solvergit: avoid "Wide character" warnings

[public-inbox.git] / lib / PublicInbox / SolverGit.pm

diff --git a/lib/PublicInbox/SolverGit.pm b/lib/PublicInbox/SolverGit.pm
index dfc28d260c7c490c1f271f25a21bc3a7dc8dd107..1b1951b58e17a733f4396f258dcfac93abc0b7c0 100644 (file)
--- a/lib/PublicInbox/SolverGit.pm
+++ b/lib/PublicInbox/SolverGit.pm
@@ -24,6 +24,14 @@ use URI::Escape qw(uri_escape_utf8);
 use POSIX qw(sysconf _SC_ARG_MAX);
 my $ARG_SIZE_MAX = (sysconf(_SC_ARG_MAX) || 4096) - 2048;
 
+# By default, "git format-patch" generates filenames with a four-digit
+# prefix, so that means 9999 patch series are OK, right? :>
+# Maybe we can make this configurable, main concern is disk space overhead
+# for uncompressed patch fragments.  Aside from space, public-inbox-httpd
+# is otherwise unaffected by having many patches, here, as it can share
+# work fairly.  Other PSGI servers may have trouble, though.
+my $MAX_PATCH = 9999;
+
 # di = diff info / a hashref with information about a diff ($di):
 # {
 #      oid_a => abbreviated pre-image oid,
@@ -117,6 +125,7 @@ sub extract_diff ($$$$$) {
 
                        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:
@@ -145,6 +154,7 @@ sub extract_diff ($$$$$) {
                        $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, $l;
@@ -381,8 +391,8 @@ sub do_git_apply ($) {
        my $patches = $self->{patches};
 
        # we need --ignore-whitespace because some patches are CRLF
-       my @cmd = qw(git apply --cached --ignore-whitespace
-                       --whitespace=warn --verbose);
+       my @cmd = (qw(git -C), $dn, qw(apply --cached --ignore-whitespace
+                       --whitespace=warn --verbose));
        my $len = length(join(' ', @cmd));
        my $total = $self->{tot};
        my $di; # keep track of the last one for "git ls-files"
@@ -392,8 +402,7 @@ sub do_git_apply ($) {
                $di = shift @$patches;
                dbg($self, "\napplying [$i/$total] " . di_url($self, $di) .
                        "\n" . join('', @{$di->{hdr_lines}}));
-               my $pn = $total + 1 - $i;
-               my $path = "$dn/$pn";
+               my $path = $total + 1 - $i;
                $len += length($path) + 1;
                push @cmd, $path;
        } while (@$patches && $len < $ARG_SIZE_MAX);
@@ -425,12 +434,15 @@ sub di_url ($$) {
 sub resolve_patch ($$) {
        my ($self, $want) = @_;
 
-       if (scalar(@{$self->{patches}}) > $self->{max_patch}) {
+       if (scalar(@{$self->{patches}}) > $MAX_PATCH) {
                die "Aborting, too many steps to $self->{oid_want}";
        }
 
        # see if we can find the blob in an existing git repo:
        my $cur_want = $want->{oid_b};
+       if ($self->{seen_oid}->{$cur_want}++) {
+               die "Loop detected solving $cur_want\n";
+       }
        if (my $existing = solve_existing($self, $want)) {
                dbg($self, "found $cur_want in " .
                        join("\n", $existing->[0]->pub_urls));
@@ -477,7 +489,6 @@ sub new {
        bless {
                gits => $ibx->{-repo_objs},
                user_cb => $user_cb,
-               max_patch => 100,
 
                # TODO: config option for searching related inboxes
                inboxes => [ $ibx ],
@@ -497,6 +508,7 @@ sub solve ($$$$$) {
 
        $self->{oid_want} = $oid_want;
        $self->{out} = $out;
+       $self->{seen_oid} = {};
        $self->{tot} = 0;
        $self->{psgi_env} = $env;
        $self->{todo} = [ { %$hints, oid_b => $oid_want } ];