Lele Gaifax wrote:
> Chase Phillips wrote:
> > Hi Lele,
> >
> > I ran the test you suggested below and it failed for me. I used a
> > snapshot from today. The error message was:
> >
> > Test that we can handle resurrected cvs directory to svn. ...
15:15:25
> > [C] Could not import checked out tree in
> > "/tmp/tailor-tests/cvsresurdirtest/cvs2svn"!
> > ERROR
>
> Uhm, I cannot understand what's going on! On both machines I tested
> this, it worked as expected...
Early on I suspected the test wasn't functioning properly (failing too
early to actually be testing what we want). I have verified that was
the case. The problem that was keeping me from advancing was svn
1.1.1's lack of the --ignore-externals flag. (I compensated for this in
my own config file but Tailor uses a preset config for tests.)
I added "ignore-externals = False;" to vcpx/tests/tailor.py's DEFAULT
config section, updated the verbose line to True, and got all of the way
through to change "four," and I found that the test failed for me:
======================================================================
ERROR: Test that we can handle resurrected cvs directory to svn.
----------------------------------------------------------------------
Traceback (most recent call last):
File "/home/chase/tailor/20060828-trial1/app/vcpx/tests/tailor.py",
line 688, in testCvsReapperedDirectoryToSubversion
t()
File "/home/chase/tailor/20060828-trial1/app/vcpx/tailor.py", line
146, in __call__
self.update()
File "/home/chase/tailor/20060828-trial1/app/vcpx/tailor.py", line
109, in update
last, conflicts = dwd.applyPendingChangesets(
File "/home/chase/tailor/20060828-trial1/app/vcpx/dualwd.py", line 97,
in applyPendingChangesets
applied=applied)
File "/home/chase/tailor/20060828-trial1/app/vcpx/source.py", line
139, in applyPendingChangesets
replay(c)
File "/home/chase/tailor/20060828-trial1/app/vcpx/dualwd.py", line
107, in replayChangeset
self.target.replayChangeset(changeset)
File "/home/chase/tailor/20060828-trial1/app/vcpx/target.py", line
126, in replayChangeset
entries)
File "/home/chase/tailor/20060828-trial1/app/vcpx/repository/svn.py",
line 556, in _commit
raise ChangesetApplicationFailure("%s returned status %d saying\n%s"
ChangesetApplicationFailure:
/tmp/tailor-tests/cvsresurdirtest/cvs2svn/svnwd $ svn commit --file
/tmp/tailorQ4mt_Ssvn bar/again returned status 1 saying
svn: '/tmp/tailor-tests/cvsresurdirtest/cvs2svn/svnwd/bar' is not a
working copy
----------------------------------------------------------------------
Ran 1 test in 15.065s
FAILED (errors=1)
At change "two," bar and bar/baz are added and committed. At change
"four," bar/again is added, the commit is attempted, and the failure
occurs -- bar is not added in change "four" before bar/again.
> > If it helps to know this, I'm using a stock CVS 1.11.21 with Jeremy
> > Barnes's sleep_past() patch and svn 1.1.1.
>
> I have CVS 1.12.9, and svn 1.2.3, but I cannot see how this could
cause
> the test breakage... I'll try to investigate further.
I will, too.
PS In the test's change "four" the svn add of bar/again exits with code
1 but Tailor doesn't exit at that point. Tailor attempts the commit
which fails. The svn add's exit message is "svn: 'bar' is not a working
copy". Maybe Tailor should fail at svn add in this case?
Chase
_______________________________________________
Tailor mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/tailor
