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

RE: make check failing on log_tests.py

From: Sander Striker <striker_at_apache.org>
Date: 2001-12-17 10:26:34 CET

Hi,

> From: Philip Martin [mailto:philip@codematters.co.uk]
> Sent: 17 December 2001 03:20
> Philip Martin <philip@codematters.co.uk> writes:
>
> > Hi
> >
> > 'make check' is failing while running log_tests.py, all the commits for
> > revisions 4 to 9 fail with 'Transaction is out of date'
> >
> > Does anyone else see this?
>
> After log_tests.py fails if I try
>
> $ cd subversion/tests/clients/cmdline
> $ svn ci working_copies/log_tests/
>
> I get the same "out of date" error. However if I use the
> electric-fence malloc/free debugger the commit succeeds:
>
> $ LD_PRELOAD=libefence.so.0.0 svn ci working_copies/log_tests/
>
> Now since the memory pools reuse malloc'd memory without free'ing it
> (which is perfectly acceptable) they defeat much of electric-fence's
> checking. However electric-fence will cause the memory to be allocated
> differently, in particular it will use more and spread it more
> widely. This behaviour is with pristine subversion code, no local
> changes.
>
> I conclude that the problem is not my configuration or build, but that
> there is a bug in the pool code, or a pool memory overwrite somewhere
> in the subversion code, and my setup just happens to trigger it.
>
> --
> Philip

Ofcourse we tried to reproduce this behaviour (wanting to rule out the
pools bug). I failed, Justin succeeded.

Justin encountered an interesting comment during investigation:

# log_tests.py: testing "svn log"
#
# ######################################################################
# ### ###
# ### YO! THIS FILE DOESN'T WORK YET. DON'T TRY TO RUN IT. THIS ###
# ### ISN'T A PROBLEM BECAUSE WE HAVEN'T ADDED IT TO THE TESTS LIST ###
# ### IN BUILD.CONF YET. ###
# ### ###
# ######################################################################

This probably is an out of date comment. If not, the
test slipped in unintentionally.

Anyhow, I updated to the latest APR and Subversion, autogen, configure,
make, make check and everything SUCCEEDs. The only significant difference
between Justins and my reproduction effort is that he has two copies of
subversion on his machine; the new source one and an old installed one
(different --prefix location). My machine only has _one_ copy of subversion.

I'm working on the pools debug version, so that efence will actually be
usefull. I'm a bit swamped till wednesday, but will try to get something
in soon.

Sander

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:53 2006

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