X-Git-Url: http://www.git.stargrave.org/?p=public-inbox.git;a=blobdiff_plain;f=lib%2FPublicInbox%2FGit.pm;h=9c96b3f0b6920acbcfd4a103e3ec14a7ec8618b6;hp=9014e021cedcb9084ac50eea515e0a71119f1e94;hb=64a5e4f9b5cda6891d3dbd719d8a010da487bfc3;hpb=af00ac5113c1e7acc203a0051d669ad9d1a43b90 diff --git a/lib/PublicInbox/Git.pm b/lib/PublicInbox/Git.pm index 9014e021..9c96b3f0 100644 --- a/lib/PublicInbox/Git.pm +++ b/lib/PublicInbox/Git.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2014-2018 all contributors +# Copyright (C) 2014-2020 all contributors # License: GPLv2 or later # # Used to read files from a git repository without excessive forking. @@ -9,12 +9,19 @@ package PublicInbox::Git; use strict; use warnings; -use POSIX qw(dup2); -require IO::Handle; -use PublicInbox::Spawn qw(spawn popen_rd); +use POSIX (); +use IO::Handle; # ->autoflush +use File::Glob qw(bsd_glob GLOB_NOSORT); +use PublicInbox::Spawn qw(popen_rd); +use PublicInbox::Tmpfile; use base qw(Exporter); our @EXPORT_OK = qw(git_unquote git_quote); +use constant MAX_INFLIGHT => + ($^O eq 'linux' ? 4096 : POSIX::_POSIX_PIPE_BUF()) + / + 65; # SHA-256 hex size + "\n" in preparation for git using non-SHA1 + my %GIT_ESC = ( a => "\a", b => "\b", @@ -97,35 +104,74 @@ sub _bidi_pipe { } return; } - my ($in_r, $in_w, $out_r, $out_w); - - pipe($in_r, $in_w) or fail($self, "pipe failed: $!"); + my ($out_r, $out_w); pipe($out_r, $out_w) or fail($self, "pipe failed: $!"); - if ($^O eq 'linux') { # 1031: F_SETPIPE_SZ - fcntl($out_w, 1031, 4096); - fcntl($in_w, 1031, 4096) if $batch eq '--batch-check'; - } - my @cmd = (qw(git), "--git-dir=$self->{git_dir}", qw(-c core.abbrev=40 cat-file), $batch); - my $redir = { 0 => fileno($out_r), 1 => fileno($in_w) }; + my $redir = { 0 => $out_r }; if ($err) { - open(my $fh, '+>', undef) or fail($self, "open.err failed: $!"); + my $id = "git.$self->{git_dir}$batch.err"; + my $fh = tmpfile($id) or fail($self, "tmpfile($id): $!"); $self->{$err} = $fh; - $redir->{2} = fileno($fh); + $redir->{2} = $fh; } - my $p = spawn(\@cmd, undef, $redir); - defined $p or fail($self, "spawn failed: $!"); + my ($in_r, $p) = popen_rd(\@cmd, undef, $redir); $self->{$pid} = $p; $out_w->autoflush(1); + if ($^O eq 'linux') { # 1031: F_SETPIPE_SZ + fcntl($out_w, 1031, 4096); + fcntl($in_r, 1031, 4096) if $batch eq '--batch-check'; + } $self->{$out} = $out_w; $self->{$in} = $in_r; } +sub read_cat_in_full ($$$) { + my ($self, $in, $left) = @_; + my $offset = 0; + my $buf = ''; + while ($left > 0) { + my $r = read($in, $buf, $left, $offset); + defined($r) or fail($self, "read failed: $!"); + $r == 0 and fail($self, 'exited unexpectedly'); + $left -= $r; + $offset += $r; + } + my $r = read($in, my $lf, 1); + defined($r) or fail($self, "read failed: $!"); + fail($self, 'newline missing after blob') if ($r != 1 || $lf ne "\n"); + \$buf; +} + +sub _cat_async_step ($$$) { + my ($self, $inflight, $in) = @_; + my $pair = shift @$inflight or die 'BUG: inflight empty'; + my ($cb, $arg) = @$pair; + local $/ = "\n"; + my $head = $in->getline; + $head =~ / missing$/ and return + eval { $cb->(undef, undef, undef, undef, $arg) }; + + $head =~ /^([0-9a-f]{40}) (\S+) ([0-9]+)$/ or + fail($self, "Unexpected result from async git cat-file: $head"); + my ($oid_hex, $type, $size) = ($1, $2, $3 + 0); + my $bref = read_cat_in_full($self, $in, $size); + eval { $cb->($bref, $oid_hex, $type, $size, $arg) }; +} + +sub cat_async_wait ($) { + my ($self) = @_; + my $inflight = delete $self->{inflight} or return; + my $in = $self->{in}; + while (scalar(@$inflight)) { + _cat_async_step($self, $inflight, $in); + } +} + sub cat_file { my ($self, $obj, $ref) = @_; my ($retried, $in, $head); - + cat_async_wait($self); again: batch_prepare($self); $self->{out}->print($obj, "\n") or fail($self, "write error: $!"); @@ -141,47 +187,12 @@ again: } return; } - $head =~ /^[0-9a-f]{40} \S+ (\d+)$/ or + $head =~ /^[0-9a-f]{40} \S+ ([0-9]+)$/ or fail($self, "Unexpected result from git cat-file: $head"); my $size = $1; - my $ref_type = $ref ? ref($ref) : ''; - - my $rv; - my $left = $size; - $$ref = $size if ($ref_type eq 'SCALAR'); - my $cb_err; - - if ($ref_type eq 'CODE') { - $rv = eval { $ref->($in, \$left) }; - $cb_err = $@; - # drain the rest - my $max = 8192; - while ($left > 0) { - my $r = read($in, my $x, $left > $max ? $max : $left); - defined($r) or fail($self, "read failed: $!"); - $r == 0 and fail($self, 'exited unexpectedly'); - $left -= $r; - } - } else { - my $offset = 0; - my $buf = ''; - while ($left > 0) { - my $r = read($in, $buf, $left, $offset); - defined($r) or fail($self, "read failed: $!"); - $r == 0 and fail($self, 'exited unexpectedly'); - $left -= $r; - $offset += $r; - } - $rv = \$buf; - } - - my $r = read($in, my $buf, 1); - defined($r) or fail($self, "read failed: $!"); - fail($self, 'newline missing after blob') if ($r != 1 || $buf ne "\n"); - die $cb_err if $cb_err; - - $rv; + $$ref = $size if $ref; + read_cat_in_full($self, $in, $size); } sub batch_prepare ($) { _bidi_pipe($_[0], qw(--batch in out pid)) } @@ -215,12 +226,21 @@ sub _destroy { my $p = delete $self->{$pid} or return; delete @$self{($in, $out)}; delete $self->{$err} if $err; # `err_c' - waitpid $p, 0; + + # PublicInbox::DS may not be loaded + eval { PublicInbox::DS::dwaitpid($p, undef, undef) }; + waitpid($p, 0) if $@; # wait synchronously if not in event loop +} + +sub cat_async_abort ($) { + my ($self) = @_; + my $inflight = delete $self->{inflight} or die 'BUG: not in async'; + cleanup($self); } sub fail { my ($self, $msg) = @_; - cleanup($self); + $self->{inflight} ? cat_async_abort($self) : cleanup($self); die $msg; } @@ -233,7 +253,6 @@ sub popen { sub qx { my ($self, @cmd) = @_; my $fh = $self->popen(@cmd); - defined $fh or return; local $/ = "\n"; return <$fh> if wantarray; local $/; @@ -255,7 +274,7 @@ sub packed_bytes { my ($self) = @_; my $n = 0; my $pack_dir = git_path($self, 'objects/pack'); - foreach my $p (glob("$pack_dir/*.pack")) { + foreach my $p (bsd_glob("$pack_dir/*.pack", GLOB_NOSORT)) { $n += -s $p; } $n @@ -273,22 +292,11 @@ sub local_nick ($) { wantarray ? ($ret) : $ret; } -# show the blob URL for cgit/gitweb/whatever -sub src_blob_url { - my ($self, $oid) = @_; - # blob_url_format = "https://example.com/foo.git/blob/%s" - if (my $bfu = $self->{blob_url_format}) { - return map { sprintf($_, $oid) } @$bfu if wantarray; - return sprintf($bfu->[0], $oid); - } - local_nick($self); -} - sub host_prefix_url ($$) { my ($env, $url) = @_; return $url if index($url, '//') >= 0; my $scheme = $env->{'psgi.url_scheme'}; - my $host_port = $env->{HTTP_HOST} || + my $host_port = $env->{HTTP_HOST} // "$env->{SERVER_NAME}:$env->{SERVER_PORT}"; "$scheme://$host_port". ($env->{SCRIPT_NAME} || '/') . $url; } @@ -301,11 +309,32 @@ sub pub_urls { local_nick($self); } -sub commit_title ($$) { - my ($self, $oid) = @_; # PublicInbox::Git, $sha1hex - my $buf = cat_file($self, $oid) or return; - utf8::decode($$buf); - ($$buf =~ /\r?\n\r?\n([^\r\n]+)\r?\n?/)[0] +sub cat_async_begin { + my ($self) = @_; + cleanup($self) if alternates_changed($self); + batch_prepare($self); + die 'BUG: already in async' if $self->{inflight}; + $self->{inflight} = []; +} + +sub cat_async ($$$;$) { + my ($self, $oid, $cb, $arg) = @_; + my $inflight = $self->{inflight} or die 'BUG: not in async'; + if (scalar(@$inflight) >= MAX_INFLIGHT) { + _cat_async_step($self, $inflight, $self->{in}); + } + + $self->{out}->print($oid, "\n") or fail($self, "write error: $!"); + push(@$inflight, [ $cb, $arg ]); +} + +sub extract_cmt_time { + my ($bref, undef, undef, undef, $modified) = @_; + + if ($$bref =~ /^committer .*?> ([0-9]+) [\+\-]?[0-9]+/sm) { + my $cmt_time = $1 + 0; + $$modified = $cmt_time if $cmt_time > $$modified; + } } # returns the modified time of a git repo, same as the "modified" field @@ -314,15 +343,13 @@ sub modified ($) { my ($self) = @_; my $modified = 0; my $fh = popen($self, qw(rev-parse --branches)); - defined $fh or return $modified; + cat_async_begin($self); local $/ = "\n"; - foreach my $oid (<$fh>) { + while (my $oid = <$fh>) { chomp $oid; - my $buf = cat_file($self, $oid) or next; - $$buf =~ /^committer .*?> (\d+) [\+\-]?\d+/sm or next; - my $cmt_time = $1; - $modified = $cmt_time if $cmt_time > $modified; + cat_async($self, $oid, \&extract_cmt_time, \$modified); } + cat_async_wait($self); $modified || time; }