about summary refs log tree commit homepage
path: root/lib
diff options
context:
space:
mode:
Diffstat (limited to 'lib')
-rw-r--r--lib/PublicInbox/Git.pm16
-rw-r--r--lib/PublicInbox/SolverGit.pm400
2 files changed, 416 insertions, 0 deletions
diff --git a/lib/PublicInbox/Git.pm b/lib/PublicInbox/Git.pm
index 90b9214a..9676086f 100644
--- a/lib/PublicInbox/Git.pm
+++ b/lib/PublicInbox/Git.pm
@@ -40,6 +40,7 @@ sub new {
         my ($class, $git_dir) = @_;
         my @st;
         $st[7] = $st[10] = 0;
+        # may contain {-wt} field (working-tree (File::Temp::Dir))
         bless { git_dir => $git_dir, st => \@st }, $class
 }
 
@@ -201,6 +202,21 @@ sub packed_bytes {
 
 sub DESTROY { cleanup(@_) }
 
+# show the blob URL for cgit/gitweb/whatever
+sub src_blob_url {
+        my ($self, $oid) = @_;
+        # blob_fmt = "https://example.com/foo.git/blob/%s"
+        if (my $bfu = $self->{blob_fmt_url}) {
+                return sprintf($bfu, $oid);
+        }
+
+        # don't show full FS path, basename should be OK:
+        if ($self->{git_dir} =~ m!/([^/]+)\z!) {
+                return "/path/to/$1";
+        }
+        '???';
+}
+
 1;
 __END__
 =pod
diff --git a/lib/PublicInbox/SolverGit.pm b/lib/PublicInbox/SolverGit.pm
new file mode 100644
index 00000000..f28768a9
--- /dev/null
+++ b/lib/PublicInbox/SolverGit.pm
@@ -0,0 +1,400 @@
+# Copyright (C) 2019 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
+# local filesystem layouts in the process.
+package PublicInbox::SolverGit;
+use strict;
+use warnings;
+use File::Temp qw();
+use Fcntl qw(SEEK_SET);
+use File::Path qw(make_path);
+use PublicInbox::Git qw(git_unquote);
+use PublicInbox::Spawn qw(spawn popen_rd);
+use PublicInbox::MsgIter qw(msg_iter msg_part_text);
+use URI::Escape qw(uri_escape_utf8);
+
+# 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 { $_ => 1 } ('', '.', '..');
+
+sub new {
+        my ($class, $gits, $inboxes) = @_;
+        bless {
+                gits => $gits,
+                inboxes => $inboxes,
+        }, $class;
+}
+
+# look for existing blobs already in git repos
+sub solve_existing ($$) {
+        my ($self, $want) = @_;
+        foreach my $git (@{$self->{gits}}) {
+                my ($oid_full, $type, $size) = $git->check($want->{oid_b});
+                if (defined($type) && $type eq 'blob') {
+                        return [ $git, $oid_full, $type, int($size) ];
+                }
+        }
+        undef;
+}
+
+# returns a hashref with information about a diff:
+# {
+#        oid_a => abbreviated pre-image oid,
+#        oid_b => abbreviated post-image oid,
+#        tmp => anonymous file handle with the diff,
+#        hdr_lines => arrayref of various header lines for mode information
+#        mode_a => original mode of oid_a (string, not integer),
+#        ibx => PublicInbox::Inbox object containing the diff
+#        smsg => PublicInbox::SearchMsg object containing diff
+#        path_a => pre-image path
+#        path_b => post-image path
+# }
+sub extract_diff ($$$$) {
+        my ($p, $re, $ibx, $smsg) = @_;
+        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 = {};
+        foreach my $l (split(/^/m, $s)) {
+                if ($l =~ /$re/) {
+                        $di->{oid_a} = $1;
+                        $di->{oid_b} = $2;
+                        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
+                        open($tmp, '+>', undef) or die "open(tmp): $!";
+                        $di->{tmp} = $tmp;
+                        $di->{hdr_lines} = $hdr_lines;
+
+                        print $tmp @$hdr_lines, $l or die "print(tmp): $!";
+
+                        # for debugging/diagnostics:
+                        $di->{ibx} = $ibx;
+                        $di->{smsg} = $smsg;
+                } elsif ($l =~ m!\Adiff --git ("?a/.+) ("?b/.+)$!) {
+                        return $di if $tmp; # got our blob, done!
+
+                        my ($path_a, $path_b) = ($1, $2);
+
+                        # 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));
+
+                        # get rid of path-traversal attempts and junk patches:
+                        foreach (@a, @b) {
+                                return if $bad_component{$_};
+                        }
+
+                        $di->{path_a} = join('/', @a);
+                        $di->{path_b} = join('/', @b);
+                        $hdr_lines = [ $l ];
+                } elsif ($tmp) {
+                        print $tmp $l or die "print(tmp): $!";
+                } elsif ($hdr_lines) {
+                        push @$hdr_lines, $l;
+                }
+        }
+        $tmp ? $di : undef;
+}
+
+sub path_searchable ($) { defined($_[0]) && $_[0] =~ m!\A[\w/\. \-]+\z! }
+
+sub find_extract_diff ($$$) {
+        my ($self, $ibx, $want) = @_;
+        my $srch = $ibx->search or return;
+
+        my $post = $want->{oid_b} or die 'BUG: no {oid_b}';
+        $post =~ /\A[a-f0-9]+\z/ or die "BUG: oid_b not hex: $post";
+
+        my $q = "dfpost:$post";
+        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 = $want->{path_b};
+        if (path_searchable($path_b)) {
+                $q .= qq{ dfn:"$path_b"};
+
+                my $path_a = $want->{path_a};
+                if (path_searchable($path_a) && $path_a ne $path_b) {
+                        $q .= qq{ dfn:"$path_a"};
+                }
+        }
+
+        my $msgs = $srch->query($q, { relevance => 1 });
+        my $re = qr/\Aindex ($pre[a-f0-9]*)\.\.($post[a-f0-9]*)(?: (\d+))?/;
+
+        my $di;
+        foreach my $smsg (@$msgs) {
+                $ibx->smsg_mime($smsg) or next;
+                msg_iter(delete($smsg->{mime}), sub {
+                        $di ||= extract_diff($_[0], $re, $ibx, $smsg);
+                });
+                return $di if $di;
+        }
+}
+
+# pure Perl "git init"
+sub do_git_init_wt ($) {
+        my ($self) = @_;
+        my $wt = File::Temp->newdir('solver.wt-XXXXXXXX', TMPDIR => 1);
+        my $dir = $wt->dirname;
+
+        foreach (qw(objects/info refs/heads)) {
+                make_path("$dir/.git/$_") or die "make_path $_: $!";
+        }
+        open my $fh, '>', "$dir/.git/config" or die "open .git/config: $!";
+        print $fh <<'EOF' or die "print .git/config $!";
+[core]
+        repositoryFormatVersion = 0
+        filemode = true
+        bare = false
+        fsyncObjectfiles = false
+        logAllRefUpdates = false
+EOF
+        close $fh or die "close .git/config: $!";
+
+        open $fh, '>', "$dir/.git/HEAD" or die "open .git/HEAD: $!";
+        print $fh "ref: refs/heads/master\n" or die "print .git/HEAD: $!";
+        close $fh or die "close .git/HEAD: $!";
+
+        my $f = '.git/objects/info/alternates';
+        open $fh, '>', "$dir/$f" or die "open: $f: $!";
+        foreach my $git (@{$self->{gits}}) {
+                print $fh "$git->{git_dir}/objects\n" or die "print $f: $!";
+        }
+        close $fh or die "close: $f: $!";
+        $wt;
+}
+
+sub extract_old_mode ($) {
+        my ($di) = @_;
+        if (grep(/\Aold mode (100644|100755|120000)$/, @{$di->{hdr_lines}})) {
+                return $1;
+        }
+        '100644';
+}
+
+sub reap ($$) {
+        my ($pid, $msg) = @_;
+        waitpid($pid, 0) == $pid or die "waitpid($msg): $!";
+        $? == 0 or die "$msg failed: $?";
+}
+
+sub prepare_wt ($$$) {
+        my ($wt_dir, $existing, $di) = @_;
+        my $oid_full = $existing->[1];
+        my ($r, $w);
+        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 @git = (qw(git -C), $wt_dir);
+
+        pipe($r, $w) or die "pipe: $!";
+        my $rdr = { 0 => fileno($r) };
+        my $pid = spawn([@git, qw(update-index -z --index-info)], {}, $rdr);
+        close $r or die "close pipe(r): $!";
+        print $w "$mode_a $oid_full\t$path_a\0" or die "print update-index: $!";
+        close $w or die "close update-index: $!";
+        reap($pid, 'update-index -z --index-info');
+
+        $pid = spawn([@git, qw(checkout-index -a -f -u)]);
+        reap($pid, 'checkout-index -a -f -u');
+}
+
+sub do_apply ($$$$) {
+        my ($out, $wt_git, $wt_dir, $di) = @_;
+
+        my $tmp = delete $di->{tmp} or die "BUG: no tmp ", di_info($di);
+        $tmp->flush or die "tmp->flush failed: $!";
+        $out->flush or die "err->flush failed: $!";
+        sysseek($tmp, 0, SEEK_SET) or die "sysseek(tmp) failed: $!";
+
+        defined(my $err_fd = fileno($out)) or die "fileno(out): $!";
+        my $rdr = { 0 => fileno($tmp), 1 => $err_fd, 2 => $err_fd };
+        my $cmd = [ qw(git -C), $wt_dir,
+                    qw(apply --whitespace=warn -3 --verbose) ];
+        reap(spawn($cmd, undef, $rdr), 'apply');
+
+        local $/ = "\0";
+        my $rd = popen_rd([qw(git -C), $wt_dir, qw(ls-files -s -z)]);
+
+        defined(my $line = <$rd>) or die "failed to read ls-files: $!";
+        chomp $line or die "no trailing \\0 in [$line] from ls-files";
+
+        my ($info, $file) = split(/\t/, $line, 2);
+        my ($mode_b, $oid_b_full, $stage) = split(/ /, $info);
+
+        defined($line = <$rd>) and die "extra files in index: $line";
+        close $rd or die "close ls-files: $?";
+
+        $file eq $di->{path_b} or
+                die "index mismatch: file=$file != path_b=$di->{path_b}";
+        my $abs_path = "$wt_dir/$file";
+        -r $abs_path or die "WT_DIR/$file not readable";
+        my $size = -s _;
+
+        print $out "OK $mode_b $oid_b_full $stage\t$file\n";
+        [ $wt_git, $oid_b_full, 'blob', $size, $di ];
+}
+
+sub di_url ($) {
+        my ($di) = @_;
+        # note: we don't pass the PSGI env here, different inboxes
+        # can have different HTTP_HOST on the same instance.
+        my $url = $di->{ibx}->base_url;
+        my $mid = $di->{smsg}->{mid};
+        defined($url) ? "<$url/$mid/>" : "<$mid>";
+}
+
+sub apply_patches ($$$$$) {
+        my ($self, $out, $wt, $found, $patches) = @_;
+        my $wt_dir = $wt->dirname;
+        my $wt_git = PublicInbox::Git->new("$wt_dir/.git");
+        $wt_git->{-wt} = $wt;
+
+        my $cur = 0;
+        my $tot = scalar @$patches;
+
+        foreach my $di (@$patches) {
+                my $i = ++$cur;
+                my $oid_a = $di->{oid_a};
+                my $existing = $found->{$oid_a};
+                my $empty_oid = $oid_a =~ /\A0+\z/;
+
+                if ($empty_oid && $i != 0) {
+                        die "empty oid at [$i/$tot] ", di_url($di);
+                }
+                if (!$existing && !$empty_oid) {
+                        die "missing $oid_a at [$i/$tot] ", di_url($di);
+                }
+
+                # prepare the worktree for patch application:
+                if ($i == 1 && $existing) {
+                        prepare_wt($wt_dir, $existing, $di);
+                }
+                unless (-f "$wt_dir/$di->{path_a}") {
+                        die "missing $di->{path_a} at [$i/$tot] ", di_url($di);
+                }
+
+                print $out "applying [$i/$tot] ", di_url($di), "\n",
+                           join('', @{$di->{hdr_lines}}), "\n"
+                        or die "print \$out failed: $!";
+
+                # apply the patch!
+                $found->{$di->{oid_b}} = do_apply($out, $wt_git, $wt_dir, $di);
+        }
+}
+
+sub dump_found ($$) {
+        my ($out, $found) = @_;
+        foreach my $oid (sort keys %$found) {
+                my ($git, $oid, $di) = @{$found->{$oid}};
+                my $loc = $di ? di_info($di) : $git->src_blob_url($oid);
+                print $out "$oid from $loc\n";
+        }
+}
+
+sub dump_patches ($$) {
+        my ($out, $patches) = @_;
+        my $tot = scalar(@$patches);
+        my $i = 0;
+        foreach my $di (@$patches) {
+                ++$i;
+                print $out "[$i/$tot] ", di_url($di), "\n";
+        }
+}
+
+# recreate $oid_b
+# Returns a 2-element array ref: [ PublicInbox::Git object, oid_full ]
+# or undef if nothing was found.
+sub solve ($$$$) {
+        my ($self, $out, $oid_b, $hints) = @_;
+
+        # should we even get here? Probably not, but somebody
+        # could be manually typing URLs:
+        return if $oid_b =~ /\A0+\z/;
+
+        my $req = { %$hints, oid_b => $oid_b };
+        my @todo = ($req);
+        my $found = {}; # { oid_abbrev => [ PublicInbox::Git, oid_full, $di ] }
+        my $patches = []; # [ array of $di hashes ]
+
+        my $max = $self->{max_steps} || 200;
+        my $steps = 0;
+
+        while (defined(my $want = pop @todo)) {
+                # see if we can find the blob in an existing git repo:
+                if (my $existing = solve_existing($self, $want)) {
+                        my $want_oid = $want->{oid_b};
+                        return $existing if $want_oid eq $oid_b; # DONE!
+
+                        $found->{$want_oid} = $existing;
+                        next; # ok, one blob resolved, more to go?
+                }
+
+                # scan through inboxes to look for emails which results in
+                # the oid we want:
+                foreach my $ibx (@{$self->{inboxes}}) {
+                        my $di = find_extract_diff($self, $ibx, $want) or next;
+
+                        unshift @$patches, $di;
+
+                        # good, we can find a path to the oid we $want, now
+                        # lets see if we need to apply more patches:
+                        my $src = $di->{oid_a};
+                        if ($src !~ /\A0+\z/) {
+                                if (++$steps > $max) {
+                                        print $out
+"Aborting, too many steps to $oid_b\n";
+
+                                        return;
+                                }
+
+                                # we have to solve it using another oid, fine:
+                                my $job = {
+                                        oid_b => $src,
+                                        path_b => $di->{path_a},
+                                };
+                                push @todo, $job;
+                        }
+                        last; # onto the next @todo item
+                }
+        }
+
+        unless (scalar(@$patches)) {
+                print $out "no patch(es) for $oid_b\n";
+                dump_found($out, $found);
+                return;
+        }
+
+        # reconstruct the oid_b blob using patches we found:
+        eval {
+                my $wt = do_git_init_wt($self);
+                apply_patches($self, $out, $wt, $found, $patches);
+        };
+        if ($@) {
+                print $out "E: $@\nfound: ";
+                dump_found($out, $found);
+                print $out "patches: ";
+                dump_patches($out, $patches);
+                return;
+        }
+
+        $found->{$oid_b};
+}
+
+1;