[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: svn commit: r1591919 - in /subversion/trunk/subversion: include/ include/private/ libsvn_fs/ libsvn_fs_base/bdb/ libsvn_fs_fs/ libsvn_fs_x/ libsvn_ra_svn/ libsvn_subr/ svnserve/ tests/ tests/libsvn_fs_fs/

From: Stefan Fuhrmann <stefan.fuhrmann_at_wandisco.com>
Date: Tue, 6 May 2014 16:46:26 +0200

On Tue, May 6, 2014 at 12:18 AM, Stefan Fuhrmann <
stefan.fuhrmann_at_wandisco.com> wrote:

> On Mon, May 5, 2014 at 2:06 PM, Stefan Sperling <stsp_at_elego.de> wrote:
>
>> On Fri, May 02, 2014 at 02:07:40PM -0000, stefan2_at_apache.org wrote:
>> > Author: stefan2
>> > Date: Fri May 2 14:07:40 2014
>> > New Revision: 1591919
>> >
>> > URL: http://svn.apache.org/r1591919
>> > Log:
>> > APR mutexes don't support recursive locking on all platforms.
>> > As a result, trying to take out the same lock twice in the
>> > same thread will cause a lock up under e.g. Linux. This patch
>> > adds an option to svn_mutex__t that detects recursive locking
>> > attempts in most cases and returns a proper error.
>>
>> > * subversion/tests/libsvn_fs_fs/fs-fs-pack-test.c
>> > (never_reached,
>> > lock_again): Callback functions required by the new test.
>> > (recursive_locking): New test expecting an SVN_ERR_RECURSIVE_LOCK.
>> > (test_funcs): Register the new test.
>>
>> Hi Stefan,
>>
>> This new test fails on the bb-openbsd bot.
>>
>> http://ci.apache.org/builders/bb-openbsd/builds/12/steps/Test/logs/faillog-ra_local-fsfs
>>
>> Can you please look into this?
>>
>> Note that the bot is deliberately compiling APR and everything else
>> without support for threads. Perhaps the test or other code doesn't
>> take this possibility into account?
>>
>
> Since double locking attempts on the same file
> may be problem even without threading support,
> we should detect recursive mutex locks there
> as well. r1592640 may do the trick. Could not
> test it, though.
>

I see that it is still failing on the buildbot and I have no clue
as to why. When I hard-code APR_HAS_THREADS to 0,
the test still passes (using the alternative code path).

All I found was a consistency edge case which did not
effect the outcome of the tests. Fixed in r1592763.

-- Stefan^2.
Received on 2014-05-06 17:03:36 CEST

This is an archived mail posted to the Subversion Dev mailing list.