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

Detecting SIGSEGVs with the test framework

From: Jani Averbach <jaa_at_jaa.iki.fi>
Date: 2005-06-02 20:11:22 CEST

On 2005-06-02 18:28+0100, Philip Martin wrote:
> Jani Averbach <jaa@jaa.iki.fi> writes:
> > On 2005-06-02 17:20+0100, Philip Martin wrote:
> >>
> >> Not really. A recent change caused 'svn unlock URL' to SEGV and the
> >> the test using it still passed. As it happens the SEGV occurred after
> >> the lock was removed, but if it had happened before the lock was
> >> removed the test would still have passed.
> >
> > I will try to include and use subprocess[1] with our test framework
> > and see if it makes any difference with 'returncode'. This is new in
> > Python 2.4 and it won't probably help with win32, so coverage won't be
> > 100%, but at least it might help a little.
> Are you proposing to detect SEGVs? This patch
> http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgNo=34876

The whole thread is here:

> was not applied because it didn't support win32.

Please propose it again, and here is my rationale why:

We have had several SIGSEV cases during last six months and almost all
the time they have been lurking a while in the code base until they
have dragged to sunlight.

If there is some über cool way to detect errors on win32, let's use
it! If there is some clever way to check correctness on OS X, same
here, use it. And if there is some way to detect SIGSEVs on UNIX,
please, use it. We have a portable code base, it is a strength, and
use that richness to comb out automatically by all possible means as
many bugs and defects as possible.

We can enforce our log format without any pre-commit-hook, and I am
pretty sure that we could enforce the old way to write non-sloppy test
cases. Every time when this SIGSEV detector found a case which could have
been detected by ordinary test setup, but which isn't, then
responsible party or parties could be flogged or dragged under keel.

BR, Jani

Jani Averbach
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Jun 2 20:12:10 2005

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

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.