Ping. This thread has received no new comments.
On 28/01/2011, at 6:40 PM, Noorul Islam K M wrote:
> Daniel Shahaf <d...@daniel.shahaf.name> writes:
>
>> Noorul Islam K M wrote on Fri, Jan 28, 2011 at 12:27:46 +0530:
>>
>>> Daniel Shahaf <d...@daniel.shahaf.name> writes:
>>>
>>>> Looks good, but I have a question:
>>>>
>>>> Noorul Islam K M wrote on Wed, Jan 26, 2011 at 13:12:54 +0530:
>>>>>
>>>>> Attached is the python test for issue 3013. This incorporates the steps
>>>>> from the shell script attached in the tracker.
>>>>>
>>>>> Log
>>>>> [[[
>>>>>
>>>>> New XFail test for issue 3013.
>>>>>
>>>>> * subversion/tests/cmdline/update_tests.py
>>>>> (update_after_switching_to_deleted_path, test_list): New XFail test
>>>>>
>>>>> Patch by: Noorul Islam K M <noorul{_AT_}collab.net>
>>>>> ]]]
>>>>>
>>>>> Thanks and Regards
>>>>> Noorul
>>>>>
>>>>
>>>>> Index: subversion/tests/cmdline/update_tests.py
>>>>> ===================================================================
>>>>> --- subversion/tests/cmdline/update_tests.py (revision 1063610)
>>>>> +++ subversion/tests/cmdline/update_tests.py (working copy)
>>>>> @@ -5347,6 +5347,34 @@
>>>>> svntest.main.run_svn(None, 'delete', os.path.join('A2', 'mu'))
>>>>> svntest.main.run_svn(None, 'update', os.path.join('A2', 'mu'))
>>>>>
>>>>> +### regression test for issue #3013
>>>>> +def update_after_switching_to_deleted_path(sbox):
>>>>> + "update after switching to deleted path"
>>>>> +
>>>>> + sbox.build()
>>>>> + wc_dir = sbox.wc_dir
>>>>> + repo_url = sbox.repo_url
>>>>> +
>>>>> + # switch to A/B
>>>>> + svntest.actions.run_and_verify_svn2(None, None, [], 0, 'switch',
>>>>> + repo_url + "/A/B", wc_dir)
>>>>> +
>>>>> + # delete A/D
>>>>> + svntest.actions.run_and_verify_svn2(None, None, [], 0, 'rm',
>>>>> + repo_url + "/A/D", '-m',
>>>>> + 'Remove A/D')
>>>>> +
>>>>> + # switch to A/D and this is known to fail
>>>>> + svntest.actions.run_and_verify_svn2(None, None,
>>>>> svntest.verify.AnyOutput,
>>>>> + 1, 'switch', repo_url + "/A/D",
>>>>> wc_dir)
>>>>> +
>>>>> + # switch to A/D@1 and this is known to succeed
>>>>> + svntest.actions.run_and_verify_svn2(None, None, [], 0, 'switch',
>>>>> + repo_url + "/A/D@1", wc_dir)
>>>>> +
>>>>> + # update should succeed
>>>>> + svntest.actions.run_and_verify_svn2(None, None, [], 0, "up", wc_dir)
>>>>> +
>>>>
>>>> Should this 'update' succeed? In my testing, updating the wc root to
>>>> a revision it does not exist in fails.
>>>
>>> "/A/D" @ revision 1 does exist. So update should succeed but it is
>>> failing. This is the issue.
>>>
>>
>> I think the issue is that the 'switch' is failing, nothing about the
>> update.
>>
>> But, in fact: both switch and update work as long as the directory
>> they're targetting is a wc subdir (as opposed to a wc root). Therefore,
>> I'm inclined to mark the issue as FIXED and adjust the test to test that
>> the switch works when the being-switched directory is not the wc root.
>>
>> Thoughts?
>>
>
> Pasting our conversation on IRC here.
>
> <noorul> danielsh_: The switch is known fail @HEAD and not @1 [12:48]
> <noorul> A/D is non-existent @HEAD and existent @1 hence first switch is known
> to fail and the second one to succeed
> <noorul> When the second one succeeds the subsequent update should update WC
> with repo_url@1 [12:49]
> <danielsh_> noorul: see mail, I think it's a different issue,
> <danielsh_> the switch actually works when you do 'svn sw $URL
> /path/to/wc/rootdir/some/subdir'
> <noorul> I think I correctly converted
>
> http://subversion.tigris.org/nonav/issues/showattachment.cgi/1157/3013.sh
> into a test [12:52]
> <danielsh_> yes
> <danielsh_> I think the 'update' shouldn't have been in the .sh script either
> btw [12:53]
> <danielsh_> but that's history
> <noorul> Before svn update if we use 'svn info' we could see that it is
> pointing to repo_url@1 [12:54]
> <noorul> Shouldn't svn up bring wc in sync with that revision?
> <noorul> A/D existing @1 the error message is misleading [12:55]
> <danielsh_> currently the 'up' fails iff the
> dir-being-updated-past-its-deletion is the wc root
> <danielsh_> I think that's sensible,
> <danielsh_> to error if the wcroot would have to be deleted,
> <danielsh_> but feel free to discuss that on dev@
> <danielsh_> re error messages: haven't checked whether they're misleading or
> not
> <danielsh_> haven't read them actually, just checked if the up succeeded or
> not and what rev it put me at
> <noorul> danielsh_: Do you mean to say that "svn up A/D" would succeed instead
> of "svn up ." [13:03]
> <danielsh_> noorul: svn rm ^/A/D; svn up A/D/; <-- succeeds [13:04]
> <danielsh_> noorul: svn co ^/A foo; cd foo; svn rm ^/A; svn up; <-- fails
> <danielsh_> that's current behaviour [13:05]
> <noorul> danielsh_: With respect to the test case [13:06]
> <noorul> I mean the test case that I submitted, do you think that svn up
> should succeed?
> <danielsh_> noorul: I think the 'svn up' doesn't belong in that test case, I
> think that issue is only about the 'switch'.
> <danielsh_> (that's my opinion; other devs may disagree with me)
>
> Thanks and Regards
> Noorul
