The patch is in this message from William Uther:
http://subversion.tigris.org/servlets/ReadMsg?list=dev&msgId=182328
In that message, he says:
As I posted before, there was a test failing on my machine, but the
failure wasn't being detected and so the test suite was PASSing.
I've attached two patches to try and fix this. The first is more
conservative, but it doesn't fix everything I'd like. It only
checks bad stderr output.
The second detects signals and non-zero exit status. As the
functions required for this don't exist on windows, this should
fall back to being the same as the first patch on windows. I
haven't tested this on windows.
This second patch reveals an error in one of the other tests:
log_tests.py 3 fails with this patch when run a second time. The
problem is that this test tries to create a repos without checking
if one is already there and gets an error:
> svn: File exists
> svn: creating Berkeley DB environment dir
`repositories/log_tests-3/db'
This wasn't detected previously. I'm working on fixing this test
and will post a patch for it soon. (The svnadmin test that I
posted a patch for earlier also fails if this patch is applied and
that test isn't fixed.)
I am very new to python, so this should be checked carefully.
He did post that other patch later; I will make a separate issue for
it. Probably that will be issue #883, but don't quote me :-).