Re: git svn clone segmentation faul issue

From: Ioannis Kappas <ioannis.kappas@rbs.com>
Date: Jun 29, 2016, 08:11 UTC
Message-ID: <loom.20160629T095219-589@post.gmane.org>
In-Reply-To: <20160628194804.GA3653@dcvr.yhbt.net>

Hi Eric,

Eric Wong <e <at> 80x24.org> writes:

Show 6 quoted lines

> 
> Ioannis.Kappas <at> rbs.com wrote:
> > Fortunately, a patch has already been submitted to subversion
> > with (github) revision
> > a074af86c8764404b28ce99d0bedcb668a321408 (at
> >

https://github.com/apache/subversion/commit/a074af86c8764404b28ce99d0bedcb66 8a321408

Show 8 quoted lines

> > ) on the trunk to handle this and a couple of other similar
> > cases. But by the looks of it has not been picked up yet in
> > the latest subversion 1.9.4 release or the 1.9.x branch,
> > perhaps because this patch was identified in sanity checks
> > rather than coming out from a perceivable production issue?
> 
> Thank you for documenting this.  Curious, does this affect older
> SVN versions or only 1.9.x?

It does not appear so, at least not in the degree of failures that I am seeing it now, i.e. failing consistently. It only came to my attention after I upgraded to the latest git/perl/subversion release from a version that was more than a year's old.

Given that the issue is with corruption in the perl interpreter stack, I suspect is due to one, or combination, of the following:

- git in latest releases makes additional perl subversion calls, thus 
affecting the perl stack usage.
- perl in latest releases affected the stack allocation profile.
- subversion perl module in latest releases refactored/added new calls that 
has changed the perl stack usage.

> 
> I don't know Perl internals well or SWIG at all; so reports
> like these are very much appreciated.
>

Thanks