Re: subversion-1.8.0-rc2: diff_test 60 fails
From: Tobias Bading <tbading_at_web.de>
Date: Sat, 25 May 2013 15:18:28 +0200
Hi Branko & Philip,
thanks a lot for your hints. I've seen your mails only now on http://mail-archives.apache.org/ and won't have access to my GNU/Linux machine till Monday. Meanwhile I'm giving 1.8 RC2 a try on Mac OS Snow Leopard.
What do you mean by "sleep for timestamps disabled"? I didn't do anything special, just running "make check" or "./diff_tests.py [-v] 60" (on an ext4 filesystem).
You've probably hit the mark with timestamps. While performing more tests yesterday, I experienced another problem, but that one was nondeterministic. Every once in a while, I'd say in about 1 out of 5-10 attempts, "./diff_tests.py -v 60" failed to create revision 4. The output claimed that the "svn ci" command was executed, but the repository only contained 3 revisions. I inserted debug outputs into the svn shell script, right before the exec, to make sure that everything is in order and file 3449_spurious has the proper contents. Everything looked perfectly fine and yet the commit did nothing at all sometimes. No "Sending..." output, no output to stdout at all. The reason for this problem could be very similar. Maybe Subversion thinks that there's nothing to commit at all, if the timestamp of file 3449_spurious didn't change? (Maybe svn could print something like "(No changes need to be committed)" in such a case, like GNU Emacs when there's nothing to save ;-)
On Monday, I'll check if a "touch 3449_spurious" fixes the broken working copy.
PS: All tests in the green on Snow Leopard now :-).
This is an archived mail posted to the Subversion Users mailing list.