diff options
Diffstat (limited to 'lib/PublicInbox/Git.pm')
-rw-r--r-- | lib/PublicInbox/Git.pm | 727 |
1 files changed, 525 insertions, 202 deletions
diff --git a/lib/PublicInbox/Git.pm b/lib/PublicInbox/Git.pm index 7eaaeb8b..aea389e8 100644 --- a/lib/PublicInbox/Git.pm +++ b/lib/PublicInbox/Git.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2014-2020 all contributors <meta@public-inbox.org> +# Copyright (C) all contributors <meta@public-inbox.org> # License: GPLv2 or later <https://www.gnu.org/licenses/gpl-2.0.txt> # # Used to read files from a git repository without excessive forking. @@ -8,19 +8,35 @@ # There are also API changes to simplify our usage and data set. package PublicInbox::Git; use strict; -use warnings; -use POSIX qw(dup2); -use IO::Handle; # ->autoflush +use v5.10.1; +use parent qw(Exporter PublicInbox::DS); +use autodie qw(socketpair read); +use POSIX (); +use Socket qw(AF_UNIX SOCK_STREAM); +use PublicInbox::Syscall qw(EPOLLIN EPOLLET); +use Errno qw(EAGAIN); use File::Glob qw(bsd_glob GLOB_NOSORT); -use PublicInbox::Spawn qw(popen_rd); +use File::Spec (); +use PublicInbox::Spawn qw(spawn popen_rd run_qx which); +use PublicInbox::IO qw(read_all try_cat); 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 +use Carp qw(croak carp); +use PublicInbox::SHA qw(sha_all); +our %HEXLEN2SHA = (40 => 1, 64 => 256); +our %OFMT2HEXLEN = (sha1 => 40, sha256 => 64); +our @EXPORT_OK = qw(git_unquote git_quote %HEXLEN2SHA %OFMT2HEXLEN + $ck_unlinked_packs); +our $in_cleanup; +our $async_warn; # true in read-only daemons + +# committerdate:unix is git 2.9.4+ (2017-05-05), so using raw instead +my @MODIFIED_DATE = qw[for-each-ref --sort=-committerdate + --format=%(committerdate:raw) --count=1]; + +use constant { + MAX_INFLIGHT => 18, # arbitrary, formerly based on PIPE_BUF + BATCH_CMD_VER => v2.36.0, # git 2.36+ +}; my %GIT_ESC = ( a => "\a", @@ -35,19 +51,40 @@ my %GIT_ESC = ( ); my %ESC_GIT = map { $GIT_ESC{$_} => $_ } keys %GIT_ESC; +my $EXE_ST = ''; # pack('dd', st_dev, st_ino); # no `q' in some 32-bit builds +my ($GIT_EXE, $GIT_VER); + +sub check_git_exe () { + $GIT_EXE = which('git') // die "git not found in $ENV{PATH}"; + my @st = stat(_) or die "stat($GIT_EXE): $!"; # can't do HiRes w/ _ + my $st = pack('dd', $st[0], $st[1]); + if ($st ne $EXE_ST) { + my $v = run_qx([ $GIT_EXE, '--version' ]); + die "$GIT_EXE --version: \$?=$?" if $?; + $v =~ /\b([0-9]+(?:\.[0-9]+){2})/ or die + "$GIT_EXE --version output: $v # unparseable"; + $GIT_VER = eval("v$1") // die "BUG: bad vstring: $1 ($v)"; + $EXE_ST = $st; + } + $GIT_EXE; +} + +sub git_version { + check_git_exe(); + $GIT_VER; +} # unquote pathnames used by git, see quote.c::unquote_c_style.c in git.git sub git_unquote ($) { return $_[0] unless ($_[0] =~ /\A"(.*)"\z/); $_[0] = $1; - $_[0] =~ s/\\([\\"abfnrtv])/$GIT_ESC{$1}/g; - $_[0] =~ s/\\([0-7]{1,3})/chr(oct($1))/ge; + $_[0] =~ s!\\([\\"abfnrtv]|[0-3][0-7]{2})!$GIT_ESC{$1}//chr(oct($1))!ge; $_[0]; } sub git_quote ($) { if ($_[0] =~ s/([\\"\a\b\f\n\r\t\013]|[^[:print:]])/ - '\\'.($ESC_GIT{$1}||sprintf("%0o",ord($1)))/egs) { + '\\'.($ESC_GIT{$1}||sprintf("%03o",ord($1)))/egs) { return qq{"$_[0]"}; } $_[0]; @@ -55,216 +92,393 @@ sub git_quote ($) { sub new { my ($class, $git_dir) = @_; - my @st; - $st[7] = $st[10] = 0; + $git_dir .= '/'; + $git_dir =~ tr!/!/!s; + chop $git_dir; # may contain {-tmp} field for File::Temp::Dir - bless { git_dir => $git_dir, st => \@st, -git_path => {} }, $class + my %dedupe = ($git_dir => undef); + bless { git_dir => (keys %dedupe)[0] }, $class } sub git_path ($$) { my ($self, $path) = @_; - $self->{-git_path}->{$path} ||= do { - local $/ = "\n"; - chomp(my $str = $self->qx(qw(rev-parse --git-path), $path)); - - # git prior to 2.5.0 did not understand --git-path - if ($str eq "--git-path\n$path") { - $str = "$self->{git_dir}/$path"; + $self->{-git_path}->{$path} //= do { + my $d = "$self->{git_dir}/$path"; + if (-e $d) { + $d; + } else { + local $/ = "\n"; + my $rdr = { 2 => \my $err }; + my $s = $self->qx([qw(rev-parse --git-path), $path], + undef, $rdr); + chomp $s; + + # git prior to 2.5.0 did not understand --git-path + $s eq "--git-path\n$path" ? $d : $s; } - $str; }; } sub alternates_changed { my ($self) = @_; my $alt = git_path($self, 'objects/info/alternates'); + use Time::HiRes qw(stat); my @st = stat($alt) or return 0; - my $old_st = $self->{st}; - # 10 - ctime, 7 - size - return 0 if ($st[10] == $old_st->[10] && $st[7] == $old_st->[7]); - $self->{st} = \@st; + + # can't rely on 'q' on some 32-bit builds, but `d' works + my $st = pack('dd', $st[10], $st[7]); # 10: ctime, 7: size + return 0 if ($self->{alt_st} // '') eq $st; + $self->{alt_st} = $st; # always a true value +} + +sub object_format { + $_[0]->{object_format} //= do { + my $fmt = $_[0]->qx(qw(config extensions.objectformat)); + $fmt eq "sha256\n" ? \'sha256' : \undef; + } } sub last_check_err { my ($self) = @_; - my $fh = $self->{err_c} or return; - sysseek($fh, 0, 0) or fail($self, "sysseek failed: $!"); - defined(sysread($fh, my $buf, -s $fh)) or - fail($self, "sysread failed: $!"); + my $fh = $self->{err_c} or return ''; + sysseek($fh, 0, 0) or $self->fail("sysseek: $!"); + my $size = -s $fh or return ''; + sysread($fh, my $buf, $size) // $self->fail("sysread: $!"); + truncate($fh, 0) or $self->fail("truncate: $!"); $buf; } -sub _bidi_pipe { - my ($self, $batch, $in, $out, $pid, $err) = @_; - if ($self->{$pid}) { - if (defined $err) { # "err_c" - my $fh = $self->{$err}; - sysseek($fh, 0, 0) or fail($self, "sysseek failed: $!"); - truncate($fh, 0) or fail($self, "truncate failed: $!"); - } - return; +sub gcf_drain { # awaitpid cb + my ($pid, $inflight, $bc) = @_; + while (@$inflight) { + my ($req, $cb, $arg) = splice(@$inflight, 0, 3); + $req = $$req if ref($req); + $bc and $req =~ s/\A(?:contents|info) //; + $req =~ s/ .*//; # drop git_dir for Gcf2Client + eval { $cb->(undef, $req, undef, undef, $arg) }; + warn "E: (in abort) $req: $@" if $@; } - my ($out_r, $out_w); - pipe($out_r, $out_w) or fail($self, "pipe failed: $!"); - my @cmd = (qw(git), "--git-dir=$self->{git_dir}", - qw(-c core.abbrev=40 cat-file), $batch); - my $redir = { 0 => $out_r }; - if ($err) { - my $id = "git.$self->{git_dir}$batch.err"; - my $fh = tmpfile($id) or fail($self, "tmpfile($id): $!"); - $self->{$err} = $fh; - $redir->{2} = $fh; +} + +sub _sock_cmd { + my ($self, $batch, $err_c) = @_; + $self->{sock} and Carp::confess('BUG: {sock} exists'); + socketpair(my $s1, my $s2, AF_UNIX, SOCK_STREAM, 0); + $s1->blocking(0); + my $opt = { pgid => 0, 0 => $s2, 1 => $s2 }; + my $gd = $self->{git_dir}; + if ($gd =~ s!/([^/]+/[^/]+)\z!/!) { + $opt->{-C} = $gd; + $gd = $1; } - my ($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'; + + # git 2.31.0+ supports -c core.abbrev=no, don't bother with + # core.abbrev=64 since not many releases had SHA-256 prior to 2.31 + my $abbr = $GIT_VER lt v2.31.0 ? 40 : 'no'; + my @cmd = ($GIT_EXE, "--git-dir=$gd", '-c', "core.abbrev=$abbr", + 'cat-file', "--$batch"); + if ($err_c) { + my $id = "git.$self->{git_dir}.$batch.err"; + $self->{err_c} = $opt->{2} = tmpfile($id, undef, 1) or + $self->fail("tmpfile($id): $!"); } - $self->{$out} = $out_w; - $self->{$in} = $in_r; -} - -sub read_cat_in_full ($$$) { - my ($self, $in, $left) = @_; - my $offset = 0; - my $buf = ''; - while ($left > 0) { - my $r = read($in, $buf, $left, $offset); - defined($r) or fail($self, "read failed: $!"); - $r == 0 and fail($self, 'exited unexpectedly'); - $left -= $r; - $offset += $r; + my $inflight = []; # TODO consider moving this into the IO object + my $pid = spawn(\@cmd, undef, $opt); + $self->{sock} = PublicInbox::IO::attach_pid($s1, $pid, + \&gcf_drain, $inflight, $self->{-bc}); + $self->{inflight} = $inflight; +} + +sub cat_async_retry ($$) { + my ($self, $old_inflight) = @_; + + # {inflight} may be non-existent, but if it isn't we delete it + # here to prevent cleanup() from waiting: + my ($sock, $epwatch) = delete @$self{qw(sock epwatch inflight)}; + $self->SUPER::close if $epwatch; + my $new_inflight = batch_prepare($self); + + while (my ($oid, $cb, $arg) = splice(@$old_inflight, 0, 3)) { + write_all($self, $oid."\n", \&cat_async_step, $new_inflight); + $oid = \$oid if !@$new_inflight; # to indicate oid retried + push @$new_inflight, $oid, $cb, $arg; } - 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; + $sock->close if $sock; # only safe once old_inflight is empty + cat_async_step($self, $new_inflight); # take one step } -sub _cat_async_step ($$$) { - my ($self, $inflight, $in) = @_; - my $pair = shift @$inflight or die 'BUG: inflight empty'; - my ($cb, $arg) = @$pair; - local $/ = "\n"; - my $head = $in->getline; - $head =~ / missing$/ and return - eval { $cb->(undef, undef, undef, undef, $arg) }; +sub gcf_inflight ($) { + my ($self) = @_; + # FIXME: the first {sock} check can succeed but Perl can complain + # about an undefined value. Not sure why or how this happens but + # t/imapd.t can complain about it, sometimes. + if ($self->{sock}) { + if (eval { $self->{sock}->can_reap }) { + return $self->{inflight}; + } elsif ($@) { + no warnings 'uninitialized'; + warn "E: $self sock=$self->{sock}: can_reap failed: ". + "$@ (continuing...)"; + } + delete @$self{qw(sock inflight)}; + } else { + $self->close; + } + undef; +} - $head =~ /^([0-9a-f]{40}) (\S+) ([0-9]+)$/ or - fail($self, "Unexpected result from async git cat-file: $head"); - my ($oid_hex, $type, $size) = ($1, $2, $3 + 0); - my $bref = read_cat_in_full($self, $in, $size); - eval { $cb->($bref, $oid_hex, $type, $size, $arg) }; +# returns true if prefetch is successful +sub async_prefetch { + my ($self, $oid, $cb, $arg) = @_; + my $inflight = gcf_inflight($self) or return; + return if @$inflight; + substr($oid, 0, 0) = 'contents ' if $self->{-bc}; + write_all($self, "$oid\n", \&cat_async_step, $inflight); + push(@$inflight, $oid, $cb, $arg); +} + +sub cat_async_step ($$) { + my ($self, $inflight) = @_; + croak 'BUG: inflight empty or odd' if scalar(@$inflight) < 3; + my ($req, $cb, $arg) = @$inflight[0, 1, 2]; + my ($bref, $oid, $type, $size); + my $head = $self->{sock}->my_readline; + my $cmd = ref($req) ? $$req : $req; + # ->fail may be called via Gcf2Client.pm + my $info = $self->{-bc} && substr($cmd, 0, 5) eq 'info '; + if ($head =~ /^([0-9a-f]{40,}) (\S+) ([0-9]+)$/) { + ($oid, $type, $size) = ($1, $2, $3 + 0); + unless ($info) { # --batch-command + $bref = $self->{sock}->my_bufread($size + 1) or + $self->fail(defined($bref) ? + 'read EOF' : "read: $!"); + chop($$bref) eq "\n" or + $self->fail('LF missing after blob'); + } + } elsif ($info && $head =~ / (missing|ambiguous)\n/) { + $type = $1; + $oid = substr($cmd, 5); # remove 'info ' + } elsif ($head =~ s/ missing\n//s) { + $oid = $head; + # ref($req) indicates it's already been retried + # -gcf2 retries internally, so it never hits this path: + if (!ref($req) && !$in_cleanup && $self->alternates_changed) { + return cat_async_retry($self, $inflight); + } + $type = 'missing'; + if ($oid eq '') { + $oid = $cmd; + $oid =~ s/\A(?:contents|info) // if $self->{-bc}; + } + } else { + my $err = $! ? " ($!)" : ''; + $self->fail("bad result from async cat-file: $head$err"); + } + splice(@$inflight, 0, 3); # don't retry $cb on ->fail + eval { $cb->($bref, $oid, $type, $size, $arg) }; + async_err($self, $req, $oid, $@, $info ? 'check' : 'cat') if $@; } sub cat_async_wait ($) { my ($self) = @_; - my $inflight = delete $self->{inflight} or return; - my $in = $self->{in}; - while (scalar(@$inflight)) { - _cat_async_step($self, $inflight, $in); + my $inflight = gcf_inflight($self) or return; + cat_async_step($self, $inflight) while (scalar(@$inflight)); +} + +sub batch_prepare ($) { + my ($self) = @_; + check_git_exe(); + if ($GIT_VER ge BATCH_CMD_VER) { + $self->{-bc} = 1; + _sock_cmd($self, 'batch-command', 1); + } else { + _sock_cmd($self, 'batch'); } } +sub _cat_file_cb { + my ($bref, $oid, $type, $size, $result) = @_; + @$result = ($bref, $oid, $type, $size); +} + sub cat_file { - my ($self, $obj, $ref) = @_; - my ($retried, $in, $head); + my ($self, $oid) = @_; + my $result = []; + cat_async($self, $oid, \&_cat_file_cb, $result); cat_async_wait($self); -again: - batch_prepare($self); - $self->{out}->print($obj, "\n") or fail($self, "write error: $!"); - - $in = $self->{in}; - local $/ = "\n"; - $head = $in->getline; - if ($head =~ / missing$/) { - if (!$retried && alternates_changed($self)) { - $retried = 1; - cleanup($self); - goto again; - } - return; - } - $head =~ /^[0-9a-f]{40} \S+ ([0-9]+)$/ or - fail($self, "Unexpected result from git cat-file: $head"); - - my $size = $1; - $$ref = $size if $ref; - read_cat_in_full($self, $in, $size); + wantarray ? @$result : $result->[0]; } -sub batch_prepare ($) { _bidi_pipe($_[0], qw(--batch in out pid)) } +sub check_async_step ($$) { + my ($ck, $inflight) = @_; + croak 'BUG: inflight empty or odd' if scalar(@$inflight) < 3; + my ($req, $cb, $arg) = @$inflight[0, 1, 2]; + chomp(my $line = $ck->{sock}->my_readline); + my ($hex, $type, $size) = split(/ /, $line); -sub check { - my ($self, $obj) = @_; - _bidi_pipe($self, qw(--batch-check in_c out_c pid_c err_c)); - $self->{out_c}->print($obj, "\n") or fail($self, "write error: $!"); - local $/ = "\n"; - chomp(my $line = $self->{in_c}->getline); - my ($hex, $type, $size) = split(' ', $line); - - # Future versions of git.git may show 'ambiguous', but for now, - # we must handle 'dangling' below (and maybe some other oddball - # stuff): + # git <2.21 would show `dangling' (2.21+ shows `ambiguous') # https://public-inbox.org/git/20190118033845.s2vlrb3wd3m2jfzu@dcvr/T/ - return if $type eq 'missing' || $type eq 'ambiguous'; + if ($hex eq 'dangling') { + my $ret = $ck->{sock}->my_bufread($type + 1); + $ck->fail(defined($ret) ? 'read EOF' : "read: $!") if !$ret; + } + splice(@$inflight, 0, 3); # don't retry $cb on ->fail + eval { $cb->(undef, $hex, $type, $size, $arg) }; + async_err($ck, $req, $hex, $@, 'check') if $@; +} + +sub check_async_wait ($) { + my ($self) = @_; + return cat_async_wait($self) if $self->{-bc}; + my $ck = $self->{ck} or return; + my $inflight = gcf_inflight($ck) or return; + check_async_step($ck, $inflight) while (scalar(@$inflight)); +} + +# git <2.36 +sub ck { + $_[0]->{ck} //= bless { git_dir => $_[0]->{git_dir} }, + 'PublicInbox::GitCheck'; +} - if ($hex eq 'dangling' || $hex eq 'notdir' || $hex eq 'loop') { - $size = $type + length("\n"); - my $r = read($self->{in_c}, my $buf, $size); - defined($r) or fail($self, "read failed: $!"); - return; +sub check_async_begin ($) { + my ($self) = @_; + cleanup($self) if alternates_changed($self); + check_git_exe(); + if ($GIT_VER ge BATCH_CMD_VER) { + $self->{-bc} = 1; + _sock_cmd($self, 'batch-command', 1); + } else { + _sock_cmd($self = ck($self), 'batch-check', 1); } +} - ($hex, $type, $size); +sub write_all { + my ($self, $buf, $read_step, $inflight) = @_; + $self->{sock} // Carp::confess 'BUG: no {sock}'; + Carp::confess('BUG: not an array') if ref($inflight) ne 'ARRAY'; + $read_step->($self, $inflight) while @$inflight >= MAX_INFLIGHT; + do { + my $w = syswrite($self->{sock}, $buf); + if (defined $w) { + return if $w == length($buf); + substr($buf, 0, $w, ''); # sv_chop + } elsif ($! != EAGAIN) { + $self->fail("write: $!"); + } + $read_step->($self, $inflight); + } while (1); } -sub _destroy { - my ($self, $in, $out, $pid, $err) = @_; - my $p = delete $self->{$pid} or return; - delete @$self{($in, $out)}; - delete $self->{$err} if $err; # `err_c' +sub check_async ($$$$) { + my ($self, $oid, $cb, $arg) = @_; + my $inflight; + if ($self->{-bc}) { # likely as time goes on +batch_command: + $inflight = gcf_inflight($self) // cat_async_begin($self); + substr($oid, 0, 0) = 'info '; + write_all($self, "$oid\n", \&cat_async_step, $inflight); + } else { # accounts for git upgrades while we're running: + my $ck = $self->{ck}; # undef OK, maybe set in check_async_begin + $inflight = ($ck ? gcf_inflight($ck) : undef) + // check_async_begin($self); + goto batch_command if $self->{-bc}; + write_all($self->{ck}, "$oid\n", \&check_async_step, $inflight); + } + push(@$inflight, $oid, $cb, $arg); +} - # PublicInbox::DS may not be loaded - eval { PublicInbox::DS::dwaitpid($p, undef, undef) }; - waitpid($p, 0) if $@; # wait synchronously if not in event loop +sub _check_cb { # check_async callback + my (undef, $hex, $type, $size, $result) = @_; + @$result = ($hex, $type, $size); } -sub cat_async_abort ($) { - my ($self) = @_; - my $inflight = delete $self->{inflight} or die 'BUG: not in async'; - cleanup($self); +sub check { + my ($self, $oid) = @_; + my $result = []; + check_async($self, $oid, \&_check_cb, $result); + check_async_wait($self); + my ($hex, $type, $size) = @$result; + + # git <2.21 would show `dangling' (2.21+ shows `ambiguous') + # https://public-inbox.org/git/20190118033845.s2vlrb3wd3m2jfzu@dcvr/T/ + return if $type =~ /\A(?:missing|ambiguous)\z/ || $hex eq 'dangling'; + ($hex, $type, $size); } sub fail { my ($self, $msg) = @_; - $self->{inflight} ? cat_async_abort($self) : cleanup($self); - die $msg; + $self->close; + croak(ref($self) . ' ' . ($self->{git_dir} // '') . ": $msg"); +} + +sub async_err ($$$$$) { + my ($self, $req, $oid, $err, $action) = @_; + $req = $$req if ref($req); # retried + my $msg = "E: $action $req ($oid): $err"; + $async_warn ? carp($msg) : $self->fail($msg); +} + +sub cmd { + my $self = shift; + [ $GIT_EXE // check_git_exe(), "--git-dir=$self->{git_dir}", @_ ] } +# $git->popen(qw(show f00)); # or +# $git->popen(qw(show f00), { GIT_CONFIG => ... }, { 2 => ... }); sub popen { - my ($self, @cmd) = @_; - @cmd = ('git', "--git-dir=$self->{git_dir}", @cmd); - popen_rd(\@cmd); + my ($self, $cmd) = splice(@_, 0, 2); + $cmd = [ 'git', "--git-dir=$self->{git_dir}", + ref($cmd) ? @$cmd : ($cmd, grep { defined && !ref } @_) ]; + popen_rd($cmd, grep { !defined || ref } @_); # env and opt } +# same args as popen above sub qx { - my ($self, @cmd) = @_; - my $fh = $self->popen(@cmd); - local $/ = "\n"; - return <$fh> if wantarray; - local $/; - <$fh> + my $fh = popen(@_); + if (wantarray) { + my @ret = <$fh>; + $fh->close; # caller should check $? + @ret; + } else { + local $/; + my $ret = <$fh>; + $fh->close; # caller should check $? + $ret; + } +} + +sub date_parse { + my $self = shift; + map { + substr($_, length('--max-age='), -1) + } $self->qx('rev-parse', map { "--since=$_" } @_); +} + +sub _active ($) { + scalar(@{gcf_inflight($_[0]) // []}) || + ($_[0]->{ck} && scalar(@{gcf_inflight($_[0]->{ck}) // []})) +} + +# check_async and cat_async may trigger the other, so ensure they're +# both completely done by using this: +sub async_wait_all ($) { + my ($self) = @_; + while (_active($self)) { + check_async_wait($self); + cat_async_wait($self); + } } # returns true if there are pending "git cat-file" processes sub cleanup { - my ($self) = @_; - _destroy($self, qw(in out pid)); - _destroy($self, qw(in_c out_c pid_c err_c)); - !!($self->{pid} || $self->{pid_c}); + my ($self, $lazy) = @_; + ($lazy && _active($self)) and + return $self->{epwatch} ? watch_async($self) : 1; + local $in_cleanup = 1; + async_wait_all($self); + $_->close for ($self, (delete($self->{ck}) // ())); + undef; } # assuming a well-maintained repo, this should be a somewhat @@ -274,85 +488,194 @@ sub packed_bytes { my ($self) = @_; my $n = 0; my $pack_dir = git_path($self, 'objects/pack'); - foreach my $p (bsd_glob("$pack_dir/*.pack", GLOB_NOSORT)) { - $n += -s $p; - } + $n += (-s $_ // 0) for (bsd_glob("$pack_dir/*.pack", GLOB_NOSORT)); $n } -sub DESTROY { cleanup(@_) } +sub DESTROY { cleanup($_[0]) } sub local_nick ($) { - my ($self) = @_; - my $ret = '???'; # don't show full FS path, basename should be OK: - if ($self->{git_dir} =~ m!/([^/]+)(?:/\.git)?\z!) { - $ret = "/path/to/$1"; - } - wantarray ? ($ret) : $ret; + $_[0]->{nick} // ($_[0]->{git_dir} =~ m!/([^/]+?)(?:/*\.git/*)?\z! ? + "$1.git" : undef); } sub host_prefix_url ($$) { my ($env, $url) = @_; return $url if index($url, '//') >= 0; - my $scheme = $env->{'psgi.url_scheme'}; my $host_port = $env->{HTTP_HOST} // "$env->{SERVER_NAME}:$env->{SERVER_PORT}"; - "$scheme://$host_port". ($env->{SCRIPT_NAME} || '/') . $url; + my $sn = $env->{SCRIPT_NAME} // ''; + "$env->{'psgi.url_scheme'}://\L$host_port\E$sn/$url"; +} + +sub base_url { # for coderepos, PSGI-only + my ($self, $env) = @_; # env - PSGI env + my $nick = $self->{nick} // return undef; + my $url = host_prefix_url($env, ''); + # for mount in Plack::Builder + $url .= '/' if substr($url, -1, 1) ne '/'; + $url . $nick . '/'; } +sub isrch {} # TODO + sub pub_urls { my ($self, $env) = @_; if (my $urls = $self->{cgit_url}) { - return map { host_prefix_url($env, $_) } @$urls; + map { host_prefix_url($env, $_) } @$urls; + } else { + (base_url($self, $env) // '???'); } - local_nick($self); } sub cat_async_begin { my ($self) = @_; - cleanup($self) if alternates_changed($self); + cleanup($self) if $self->alternates_changed; + die 'BUG: already in async' if gcf_inflight($self); batch_prepare($self); - die 'BUG: already in async' if $self->{inflight}; - $self->{inflight} = []; } sub cat_async ($$$;$) { my ($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}); + my $inflight = gcf_inflight($self) // cat_async_begin($self); + substr($oid, 0, 0) = 'contents ' if $self->{-bc}; + write_all($self, $oid."\n", \&cat_async_step, $inflight); + push(@$inflight, $oid, $cb, $arg); +} + +# returns the modified time of a git repo, same as the "modified" field +# of a grokmirror manifest +sub modified ($;$) { + my $fh = $_[1] // popen($_[0], @MODIFIED_DATE); + (split(/ /, <$fh> // time))[0] + 0; # integerize for JSON +} + +sub cat_desc ($) { + my $desc = try_cat($_[0]); + chomp $desc; + utf8::decode($desc); + $desc =~ s/\s+/ /smg; + $desc eq '' ? undef : $desc; +} + +sub description { + cat_desc("$_[0]->{git_dir}/description") // 'Unnamed repository'; +} + +sub cloneurl { + my ($self, $env) = @_; + $self->{cloneurl} // do { + my @urls = split(/\s+/s, try_cat("$self->{git_dir}/cloneurl")); + scalar(@urls) ? ($self->{cloneurl} = \@urls) : undef; + } // [ substr(base_url($self, $env), 0, -1) ]; +} + +# for grokmirror, which doesn't read gitweb.description +# templates/hooks--update.sample and git-multimail in git.git +# only match "Unnamed repository", not the full contents of +# templates/this--description in git.git +sub manifest_entry { + my ($self, $epoch, $default_desc) = @_; + check_git_exe(); + my $gd = $self->{git_dir}; + my @git = ($GIT_EXE, "--git-dir=$gd"); + my $sr = popen_rd([@git, 'show-ref']); + my $own = popen_rd([@git, qw(config gitweb.owner)]); + my $mod = popen_rd([@git, @MODIFIED_DATE]); + my $buf = description($self); + if (defined $epoch && index($buf, 'Unnamed repository') == 0) { + $buf = "$default_desc [epoch $epoch]"; } + my $ent = { description => $buf, reference => undef }; + if (open(my $alt, '<', "$gd/objects/info/alternates")) { + # n.b.: GitPython doesn't seem to handle comments or C-quoted + # strings like native git does; and we don't for now, either. + local $/ = "\n"; + chomp(my @alt = <$alt>); - $self->{out}->print($oid, "\n") or fail($self, "write error: $!"); - push(@$inflight, [ $cb, $arg ]); + # grokmirror only supports 1 alternate for "reference", + if (scalar(@alt) == 1) { + $buf = File::Spec->rel2abs($alt[0], "$gd/objects"); + $buf =~ s!/[^/]+/?\z!!; # basename + $ent->{reference} = $buf; + } + } + $ent->{fingerprint} = sha_all(1, $sr)->hexdigest; + $sr->close or return; # empty, uninitialized git repo + $ent->{modified} = modified(undef, $mod); + chomp($buf = <$own> // ''); + utf8::decode($buf); + $ent->{owner} = $buf eq '' ? undef : $buf; + $ent; } -sub extract_cmt_time { - my ($bref, undef, undef, undef, $modified) = @_; +our $ck_unlinked_packs = $^O eq 'linux' ? sub { + # FIXME: port gcf2-like over to git.git so we won't need to + # deal with libgit2 + my $s = try_cat "/proc/$_[0]/maps"; + $s =~ /\.(?:idx|pack) \(deleted\)/s ? 1 : undef; +} : undef; - if ($$bref =~ /^committer .*?> ([0-9]+) [\+\-]?[0-9]+/sm) { - my $cmt_time = $1 + 0; - $$modified = $cmt_time if $cmt_time > $$modified; +# returns true if there are pending cat-file processes +sub cleanup_if_unlinked { + my ($self) = @_; + $ck_unlinked_packs or return cleanup($self, 1); + # Linux-specific /proc/$PID/maps access + # TODO: support this inside git.git + my $nr_live = 0; + for my $obj ($self, ($self->{ck} // ())) { + my $sock = $obj->{sock} // next; + my $pid = $sock->attached_pid // next; + $ck_unlinked_packs->($pid) and return cleanup($self, 1); + ++$nr_live; } + $nr_live; } -# returns the modified time of a git repo, same as the "modified" field -# of a grokmirror manifest -sub modified ($) { +sub event_step { my ($self) = @_; - my $modified = 0; - my $fh = popen($self, qw(rev-parse --branches)); - cat_async_begin($self); - local $/ = "\n"; - while (my $oid = <$fh>) { - chomp $oid; - cat_async($self, $oid, \&extract_cmt_time, \$modified); + my $inflight = gcf_inflight($self); + if ($inflight && @$inflight) { + $self->cat_async_step($inflight); + return $self->close unless $self->{sock}; + # don't loop here to keep things fair, but we must requeue + # if there's already-read data in pi_io_rbuf + $self->requeue if $self->{sock}->has_rbuf; + } +} + +sub schedule_cleanup { + my ($self) = @_; + PublicInbox::DS::add_uniq_timer($self+0, 30, \&cleanup, $self, 1); +} + +# idempotently registers with DS epoll/kqueue/select/poll +sub watch_async ($) { + my ($self) = @_; + schedule_cleanup($self); + $self->{epwatch} //= do { + $self->SUPER::new($self->{sock}, EPOLLIN); + \undef; } - cat_async_wait($self); - $modified || time; } +sub close { + my ($self) = @_; + my $sock = $self->{sock}; + delete @$self{qw(-bc err_c inflight)}; + delete($self->{epwatch}) ? $self->SUPER::close : delete($self->{sock}); + $sock->close if $sock; # calls gcf_drain via awaitpid +} + +package PublicInbox::GitCheck; # only for git <2.36 +use v5.12; +our @ISA = qw(PublicInbox::Git); +no warnings 'once'; + +# for event_step +*cat_async_step = \&PublicInbox::Git::check_async_step; + 1; __END__ =pod |