inboxes to look for emails which results in
557 my $ibx = shift(@{$want->{try_ibxs}}) or die 'BUG: {try_ibxs} empty';
558 if (my $diffs = find_extract_diffs($self, $ibx, $want)) {
559 unshift @{$self->{patches}}, @$diffs;
560 dbg($self, "found $cur_want in ".
561 join(" ||\n\t", map { di_url($self, $_) } @$diffs));
563 # good, we can find a path to the oid we $want, now
564 # lets see if we need to apply more patches:
565 my $di = $diffs->[0];
566 my $src = $di->{oid_a};
568 unless ($src =~ /\A0+\z/) {
569 # we have to solve it using another oid, fine:
570 my $job = { oid_b => $src, path_b => $di->{path_a} };
571 push @{$self->{todo}}, $job;
573 return next_step($self); # onto the next todo item
576 if (scalar @{$want->{try_ibxs}}) { # do we have more inboxes to try?
577 push @{$self->{todo}}, $want;
578 return next_step($self);
581 if (length($cur_want) > $OID_MIN) { # maybe a shorter OID will work
582 delete $want->{try_ibxs}; # drop empty arrayref
584 dbg($self, "retrying $want->{oid_b} as $cur_want");
585 $want->{oid_b} = $cur_want;
586 push @{$self->{todo}}, $want;
587 return next_step($self); # retry with shorter abbrev
590 dbg($self, "could not find $cur_want");
591 eval { done($self, undef) };
595 # this API is designed to avoid creating self-referential structures;
596 # so user_cb never references the SolverGit object
598 my ($class, $ibx, $user_cb, $uarg) = @_;
601 gits => $ibx->{-repo_objs},
604 # -cur_di, -qsp, -msg => temporary fields for Qspawn callbacks
606 # TODO: config option for searching related inboxes
611 # recreate $oid_want using $hints
612 # hints keys: path_a, path_b, oid_a (note: `oid_b' is NOT a hint)
613 # Calls {user_cb} with: [ ::Git object, oid_full, type, size, di (diff_info) ]
614 # with found object, or undef if nothing was found
615 # Calls {user_cb} with a string error on fatal errors
617 my ($self, $env, $out, $oid_want, $hints) = @_;
619 # should we even get here? Probably not, but somebody
620 # could be manually typing URLs:
621 return done($self, undef) if $oid_want =~ /\A0+\z/;
623 $self->{oid_want} = $oid_want;
625 $self->{seen_oid} = {};
627 $self->{psgi_env} = $env;
628 $self->{todo} = [ { %$hints, oid_b => $oid_want } ];
629 $self->{patches} = []; # [ $di, $di, ... ]
630 $self->{found} = {}; # { abbr => [ ::Git, oid, type, size, $di ] }
631 $self->{tmp} = File::Temp->newdir("solver.$oid_want-XXXXXXXX", TMPDIR => 1);
633 dbg($self, "solving $oid_want ...");
634 if (my $async = $env->{'pi-httpd.async'}) {
635 # PublicInbox::HTTPD::Async->new
636 $async->(undef, undef, $self);
638 event_step($self) while $self->{user_cb};