We normally try to avoid having two auto-gc operations run
at the same time, because it wastes resources. This was done
long ago in 64a99eb47 (gc: reject if another gc is running,
unless --force is given, 2013-08-08).
When we do a detached auto-gc, we run the ref-related
commands _before_ detaching, to avoid confusing lock
contention. This was done by 62aad1849 (gc --auto: do not
lock refs in the background, 2014-05-25).
These two features do not interact well. The pre-detach
operations are run before we check the gc.pid lock, meaning
that on a busy repository we may run many of them
concurrently. Ideally we'd take the lock before spawning any
operations, and hold it for the duration of the program.
This is tricky, though, with the way the pid-file interacts
with the daemonize() process. Other processes will check
that the pid recorded in the pid-file still exists. But
detaching causes us to fork and continue running under a
new pid. So if we take the lock before detaching, the
pid-file will have a bogus pid in it. We'd have to go back
and update it with the new pid after detaching. We'd also
have to play some tricks with the tempfile subsystem to
tweak the "owner" field, so that the parent process does not
clean it up on exit, but the child process does.
Instead, we can do something a bit simpler: take the lock
only for the duration of the pre-detach work, then detach,
then take it again for the post-detach work. Technically,
this means that the post-detach lock could lose to another
process doing pre-detach work. But in the long run this
works out.
That second process would then follow-up by doing
post-detach work. Unless it was in turn blocked by a third
process doing pre-detach work, and so on. This could in
theory go on indefinitely, as the pre-detach work does not
repack, and so need_to_gc() will continue to trigger. But
in each round we are racing between the pre- and post-detach
locks. Eventually, one of the post-detach locks will win the
race and complete the full gc. So in the worst case, we may
racily repeat the pre-detach work, but we would never do so
simultaneously (it would happen via a sequence of serialized
race-wins).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The last test in 't6500-gc', 'background auto gc does not run if
gc.log is present and recent but does if it is old', added in
a831c06a2 (gc: ignore old gc.log files, 2017-02-10), may sporadically
trigger an error message from the test harness:
rm: cannot remove 'trash directory.t6500-gc/.git/objects': Directory not empty
The test in question ends with executing an auto gc in the backround,
which occasionally takes so long that it's still running when
'test_done' is about to remove the trash directory. This 'rm -rf
$trash' in the foreground might race with the detached auto gc to
create and delete files and directories, and gc might (re-)create a
path that 'rm' already visited and removed, triggering the above error
message when 'rm' attempts to remove its parent directory.
Commit bb05510e5 (t5510: run auto-gc in the foreground, 2016-05-01)
fixed the same problem in a different test script by simply
disallowing background gc. Unfortunately, what worked there is not
applicable here, because the purpose of this test is to check the
behavior of a detached auto gc.
Make sure that the test doesn't continue before the gc is finished in
the background with a clever bit of shell trickery:
- Open fd 9 in the shell, to be inherited by the background gc
process, because our daemonize() only closes the standard fds 0,
1 and 2.
- Duplicate this fd 9 to stdout.
- Read 'git gc's stdout, and thus fd 9, through a command
substitution. We don't actually care about gc's output, but this
construct has two useful properties:
- This read blocks until stdout or fd 9 are open. While stdout is
closed after the main gc process creates the background process
and exits, fd 9 remains open until the backround process exits.
- The variable assignment from the command substitution gets its
exit status from the command executed within the command
substitution, i.e. a failing main gc process will cause the test
to fail.
Note, that this fd trickery doesn't work on Windows, because due to
MSYS limitations the git process only inherits the standard fds 0, 1
and 2 from the shell. Luckily, it doesn't matter in this case,
because on Windows daemonize() is basically a noop, thus 'git gc
--auto' always runs in the foreground.
And since we can now continue the test reliably after the detached gc
finished, check that there is only a single packfile left at the end,
i.e. that the detached gc actually did what it was supposed to do.
Also add a comment at the end of the test script to warn developers of
future tests about this issue of long running detached gc processes.
Helped-by: Jeff King <peff@peff.net>
Helped-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A server can end up in a state where there are lots of unreferenced
loose objects (say, because many users are doing a bunch of rebasing
and pushing their rebased branches). Running "git gc --auto" in
this state would cause a gc.log file to be created, preventing
future auto gcs, causing pack files to pile up. Since many git
operations are O(n) in the number of pack files, this would lead to
poor performance.
Git should never get itself into a state where it refuses to do any
maintenance, just because at some point some piece of the maintenance
didn't make progress.
Teach Git to ignore gc.log files which are older than (by default)
one day old, which can be tweaked via the gc.logExpiry configuration
variable. That way, these pack files will get cleaned up, if
necessary, at least once per day. And operators who find a need for
more-frequent gcs can adjust gc.logExpiry to meet their needs.
There is also some cleanup: a successful manual gc, or a
warning-free auto gc with an old log file, will remove any old
gc.log files.
It might still happen that manual intervention is required
(e.g. because the repo is corrupt), but at the very least it won't
be because Git is too dumb to try again.
Signed-off-by: David Turner <dturner@twosigma.com>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When git gc --auto does an incremental repack of loose objects, we do
not expect to be able to write a bitmap; it is very likely that
objects in the new pack will have references to objects outside of the
pack. So we shouldn't try to write a bitmap, because doing so will
likely issue a warning.
This warning was making its way into gc.log. When the gc.log was
present, future auto gc runs would refuse to run.
Patch by Jeff King.
Bug report, test, and commit message by David Turner.
Signed-off-by: David Turner <dturner@twosigma.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is quite possible for, say, a remote HEAD to become broken, e.g.
when the default branch was renamed.
We should still be able to pack our objects when such a thing happens;
simply ignore broken symrefs (because they cannot matter for the packing
process anyway).
This fixes https://github.com/git-for-windows/git/issues/423
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This file isn't really harmful, but isn't useful either, and can create
minor annoyance for the user:
* It's confusing, as the presence of a *.pid file often implies that a
process is currently running. A user running "ls .git/" and finding
this file may incorrectly guess that a "git gc" is currently running.
* Leaving this file means that a "git gc" in an already gc-ed repo is
no-longer a no-op. A user running "git gc" in a set of repositories,
and then synchronizing this set (e.g. rsync -av, unison, ...) will see
all the gc.pid files as changed, which creates useless noise.
This patch unlinks the file after the garbage collection is done, so that
gc.pid is actually present only during execution.
Future versions of Git may want to use the information left in the gc.pid
file (e.g. for policies like "don't attempt to run a gc if one has
already been ran less than X hours ago"). If so, this patch can safely be
reverted. For now, let's not bother the users.
Explained-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Improved-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use the i18n-specific test functions in test scripts for parseopt tests.
This issue was was introduced in v1.7.10.1-488-g54e6d:
54e6d i18n: parseopt: lookup help and argument translations when showing usage
and been broken under GETTEXT_POISON=YesPlease since.
Signed-off-by: Jiang Xin <worldhello.net@gmail.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Given a request for command-line usage information rather than some
more substantial action, the only friendly thing to do is to report
the usage information as soon as possible and exit.
Without this change, as "git gc" glances over the repository, it can
be distracted by the desire to report a malformed configuration file.
Noticed while working through reports from Duy's repository access
checker.
[jn: with rewritten log message and tests]
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jeff King
SZEDER Gábor
David Turner
Johannes Schindelin
Jonathan Nieder
Jiang Xin
Nguyễn Thái Ngọc Duy