| Date | Commit message (Collapse) |
|---|
|
The "-w" perlop always succeeds as root, so we need to check
st_mode for writability bits to detect directories we shouldn't
write to.
Reported-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
Link: https://public-inbox.org/meta/20210927124056.kj5okiefvs4ztk27@meerkat.local/
|
|
We still need Email::MIME to test against old revisions.
We'll also depend on the revision just prior to the
manifest.js.gz introduction to avoid loading Danga::Socket,
since it was getting loaded even with `plackup'.
Finally, we'll disable Inline::C usage with old Spawn.pm
since our old code included alloca.h, which is not
portable to FreeBSD.
|
|
There may still be pre-manifest.js.gz versions of
PublicInbox::WWW running and serving v2 inboxes.
While -clone and "add-external --mirror" were working, -fetch
was failing due to 301 redirect to $INBOX_URL/manifest.js.gz/
and not the expected 404. Update the code to deal with a JSON
decode error (from the 301) and ensure v2 epochs detection is
correct (and not using a shadowed variable).
|
|
This makes it easier for users to enable fetching on a
previously read-only epoch. Prior to this change, users were
required to delete manifest.js.gz in addition to adding the
writable bit. Now, they just have to "chmod +w $EPOCH_DIR".
|
|
There may still be pre-manifest.js.gz versions of PublicInbox::WWW.
running and serving v2 inboxes.
Since $INBOX_URL/manifest.js.gz was not understood, it was
assumed to be a Message-ID and 301-ed to
"$INBOX_URL/manifest.js.gz/" with a trailing slash, so our 404
checks were invalid. Update our fallbacks to deal with 301
by catching JSON decoding errors to trigger HTML scraping.
For HTML parsing, be sure to not be fooled by potential
user-generated content and only scan the part after the last
<hr>.
We also need to avoid propagating $? from curl unnecessarily
when we can continue safely.
Finally, update v2mirror.t with tests to use PublicInbox::WWW
from our "v1.1.0-pre1" tag to ensure these code paths get tested
|
|
Partial (v2) clones should be useful addition for users wanting
to conserve storage while having fast access to recent messages.
Continuing work started in 876e74283ff3 (fetch: ignore
non-writable epoch dirs, 2021-09-17), this creates bare,
read-only epoch git repos. These git repos have the remotes
pre-configured, but does not fetch any objects.
The goal is to allow users to set the writable bit on a
previously-skipped epoch and start fetching it.
Shell completion support may not be necessary given how short
the epoch ranges are, here.
Cc: Luis Chamberlain <mcgrof@kernel.org>
Link: https://public-inbox.org/meta/20210917002204.GA13112@dcvr/T/#u
|
|
This will eventually be useful for maintaing partial mirrors.
Keeping inline with the original public-inbox-fetch philosophy,
there are no additional config files to manage:
the user merely needs to remove write permissions to an $N.git
directory to prevent it from being updated.
Re-enabling updates just requires restoring write permission.
|
|
As noted in the new manpage entry, this is useful for avoiding
public-inbox-index invocations when there's nothing to update.
We use 127 to match "grok-pull", and also because it doesn't
conflict with any of the current curl(1) exit codes.
|
|
IMHO, this greatly improves code sharing and organization
between v2, extindex, and lei/store. Common git-related
logic for these is lightly-refactored and easier to reason
about.
The impetus for this big change was to ensure inboxes
created+managed by public-inbox-{clone,fetch} could have
alternates and configs setup properly without depending on
SQLite (via V2Writable). This change does that while
making old code shorter and better factored.
|
|
t/v2mirror.t and t/lei-mirror.t are now skipped when curl
is missing (instead of failing in appropriate places).
A bunch of which() checks are updated to use require_cmd
to avoid explicitly loading Spawn.
|
|
Timestamp comparisons only have 1 second granularity, which
isn't nearly enough for our test cases, and probably not for
real world use for "git send-email" bursts and fast SMTP
servers.
We'll continue to check modification times inside the manifest,
though, in case an extremely rare SHA-1 collision is found...
|
|
It was also totally broken by the change to use manifest.js.gz
for v1 :x
Fixes: ffb7fbda6869db4b ("fetch: use manifest.js.gz for v1")
|
|
Using create_inbox doesn't seem worth the trouble, here, at the
moment, but disabling fsync(2) gives a noticeable speedup on
my system even with an SSD.
|
|
IPv4 gets plenty of real-world coverage, and apparently there's
Debian buildd hosts which lack IPv4(*). So ensure everything
can work on IPv6 and not cause problems for odd setups.
(*) https://bugs.debian.org/979432
|
|
Using "make update-copyrights" after setting GNULIB_PATH in my
config.mak
|
|
Nearly all of the search uses in the production code rely on
a Xapian mset iterator being returned (instead of an array
of $smsg objects). So default to returning the mset and move
the burden of smsg array conversion into the test cases.
|
|
{over_ro} being a part of the Search object is a historical
oddity which will go away, soon. Lets start removing its use in
tests and rarely-used helper scripts.
|
|
This gives better page cache utilization for Xapian indexing on
slow storage by improving locality for random I/O activity on
the Xapian DB.
Instead of doing a single-pass to index both SQLite and Xapian;
this indexes them separately. The first pass is identical to
indexlevel=basic: it indexes both over.sqlite3 and msgmap.sqlite3.
Subsequent passes only operate on a single Xapian shard for
documents belonging to that shard. Given enough shards, each
individual shard can be made small enough to fit into the kernel
page cache and avoid HDD seeks for read activity.
Doing rough tests with a busy system with a 7200 RPM HDD with ext4,
full indexing of LKML (9 epochs) goes from ~80 hours (-j0) to
~30 hours (-j8) with 16GB RAM with 7 shards configured and fsync(2)
disabled (--no-sync) and `--batch-size=10m'.
|
|
On a powerful (by my standards) machine with 16GB RAM and an
7200 RPM HDD marketed for "enterprise" use, indexing a 8.1G (in
git) LKML snapshot from Sep 2019 did not finish after 7 days
with the default number (3) of Xapian shards (`--jobs=4') and
`--batch-size=10m'.
Indexing starts off fast, but progressively get slower as
contents of the inbox (including Xapian + SQLite DBs) could no
longer be cached by the kernel. Once the on-disk size
increased, HDD seek contention between the Xapian shard workers
slowed the process down to a crawl.
With a single shard, it still took around 3.5 days to index on
the HDD. That's not good, but it's far better than not
finishing after 7 days. So allow unfortunate HDD users to
easily specify a single shard on public-inbox-init.
For reference, a freshly TRIM-ed low-end TLC SSD on the SATA II
bus on the same machine indexes that same snapshot of LKML in
~7 hours with 3 shards and the same 10m batch size. In the past,
a higher-end consumer grade MLC SSDs on similar hardware indexed
a similarly sized-data set in ~4 hours.
|
|
InboxWritable should only set $v2w->{parallel} if the $parallel
flag is defined to 0 or 1. We want indexing a new inbox to
utilize SMP, just like --reindex.
-index once again allows -j0/--jobs=0 to force single-process
use, and we'll be ensuring that works in tests to maintain
performance on small systems.
Fixes: 61a2fff5b34a3e32 ("admin: move index_inbox over")
|
|
We no longer load or use Email::MIME outside of comparison
tests.
|
|
In normal mail paths, we can rely on MTAs being configured with
reasonable limits in the -watch and -mda mail injection paths.
However, the MTA is bypassed in a git-only delivery path, a BOFH
could inject a large message and DoS users attempting to mirror
a public-inbox.
This doesn't protect unindexed WWW interfaces from Email::MIME
memory explosions on v1 inboxes. Probably nobody cares about
unindexed WWW interfaces anymore, especially now that Xapian is
optional for indexing.
|
|
Barely noticeable on Linux, but this gives a 1-2% speedup
on a FreeBSD 11.3 VM and lets us use built-in redirects
rather than relying on /bin/sh.
|
|
We will occasionally see legit messages with zero lines,
be sure we index that count for NNTP clients.
I'm not sure about bytes being zero (aside from purged
messages), but we should've dealt with that earlier up
the stack.
|
|
I didn't wait until September to do it, this year!
|
|
This cuts down on lines of code in individual test cases and
fixes some misnamed error messages by using "$0" consistently.
This will also provide us with a method of swapping out
dependencies which provide equivalent functionality (e.g
"Xapian" SWIG can replace "Search::Xapian" XS bindings).
|
|
We want to be able to use run_script with *.t files, so
t/common.perl putting subs into the top-level "main" namespace
won't work. Instead, make it a module which uses Exporter
like other libraries.
|
|
These usages of file-local global variables make the *.t files
incompatible with run_script(). Instead, use anonymous subs,
"our", or pass the parameter as appropriate.
|
|
Newer versions of git enable the commit graph by default.
Since we blow away our temporary directories every test,
generating graphis is a waste and clutters stderr with
"Computing commit graph generation numbers" messages.
|
|
We'll also introduce a tmpdir() API to give tempdirs
consistent names.
|
|
We can shave several hundred milliseconds off tests which spawn
daemons by preloading and avoiding startup time for common
modules which are already loaded in the parent process.
This also gives ENV{TAIL} support to all tests which support
daemons which log to stdout/stderr.
|
|
We need to be careful and explicitly close FDs before doing
-index, since we can't rely on FD_CLOEXEC without execve(2)
syscalls.
|
|
Not taking advantage of faster run modes in run_script, yet
since some lifetime problems need to be sorted.
|
|
And explicitly test for respawning in t/httpd-corner.t
There's no need to have an extra entries in the process table
for most tests we run, since that's not what we're testing.
|
|
"mainrepo" ws a bad name and artifact from the early days when I
intended for there to be a "spamrepo" (now just the
ENV{PI_EMERGENCY} Maildir). With v2, "mainrepo" can be
especially confusing, since v2 needs at least two git
repositories (epoch + all.git) to function and we shouldn't
confuse users by having them point to a git repository for v2.
Much of our documentation already references "INBOX_DIR" for
command-line arguments, so use "inboxdir" as the
git-config(1)-friendly variant for that.
"mainrepo" remains supported indefinitely for compatibility.
Users may need to revert to old versions, or may be referring
to old documentation and must not be forced to change config
files to account for this change.
So if you're using "mainrepo" today, I do NOT recommend changing
it right away because other bugs can lurk.
Link: https://public-inbox.org/meta/874l0ice8v.fsf@alyssa.is/
|
|
|
|
IO::Socket:*->new options are verbose and we can save
a bunch of code by putting this into t/common.perl,
since the related spawn_listener stuff is already there.
|
|
And use it in manifest.js.
To ease maintaining mirrors with grokmirror(1), we can accept
a "git/" directory prefix before the epoch, and ".git" suffix
after the epoch number.
We maintain compatibility with "$INBOX/$EPOCH" cloning, of
course, and it's still easier-to-type on the command-line.
|
|
In retrospect, introducing V1Writable was unnecessary and
InboxWritable->importer is in a better position to abstract
away differences between v1 and v2 writers.
So teach InboxWritable to initialize inboxes and get rid
of V1Writable.
|
|
PublicInbox::DS works for every platform we we care about,
nowadays; so checking for it is a waste of time. Cleanup a
few POSIX and Socket imports while we're in the area.
|
|
These modules are unmaintained upstream at the moment, but I'll
be able to help with the intended maintainer once/if CPAN
ownership is transferred. OTOH, we've been waiting for that
transfer for several years, now...
Changes I intend to make:
* EPOLLEXCLUSIVE for Linux
* remove unused fields wasting memory
* kqueue bugfixes e.g. https://rt.cpan.org/Ticket/Display.html?id=116615
* accept4 support
And some lower priority experiments:
* switch to EV_ONESHOT / EPOLLONESHOT (incompatible changes)
* nginx-style buffering to tmpfile instead of string array
* sendfile off tmpfile buffers
* io_uring maybe?
|
|
This allows v1 tests to continue working on git 1.8.0 for
now. This allows git 2.1.4 packaged with Debian 8 ("jessie")
to run old tests, at least.
I suppose it's safe to drop Debian 7 ("wheezy") due to our
dependency on git 1.8.0 for "merge-base --is-ancestor".
Writing V2 repositories requires git 2.6 for "get-mark"
support, so mask out tests for older gits.
|
|
IPC::Run provides a nice simplification in several places; and
we already use it (optionally) on a lot of tests.
For the non-test code, we still rely on our vfork-capable
Inline::C stuff since real-world server processes can get large
enough to where vfork is an advantage. Maybe Perl5 can use
CLONE_VFORK somehow, one day:
https://rt.perl.org/Ticket/Display.html?id=128227
Ohg V'q engure cbeg choyvp-vaobk gb Ehol :C
|
|
The normal behavior is to prevent the deleted messages from
being indexed in the first place. However, when fetching
incrementally via git; public-inbox-index needs to account for
deleted files which were created outside of the most recent
fetch/reindexing window.
Reported-by: Eric W. Biederman <ebiederm@xmission.com>
|
|
For v2 repositories with multiple epochs, we must not forget
about earlier epochs in clones. Ensure we update the alternates
file with all known epochs up to the current one.
Reported-by: Eric W. Biederman <ebiederm@xmission.com>
https://public-inbox.org/meta/871scj2vzi.fsf@xmission.com/
|
|
Xapian v1.2.21..v1.2.24 (inclusive) use OFD locks but failed to
set the close-on-exec flag on those locks. So we must continue
to work around those old versions by ensuring Xapian file
descriptors aren't held any longer than necessary when in
long-running git processes.
Reported-by: Konstantin Ryabitsev <konstantin@linuxfoundation.org>
|
|
Xapian is size-intensive and SQLite is not strictly necessary for v1.
|
|
This is important for people running mirrors via "git fetch",
as they need to be kept up-to-date. Purging is also now
supported in mirrors.
The short-lived "--regenerate" option is gone and is now
implicitly enabled as a result. It's still cheap when
article number regeneration is unnecessary, as we track
the range for each git repository.
|