From: Eric Wong <e@80x24.org>
To: meta@public-inbox.org
Subject: [PATCH 87/95] lei_mirror: eliminate circular references
Date: Mon, 28 Nov 2022 05:32:24 +0000 [thread overview]
Message-ID: <20221128053232.291618-88-e@80x24.org> (raw)
In-Reply-To: <20221128053232.291618-1-e@80x24.org>
...by using local-ized globals. While non-globals could work,
eliminating the {todo} and {fgrp_todo} refs in all sub-refs
is more error-prone and the `local' construct is convenient.
This allows us to get rid of the `delete $fgrp->{-fini}' call
in pack_refs and eliminates the indiscriminate reaping of all
processes before calling fgrp_fetch_all. This means we can
fully depend on DESTROY to provide predictable dependency
handling while supporting parallelization.
Global $TODO and $FGRP_TODO now become SCALAR refs on
consumption so they can act as assertions to detect future bugs.
---
lib/PublicInbox/LeiMirror.pm | 43 ++++++++++++++++++++----------------
1 file changed, 24 insertions(+), 19 deletions(-)
diff --git a/lib/PublicInbox/LeiMirror.pm b/lib/PublicInbox/LeiMirror.pm
index 47db9ccd..b30cc519 100644
--- a/lib/PublicInbox/LeiMirror.pm
+++ b/lib/PublicInbox/LeiMirror.pm
@@ -21,6 +21,8 @@ use PublicInbox::OnDestroy;
use Digest::SHA qw(sha256_hex sha1_hex);
our $LIVE; # pid => callback
+our $FGRP_TODO; # objstore -> [ fgrp mirror objects ]
+our $TODO; # reference => [ non-fgrp mirror objects ]
sub keep_going ($) {
$LIVE && (!$_[0]->{lei}->{child_error} ||
@@ -324,7 +326,6 @@ sub fgrp_update {
sub pack_dst { # packs lightweight satellite repos
my ($fgrp) = @_;
pack_refs($fgrp, $fgrp->{cur_dst});
- delete($fgrp->{-fini}) // die 'BUG: no {-fini}'; # call v1_done
}
sub pack_refs {
@@ -362,7 +363,8 @@ sub fgrpv_done {
sub fgrp_fetch_all {
my ($self) = @_;
- my $todo = delete $self->{fgrp_todo} or return;
+ my $todo = $FGRP_TODO;
+ $FGRP_TODO = \'BUG on further use';
keys(%$todo) or return;
# Rely on the fgrptmp remote groups in the config file rather
@@ -510,7 +512,7 @@ sub resume_fetch {
}
sub fgrp_enqueue {
- my ($fgrp) = @_;
+ my ($fgrp, $end) = @_; # $end calls fgrp_fetch_all
return if !keep_going($fgrp);
my $opt = { 2 => $fgrp->{lei}->{2} };
# --no-tags is required to avoid conflicts
@@ -523,12 +525,11 @@ sub fgrp_enqueue {
$fgrp->{dry_run} ? $fgrp->{lei}->qerr("# @cmd @kv") :
run_die([@cmd, @kv], undef, $opt);
}
- $fgrp->{fgrp_todo} // die 'BUG: no fgrp_todo';
- push @{$fgrp->{fgrp_todo}->{$fgrp->{-osdir}}}, $fgrp;
+ push @{$FGRP_TODO->{$fgrp->{-osdir}}}, $fgrp;
}
sub clone_v1 {
- my ($self, $nohang) = @_;
+ my ($self, $end) = @_;
my $lei = $self->{lei};
my $curl = $self->{curl} //= PublicInbox::LeiCurl->new($lei) or return;
my $uri = URI->new($self->{cur_src} // $self->{src});
@@ -540,7 +541,8 @@ sub clone_v1 {
my $resume = -d $dst;
if (my $fgrp = forkgroup_prep($self, $uri)) {
$fgrp->{-fini} = $fini;
- $resume ? cmp_fp_do($fgrp, \&fgrp_enqueue) : fgrp_enqueue($fgrp)
+ $resume ? cmp_fp_do($fgrp, \&fgrp_enqueue, $end)
+ : fgrp_enqueue($fgrp, $end);
} elsif ($resume) {
cmp_fp_do($self, \&resume_fetch, $uri, $fini);
} else { # normal clone
@@ -562,10 +564,10 @@ sub clone_v1 {
my $d = $self->{-ent} ? $self->{-ent}->{description} : undef;
$self->{'txt.description'} = $d if defined $d;
- (!defined($d) && !$nohang) and
+ (!defined($d) && !$end) and
_get_txt_start($self, 'description', $fini);
- $nohang or do_reap($self, 1); # for non-manifest clone
+ $end or do_reap($self, 1); # for non-manifest clone
}
sub parse_epochs ($$) {
@@ -785,7 +787,6 @@ sub clone_v2_prep ($$;$) {
my $dst = $self->{cur_dst} // $self->{dst};
my $want = parse_epochs($lei->{opt}->{epoch}, $v2_epochs);
my $task = $m ? bless { %$self }, __PACKAGE__ : $self;
- delete $task->{todo}; # $self->{todo} still exists
my (@skip, $desc);
my $fini = PublicInbox::OnDestroy->new($$, \&v2_done, $task);
for my $nr (sort { $a <=> $b } keys %$v2_epochs) {
@@ -813,7 +814,7 @@ failed to extract epoch number from $src
$etask->{cur_dst} = $edst;
$etask->{-is_epoch} = $fini;
my $ref = $ent->{reference} // '';
- push @{$self->{todo}->{$ref}}, $etask;
+ push @{$TODO->{$ref}}, $etask;
$self->{any_want}->{$key} = 1;
} else { # create a placeholder so users only need to chmod +w
init_placeholder($src, $edst, $ent);
@@ -917,7 +918,9 @@ sub multi_inbox ($$$) {
sub clone_all {
my ($self, $m) = @_;
- my $todo = delete $self->{todo};
+ my $todo = $TODO;
+ $TODO = \'BUG on further use';
+ my $end = PublicInbox::OnDestroy->new($$, \&fgrp_fetch_all, $self);
{
my $nodep = delete $todo->{''};
@@ -929,7 +932,7 @@ sub clone_all {
# handle no-dependency repos, first
for (@$nodep) {
- clone_v1($_, 1);
+ clone_v1($_, $end);
return if !keep_going($self);
}
}
@@ -947,14 +950,16 @@ EOM
}
my $y = delete $todo->{$x} // next; # already done
for (@$y) {
- clone_v1($_, 1);
+ clone_v1($_, $end);
return if !keep_going($self);
}
last; # restart %$todo iteration
}
}
- do_reap($self, 1); # finish all fingerprint checks
- fgrp_fetch_all($self);
+
+ # $end->DESTROY will call fgrp_fetch_all once all references
+ # in $LIVE are gone, and do_reap will eventually drain $LIVE
+ $end = undef;
do_reap($self, 1);
}
@@ -1007,8 +1012,6 @@ sub try_manifest {
my ($path_pfx, $n, $multi) = multi_inbox($self, \$path, $m);
return $lei->child_error(1, $multi) if !ref($multi);
my $v2 = delete $multi->{v2};
- local $self->{todo} = {};
- local $self->{fgrp_todo} = {}; # { objstore_dir => [fgrp, ...] }
if ($v2) {
for my $name (sort keys %$v2) {
my $epochs = delete $v2->{$name};
@@ -1054,7 +1057,7 @@ E: `$task->{cur_dst}' must not contain newline
EOM
$task->{cur_src} .= '/';
my $dep = $task->{-ent}->{reference} // '';
- push @{$self->{todo}->{$dep}}, $task; # for clone_all
+ push @{$TODO->{$dep}}, $task; # for clone_all
$self->{any_want}->{$name} = 1;
}
}
@@ -1112,6 +1115,8 @@ sub do_mirror { # via wq_io_do or public-inbox-clone
$self->{"-$k"} = $v;
}
local $LIVE = {};
+ local $TODO = {};
+ local $FGRP_TODO = {};
my $iv = $lei->{opt}->{'inbox-version'} //
return start_clone_url($self);
return clone_v1($self) if $iv == 1;
next prev parent reply other threads:[~2022-11-28 5:32 UTC|newest]
Thread overview: 96+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-28 5:30 [PATCH 00/95] clone: multi-inbox/repo support Eric Wong
2022-11-28 5:30 ` [PATCH 01/95] clone: support multi-inbox clone Eric Wong
2022-11-28 5:30 ` [PATCH 02/95] clone: support --include and --exclude with multi-clone Eric Wong
2022-11-28 5:31 ` [PATCH 03/95] clone: parallelize v2 epoch clones Eric Wong
2022-11-28 5:31 ` [PATCH 04/95] lei_mirror: async config retrieval for v2 w/ manifest Eric Wong
2022-11-28 5:31 ` [PATCH 05/95] lei_mirror: rely on DESTROY to index v2 inbox Eric Wong
2022-11-28 5:31 ` [PATCH 06/95] lei_mirror: rely on global process reaper Eric Wong
2022-11-28 5:31 ` [PATCH 07/95] clone: support parallel v1 clones Eric Wong
2022-11-28 5:31 ` [PATCH 08/95] lei_mirror: default to single job by default Eric Wong
2022-11-28 5:31 ` [PATCH 09/95] lei_mirror: move directory creation to v2-only path Eric Wong
2022-11-28 5:31 ` [PATCH 10/95] lei_mirror: retrieve description text asynchronously, too Eric Wong
2022-11-28 5:31 ` [PATCH 11/95] switch inotify/kevent stuff to v5.12 Eric Wong
2022-11-28 5:31 ` [PATCH 12/95] manifest: update module blurb + v5.12 Eric Wong
2022-11-28 5:31 ` [PATCH 13/95] lei_mirror: simplify _get_txt_start callers Eric Wong
2022-11-28 5:31 ` [PATCH 14/95] lei_mirror: elide description retrieval for v1|coderepo Eric Wong
2022-11-28 5:31 ` [PATCH 15/95] lei_mirror: add a hint for skipped epoch permissions Eric Wong
2022-11-28 5:31 ` [PATCH 16/95] lei_mirror: consolidate clone process management Eric Wong
2022-11-28 5:31 ` [PATCH 17/95] lei_mirror: load File::Path unconditionally Eric Wong
2022-11-28 5:31 ` [PATCH 18/95] lei_mirror: load most modules up-front Eric Wong
2022-11-28 5:31 ` [PATCH 19/95] lei_mirror: set gitweb.owner from manifest Eric Wong
2022-11-28 5:31 ` [PATCH 20/95] clone: support --dry-run / -n flag Eric Wong
2022-11-28 5:31 ` [PATCH 21/95] lei_mirror: initialize placeholders with "head" from manifest Eric Wong
2022-11-28 5:31 ` [PATCH 22/95] lei_mirror: support {reference} for v1 manifest clones Eric Wong
2022-11-28 5:31 ` [PATCH 23/95] lei_mirror: reduce noise on interrupted clones Eric Wong
2022-11-28 5:31 ` [PATCH 24/95] clone: support --inbox-config option Eric Wong
2022-11-28 5:31 ` [PATCH 25/95] lei_mirror: retrieve v2 description properly Eric Wong
2022-11-28 5:31 ` [PATCH 26/95] lei_mirror: reduce scope of v2 lock Eric Wong
2022-11-28 5:31 ` [PATCH 27/95] lei_mirror: allow --epoch on mixed v1/v2 clones Eric Wong
2022-11-28 5:31 ` [PATCH 28/95] lei_mirror: fix infinite loop in dependency resolution Eric Wong
2022-11-28 5:31 ` [PATCH 29/95] lei_mirror: defend against infinite loops Eric Wong
2022-11-28 5:31 ` [PATCH 30/95] lei_mirror: do not fetch descriptions if using manifest Eric Wong
2022-11-28 5:31 ` [PATCH 31/95] lei_mirror: require PublicInbox::Lock at use Eric Wong
2022-11-28 5:31 ` [PATCH 32/95] lei_mirror: fix glob semantics to match end-of-path Eric Wong
2022-11-28 5:31 ` [PATCH 33/95] lei_mirror: differentiate -entv vs -ent Eric Wong
2022-11-28 5:31 ` [PATCH 34/95] lei_mirror: support manifest {references} for v2 epochs Eric Wong
2022-11-28 5:31 ` [PATCH 35/95] lei_mirror: simplify v2 code paths Eric Wong
2022-11-28 5:31 ` [PATCH 36/95] clone: support --inbox-version Eric Wong
2022-11-28 5:31 ` [PATCH 37/95] lei_mirror: require Perl v5.12+ Eric Wong
2022-11-28 5:31 ` [PATCH 38/95] lei_mirror: ensure curl exits 22 on HTTP 404 responses Eric Wong
2022-11-28 5:31 ` [PATCH 39/95] lei_mirror: cleanup File::Temp OO usage Eric Wong
2022-11-28 5:31 ` [PATCH 40/95] lei_mirror: add `index' target to generated Makefile Eric Wong
2022-11-28 5:31 ` [PATCH 41/95] lei_mirror: do not write Makefile for --inbox-config=never Eric Wong
2022-11-28 5:31 ` [PATCH 42/95] lei_mirror: hoist out dump_manifest sub Eric Wong
2022-11-28 5:31 ` [PATCH 43/95] lei_mirror: avoid convoluted lazy_cb usage Eric Wong
2022-11-28 5:31 ` [PATCH 44/95] lei_mirror: simplify clone_v2_prep Eric Wong
2022-11-28 5:31 ` [PATCH 45/95] lei_mirror: support --objstore and forkgroups Eric Wong
2022-11-28 5:31 ` [PATCH 46/95] lei_mirror: cleanup process reaping logic Eric Wong
2022-11-28 5:31 ` [PATCH 47/95] lei_mirror: ensure git <1.8.5 fallback can use torsocks Eric Wong
2022-11-28 5:31 ` [PATCH 48/95] clone: flesh out --objstore behavior and document Eric Wong
2022-11-28 5:31 ` [PATCH 49/95] lei_mirror: always pack refs for coderepos Eric Wong
2022-11-28 5:31 ` [PATCH 50/95] lei_mirror: set description for non-inboxes, too Eric Wong
2022-11-28 5:31 ` [PATCH 51/95] lei_mirror: force --no-tags when fetching forkgroups Eric Wong
2022-11-28 5:31 ` [PATCH 52/95] lei_mirror: preserve permissions of existing alternates file Eric Wong
2022-11-28 5:31 ` [PATCH 53/95] lei_mirror: do not show ref updates w/o --verbose Eric Wong
2022-11-28 5:31 ` [PATCH 54/95] lei_mirror: drop git <1.8.5 support Eric Wong
2022-11-28 5:31 ` [PATCH 55/95] lei_mirror: make basename more descriptive Eric Wong
2022-11-28 5:31 ` [PATCH 56/95] lei_mirror: fix --dry-run for forkgroups Eric Wong
2022-11-28 5:31 ` [PATCH 57/95] lei_mirror: forkgroups use `git fetch --multiple' Eric Wong
2022-11-28 5:31 ` [PATCH 58/95] clone: move --dry-run handling to lei_mirror Eric Wong
2022-11-28 5:31 ` [PATCH 59/95] clone: drop unnecessary requires Eric Wong
2022-11-28 5:31 ` [PATCH 60/95] clone: use v5.12 Eric Wong
2022-11-28 5:31 ` [PATCH 61/95] clone: require `--objstore=' for default location Eric Wong
2022-11-28 5:31 ` [PATCH 62/95] lei_mirror: shorten remote names Eric Wong
2022-11-28 5:32 ` [PATCH 63/95] fetch: use v5.12 Eric Wong
2022-11-28 5:32 ` [PATCH 64/95] fetch: eliminate File::Temp->filename var Eric Wong
2022-11-28 5:32 ` [PATCH 65/95] lei_mirror: properly pack-refs in non-forkgroup repos Eric Wong
2022-11-28 5:32 ` [PATCH 66/95] lei_mirror: show child error error code Eric Wong
2022-11-28 5:32 ` [PATCH 67/95] on_destroy: support ->cancel callback Eric Wong
2022-11-28 5:32 ` [PATCH 68/95] lei_mirror: support resuming multi-repo clones Eric Wong
2022-11-28 5:32 ` [PATCH 69/95] lei_mirror: check fingerprints before fetching Eric Wong
2022-11-28 5:32 ` [PATCH 70/95] clone: support loading manifest.js.gz from destination Eric Wong
2022-11-28 5:32 ` [PATCH 71/95] lei_mirror: delay configuring forkgroups Eric Wong
2022-11-28 5:32 ` [PATCH 72/95] clone: canonicalize destination path from CLI Eric Wong
2022-11-28 5:32 ` [PATCH 73/95] clone|fetch: support passing --prune(-tags) to `git fetch' Eric Wong
2022-11-28 5:32 ` [PATCH 74/95] lei_mirror: avoid needless FD passing Eric Wong
2022-11-28 5:32 ` [PATCH 75/95] clone: support --keep-going/-k like make(1) Eric Wong
2022-11-28 5:32 ` [PATCH 76/95] lei_mirror: don't warn on missing manifest on initial clone Eric Wong
2022-11-28 5:32 ` [PATCH 77/95] lei_mirror: respect `./' and `../' prefixes for CLI args Eric Wong
2022-11-28 5:32 ` [PATCH 78/95] lei_mirror: --manifest= affects destination, too Eric Wong
2022-11-28 5:32 ` [PATCH 79/95] lei_mirror: update fingerprints when writing local manifest.js.gz Eric Wong
2022-11-28 5:32 ` [PATCH 80/95] lei_mirror: remove janky mirror.done stamp file Eric Wong
2022-11-28 5:32 ` [PATCH 81/95] lei_mirror: simplify most process spawning Eric Wong
2022-11-28 5:32 ` [PATCH 82/95] lei_mirror: run v1_done earlier on forkgroup done Eric Wong
2022-11-28 5:32 ` [PATCH 83/95] lei_mirror: simplify forkgroup-related subs Eric Wong
2022-11-28 5:32 ` [PATCH 84/95] lei_mirror: shorten scope mirror objects Eric Wong
2022-11-28 5:32 ` [PATCH 85/95] lei_mirror: set {head} from manifest Eric Wong
2022-11-28 5:32 ` [PATCH 86/95] lei_mirror: support {symlinks} " Eric Wong
2022-11-28 5:32 ` Eric Wong [this message]
2022-11-28 5:32 ` [PATCH 88/95] lei_mirror: use curl -z/--timecond if manifest exists Eric Wong
2022-11-28 5:32 ` [PATCH 89/95] lei_mirror: avoid redundant curl `-f' use Eric Wong
2022-11-28 5:32 ` [PATCH 90/95] lei_mirror: omit trailing slash for git remote.*.url Eric Wong
2022-11-28 5:32 ` [PATCH 91/95] lei_mirror: set info/web/last-modified from manifest Eric Wong
2022-11-28 5:32 ` [PATCH 92/95] lei_mirror: don't clobber inbox.config.example if it exists Eric Wong
2022-11-28 5:32 ` [PATCH 93/95] lei_mirror: break out of fgrp fetch iteration early Eric Wong
2022-11-28 5:32 ` [PATCH 94/95] clone: support --project-list= for cgit Eric Wong
2022-11-28 5:32 ` [PATCH 95/95] lei_mirror: handle forkgroup changes Eric Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
List information: https://public-inbox.org/README
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20221128053232.291618-88-e@80x24.org \
--to=e@80x24.org \
--cc=meta@public-inbox.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://80x24.org/public-inbox.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).