Oh, I fully understand when things break on alternate setups or platforms. If it passes on "ra_local" vs "ra_*" ... then yeah. Totally normal.
It is just that the bots are showing breakage so, so often lately. I just got a bit frustrated. Sorry. On Sun, Apr 24, 2011 at 16:02, Bert Huijben <b...@qqmail.nl> wrote: > > >> -----Original Message----- >> From: Greg Stein [mailto:gst...@gmail.com] >> Sent: zondag 24 april 2011 14:27 >> To: dev@subversion.apache.org >> Subject: Re: svn commit: r1096264 - in /subversion/trunk/subversion: >> include/svn_wc.h libsvn_client/diff.c libsvn_wc/deprecated.c >> libsvn_wc/diff.c >> >> This broke a ton of diff tests. Did you even run them before committing >> this? >> > >> Seriously, I'm really tired of all the buildbot breakage. > > Well... Buy me a dozen machines and some rackspace and maybe I can test > every commit on every configuration before I commit. (And maybe we can > release Subversion 1.7 in 2016 if everybody needs its own rack of > buildservers) > > I ran the test suite 3 times over ra local before committing this patch and > without a real problem. > > Just now I found the memory corruption that was specific to how ra_neon and > ra_serf don't keep their paths valid while the baton is. > > > The buildbots are there to help us find these problems. If we should never > break builds we really don't need them. > > Bert > >
