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

Re: Enabling svnserve log in svnserveautocheck

From: Stefan Fuhrmann <stefan.fuhrmann_at_wandisco.com>
Date: Wed, 14 Jan 2015 18:11:41 +0100

On Wed, Jan 14, 2015 at 5:52 PM, Stefan Sperling <stsp_at_elego.de> wrote:

> On Wed, Jan 14, 2015 at 05:45:30PM +0100, Stefan Fuhrmann wrote:
> > We are currently seeing test failures on the BSD bot
> > for ra_svn x bdb:
> >
> > http://ci.apache.org/builders/bb-openbsd/builds/275
> > http://ci.apache.org/builders/bb-openbsd/builds/267
> > (and probably more)
> >
> > It seems that the commit is failing on the server side
> > but, by default, there is no log available on the server.
> >
> > So, I'd like to commit a small patch that enables the
> > svnserve log in svnserveautocheck. My only concern
> > is that it adds ~11MB to the log file per test suite run.
> > The log location would be
> >
> > $ABS_BUILDDIR/subversion/tests/svnserveautocheck.log
> >
> > Would that be a problem for anyone running buildbots
> > on RAM disks?
> >
> > -- Stefan^2.
>
> The OpenBSD bot does not use davautocheck.
> Anything you commit there won't affect it.
>

Ok. So, could you add "--log-file /some/path/log" to the svnserve
command line and preserve the log afterwards?

> What part of the build logs is missing? I can try to provide
> any details you need.
>

The client-side logs current show something like

subversion/libsvn_ra_svn/editorp.c:418: (apr_err=SVN_ERR_COMPOSED_ERROR)
svn: E200042: Additional errors:
subversion/libsvn_ra_svn/editorp.c:418,
subversion/libsvn_ra_svn/marshal.c:1546,
:213,

which means that the server did not crash but returned
"failure" instead of "success". So, chances are that no
server sub-process got killed. That may still be random
chance, though.

> Do you want a backtrace from the crash? Unfortunately the binary
> has already been rebuilt so the current core dump is useless but
> I can try to get it at the right moment.
>

If the / a server process should crash, then a backtrace
would be helpful. Right now, it looks like the commit is
failing but it would be nice to narrow that down a bit first.

-- Stefan^2.
Received on 2015-01-14 18:12:55 CET

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