You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Karl Fogel <kf...@google.com> on 2006/09/19 22:49:35 UTC
Re: svn commit: r21546 - trunk/subversion/tests/libsvn_repos
philip@tigris.org writes:
> Log:
> * subversion/tests/libsvn_repos/repos-test.c (commit_editor_authz): Tweak
> pool usage to avoid a core dump with pool debugging and --enable-dso.
>
> Patch by: Malcolm Rowe <ma...@farside.org.uk>
How exactly does this prevent a core dump? Since 'subpool' is just a
child of 'pool', I can't see a safety issue with allocating in the
latter instead of the former. A scalability issue, sure, but safety?
How does the combination of pool debugging and --enable-dso cause such
a counterintuitive result?
-Karl
> Modified:
> trunk/subversion/tests/libsvn_repos/repos-test.c
>
> Modified: trunk/subversion/tests/libsvn_repos/repos-test.c
> URL: http://svn.collab.net/viewvc/svn/trunk/subversion/tests/libsvn_repos/repos-test.c?pathrev=21546&r1=21545&r2=21546
> ==============================================================================
> --- trunk/subversion/tests/libsvn_repos/repos-test.c (original)
> +++ trunk/subversion/tests/libsvn_repos/repos-test.c Mon Sep 18 16:38:39 2006
> @@ -1561,7 +1561,7 @@
> /* Open directory /A, to which we have read/write access. */
> SVN_ERR(editor->open_directory("/A", root_baton,
> SVN_INVALID_REVNUM,
> - pool, &dir_baton));
> + subpool, &dir_baton));
>
> /* Test denied file addition. Denied because of a conflicting rule
> on the file path itself. */
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: svn-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: svn-help@subversion.tigris.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21546 - trunk/subversion/tests/libsvn_repos
Posted by Philip Martin <ph...@codematters.co.uk>.
Karl Fogel <kf...@google.com> writes:
> Philip Martin <ph...@codematters.co.uk> writes:
>> Karl Fogel <kf...@google.com> writes:
>>> How exactly does this prevent a core dump? Since 'subpool' is just a
>>> child of 'pool', I can't see a safety issue with allocating in the
>>> latter instead of the former. A scalability issue, sure, but safety?
>>
>> See this thread
>>
>> http://svn.haxx.se/dev/archive-2006-03/1242.shtml
>
> Thanks. I edited the log message (sorry for the double propedit, I
> found your later message in the thread that explained it better).
I think the use of pool rather than subpool in the original code was
unintentional, although few people encountered the problem because
--enable-dso is not widely tested. I didn't commit the patch
immediately since I wasn't sure whether it was simply hiding an editor
bug, but yesterday I was going though some old patches in my working
copy and decided that unbreaking the test suite was reason enough to
commit it. However, I realise now I didn't reproduce the failure with
--enable-dso using a recent trunk build, I just ran regression tests
with the patch applied. I suspect that the new DSO pool handling has
made the patch redundant and that had I tried to reproduce the problem
I would not have been able to do so. The patch doesn't hurt but I
suspect it no longer really fixes a core dump.
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21546 - trunk/subversion/tests/libsvn_repos
Posted by Karl Fogel <kf...@google.com>.
Philip Martin <ph...@codematters.co.uk> writes:
> Karl Fogel <kf...@google.com> writes:
>> How exactly does this prevent a core dump? Since 'subpool' is just a
>> child of 'pool', I can't see a safety issue with allocating in the
>> latter instead of the former. A scalability issue, sure, but safety?
>
> See this thread
>
> http://svn.haxx.se/dev/archive-2006-03/1242.shtml
Thanks. I edited the log message (sorry for the double propedit, I
found your later message in the thread that explained it better).
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Re: svn commit: r21546 - trunk/subversion/tests/libsvn_repos
Posted by Philip Martin <ph...@codematters.co.uk>.
Karl Fogel <kf...@google.com> writes:
> philip@tigris.org writes:
>> Log:
>> * subversion/tests/libsvn_repos/repos-test.c (commit_editor_authz): Tweak
>> pool usage to avoid a core dump with pool debugging and --enable-dso.
>>
>> Patch by: Malcolm Rowe <ma...@farside.org.uk>
>
> How exactly does this prevent a core dump? Since 'subpool' is just a
> child of 'pool', I can't see a safety issue with allocating in the
> latter instead of the former. A scalability issue, sure, but safety?
See this thread
http://svn.haxx.se/dev/archive-2006-03/1242.shtml
--
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org