* --reindex buggy wrt NNTP numbers, fixes coming
@ 2019-05-25 19:57 Eric Wong
2019-05-27 18:45 ` [PATCH 0/3] fix --reindex skipping NNTP article numbers Eric Wong
0 siblings, 1 reply; 5+ messages in thread
From: Eric Wong @ 2019-05-25 19:57 UTC (permalink / raw)
To: meta
I was doing some cleanups and adding progress reporting to
V2Writable for public-inbox-index when I noticed NNTP article
numbers were off (too high, gaps)
And the v1 index_sync path seems buggy in this regard, too.
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 0/3] fix --reindex skipping NNTP article numbers
2019-05-25 19:57 --reindex buggy wrt NNTP numbers, fixes coming Eric Wong
@ 2019-05-27 18:45 ` Eric Wong
2019-05-27 18:45 ` [PATCH 1/3] t/v1reindex.t: fix typo in setting `indexlevel' Eric Wong
` (2 more replies)
0 siblings, 3 replies; 5+ messages in thread
From: Eric Wong @ 2019-05-27 18:45 UTC (permalink / raw)
To: meta
I was working on v2 cleanups and progress-reporting,
and speeding up no-op 'public-inbox-index' invocations when
I noticed --reindex showing higher-than-expected NNTP
article numbers with the verbose output I'm working on.
There's some (apparently) inconsequential typo fixes, too.
I think the v1 --reindex code is alright, after all, but
maybe I haven't been paying enough attention to it...
Also, I've noticed the v1 indexing code works around a Xapian
v1.2.20..v1.2.24 bug wrt OFD locks which the newer v2 code
doesn't (since v2 was developed mainly on Xapian 1.4.3+).
Not sure who uses Xapian 1.2, still; but it might still be
worth fixing for folks on older systems.
Eric Wong (3):
t/v1reindex.t: fix typo in setting `indexlevel'
searchidx: fix obvious typo
v2: fix reindex skipping NNTP article numbers
lib/PublicInbox/SearchIdx.pm | 2 +-
lib/PublicInbox/V2Writable.pm | 25 ++++++++++++++++++++++++-
t/indexlevels-mirror.t | 25 +++++++++++++++++++++++++
t/v1reindex.t | 2 +-
4 files changed, 51 insertions(+), 3 deletions(-)
--
EW
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH 1/3] t/v1reindex.t: fix typo in setting `indexlevel'
2019-05-27 18:45 ` [PATCH 0/3] fix --reindex skipping NNTP article numbers Eric Wong
@ 2019-05-27 18:45 ` Eric Wong
2019-05-27 18:45 ` [PATCH 2/3] searchidx: fix obvious typo Eric Wong
2019-05-27 18:45 ` [PATCH 3/3] v2: fix reindex skipping NNTP article numbers Eric Wong
2 siblings, 0 replies; 5+ messages in thread
From: Eric Wong @ 2019-05-27 18:45 UTC (permalink / raw)
To: meta
It did not cause a test failure because the default fallback
is `indexlevel=full'
---
t/v1reindex.t | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/t/v1reindex.t b/t/v1reindex.t
index 402ecd7..35275fb 100644
--- a/t/v1reindex.t
+++ b/t/v1reindex.t
@@ -223,7 +223,7 @@ ok(!-d $xap, 'Xapian directories removed again');
my @warn;
local $SIG{__WARN__} = sub { push @warn, @_ };
my %config = %$ibx_config;
- $config{indexleve} = 'medium';
+ $config{indexlevel} = 'medium';
my $ibx = PublicInbox::Inbox->new(\%config);
my $rw = PublicInbox::SearchIdx->new($ibx, 1);
eval { $rw->index_sync };
--
EW
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 2/3] searchidx: fix obvious typo
2019-05-27 18:45 ` [PATCH 0/3] fix --reindex skipping NNTP article numbers Eric Wong
2019-05-27 18:45 ` [PATCH 1/3] t/v1reindex.t: fix typo in setting `indexlevel' Eric Wong
@ 2019-05-27 18:45 ` Eric Wong
2019-05-27 18:45 ` [PATCH 3/3] v2: fix reindex skipping NNTP article numbers Eric Wong
2 siblings, 0 replies; 5+ messages in thread
From: Eric Wong @ 2019-05-27 18:45 UTC (permalink / raw)
To: meta
We can't pass an empty string to `git merge-base --is-ancestor'
AFAIK, this did NOT present issues in the current test suite.
---
lib/PublicInbox/SearchIdx.pm | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm
index 9c29106..b963805 100644
--- a/lib/PublicInbox/SearchIdx.pm
+++ b/lib/PublicInbox/SearchIdx.pm
@@ -694,7 +694,7 @@ sub _last_x_commit {
$lx = $lm;
}
# Use last_commit from msgmap if it is older or unset
- if (!$lm || ($lx && $lx && is_ancestor($self->{git}, $lm, $lx))) {
+ if (!$lm || ($lx && $lm && is_ancestor($self->{git}, $lm, $lx))) {
$lx = $lm;
}
$lx;
--
EW
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 3/3] v2: fix reindex skipping NNTP article numbers
2019-05-27 18:45 ` [PATCH 0/3] fix --reindex skipping NNTP article numbers Eric Wong
2019-05-27 18:45 ` [PATCH 1/3] t/v1reindex.t: fix typo in setting `indexlevel' Eric Wong
2019-05-27 18:45 ` [PATCH 2/3] searchidx: fix obvious typo Eric Wong
@ 2019-05-27 18:45 ` Eric Wong
2 siblings, 0 replies; 5+ messages in thread
From: Eric Wong @ 2019-05-27 18:45 UTC (permalink / raw)
To: meta
`public-inbox-index --reindex' could cause NNTP article number
gaps to form when it also has to deal with new,
never-before-seen commits in mirrors running off `git fetch'.
Fix this by running two distinct invocations of ->index_sync;
once to only reindex old commits, and a second time to index
new commits.
This does not appear to be a problem on v1 at the moment,
but I'll need more time to analyze this.
---
lib/PublicInbox/V2Writable.pm | 25 ++++++++++++++++++++++++-
t/indexlevels-mirror.t | 25 +++++++++++++++++++++++++
2 files changed, 49 insertions(+), 1 deletion(-)
diff --git a/lib/PublicInbox/V2Writable.pm b/lib/PublicInbox/V2Writable.pm
index cd08acd..331c4f4 100644
--- a/lib/PublicInbox/V2Writable.pm
+++ b/lib/PublicInbox/V2Writable.pm
@@ -850,11 +850,19 @@ sub index_prepare {
my $pr = $opts->{-progress};
my $regen_max = 0;
my $head = $self->{-inbox}->{ref_head} || 'refs/heads/master';
+
+ # reindex stops at the current heads and we later rerun index_sync
+ # without {reindex}
+ my $reindex_heads = last_commits($self, $epoch_max) if $opts->{reindex};
+
for (my $i = $epoch_max; $i >= 0; $i--) {
die 'BUG: already indexing!' if $self->{reindex_pipe};
my $git_dir = git_dir_n($self, $i);
-d $git_dir or next; # missing parts are fine
my $git = PublicInbox::Git->new($git_dir);
+ if ($reindex_heads) {
+ $head = $reindex_heads->[$i] or next;
+ }
chomp(my $tip = $git->qx(qw(rev-parse -q --verify), $head));
next if $?; # new repo
@@ -959,7 +967,14 @@ sub index_sync {
my $high = $self->{mm}->num_highwater();
my $regen = $self->index_prepare($opts, $epoch_max, $ranges);
- $$regen += $high if $high;
+ if ($opts->{reindex}) {
+ # reindex should NOT see new commits anymore, if we do,
+ # it's a problem and we need to notice it via die()
+ $$regen = -1;
+ } else {
+ $$regen += $high;
+ }
+
my $D = {}; # "$mid\0$cid" => $oid
my @cmd = qw(log --raw -r --pretty=tformat:%H
--no-notes --no-color --no-abbrev --no-renames);
@@ -1001,6 +1016,14 @@ sub index_sync {
$git->cleanup;
}
$self->done;
+
+ # reindex does not pick up new changes, so we rerun w/o it:
+ if ($opts->{reindex}) {
+ my %again = %$opts;
+ $mm_tmp = undef;
+ delete @again{qw(reindex -skip_lock)};
+ index_sync($self, \%again);
+ }
}
1;
diff --git a/t/indexlevels-mirror.t b/t/indexlevels-mirror.t
index ce138fe..1251136 100644
--- a/t/indexlevels-mirror.t
+++ b/t/indexlevels-mirror.t
@@ -105,9 +105,17 @@ sub import_index_incremental {
is_deeply([sort { $a cmp $b } map { $_->{mid} } @$msgs],
['m@1','m@2'], 'got both messages in master');
+ my @rw_nums = map { $_->{num} } @{$ibx->over->query_ts(0, 0)};
+ is_deeply(\@rw_nums, [1, 2], 'master has expected NNTP articles');
+
+ my @ro_nums = map { $_->{num} } @{$ro_mirror->over->query_ts(0, 0)};
+ is_deeply(\@ro_nums, [1, 2], 'mirror has expected NNTP articles');
+
# remove message from master
ok($im->remove($mime), '2nd message removed');
$im->done;
+ @rw_nums = map { $_->{num} } @{$ibx->over->query_ts(0, 0)};
+ is_deeply(\@rw_nums, [1], 'unindex NNTP article'.$v.$level);
if ($level ne 'basic') {
is(system(@xcpdb, $mirror), 0, "v$v xcpdb OK");
@@ -132,6 +140,23 @@ sub import_index_incremental {
($nr, $msgs) = $ro_mirror->search->reopen->query('m:m@2');
is($nr, 0, "v$v m\@2 gone from Xapian in mirror on $level");
}
+
+ # add another message to master and have the mirror
+ # sync and reindex it
+ my @expect = map { $_->{num} } @{$ibx->over->query_ts(0, 0)};
+ foreach my $i (3..5) {
+ $mime->header_set('Message-ID', "<m\@$i>");
+ ok($im->add($mime), "#$i message added");
+ push @expect, $i;
+ }
+ $im->done;
+ is(system('git', "--git-dir=$fetch_dir", qw(fetch -q)), 0, 'fetch OK');
+ is(system($index, '--reindex', $mirror), 0,
+ "v$v index --reindex mirror OK");
+ @ro_nums = map { $_->{num} } @{$ro_mirror->over->query_ts(0, 0)};
+ @rw_nums = map { $_->{num} } @{$ibx->over->query_ts(0, 0)};
+ is_deeply(\@rw_nums, \@expect, "v$v master has expected NNTP articles");
+ is_deeply(\@ro_nums, \@expect, "v$v mirror matches master articles");
}
# we can probably cull some other tests and put full/medium tests, here
--
EW
^ permalink raw reply related [flat|nested] 5+ messages in thread
end of thread, other threads:[~2019-05-27 18:45 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-05-25 19:57 --reindex buggy wrt NNTP numbers, fixes coming Eric Wong
2019-05-27 18:45 ` [PATCH 0/3] fix --reindex skipping NNTP article numbers Eric Wong
2019-05-27 18:45 ` [PATCH 1/3] t/v1reindex.t: fix typo in setting `indexlevel' Eric Wong
2019-05-27 18:45 ` [PATCH 2/3] searchidx: fix obvious typo Eric Wong
2019-05-27 18:45 ` [PATCH 3/3] v2: fix reindex skipping NNTP article numbers Eric Wong
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).