On Thu, Apr 1, 2010 at 6:07 PM, Paul Burba <ptburba_at_gmail.com> wrote:
> On Thu, Apr 1, 2010 at 9:50 AM, Philip Martin
> <philip.martin_at_wandisco.com> wrote:
>> Philip Martin <philip.martin_at_wandisco.com> writes:
>>
>>> Paul Burba <ptburba_at_gmail.com> writes:
>>>
>>>>> My bad, I didn't pass --url! I was still half asleep when I did that.
>>>>> The tests pass for me over ra_neon, but all except 77 fail over
>>>>> ra_serf.
>>>>
>>>> All 5 corresponding tests (45, 76, 78, 123, 125 for those playing at
>>>> home) pass on trunk [fsfs | ra_serf]
>>>
>>> Agreed. I get the same 4 failures as you, and they pass on trunk.
>>
>> This is the revision that causes the failures:
>>
>> r923779 | hwright | 2010-03-16 14:48:28 +0000 (Tue, 16 Mar 2010) | 16 lines
>>
>> Reintegrate the 1.6.x-issue-3242-partial branch:
>>
>> * ^/subversion/branches/1.6.x-issue-3242-partial
>> Merge a portion of the issue 3242 work, specifically the bit that
>> makes mergeinfo queries in the client layer happen against the
>> current RA session URL instead of forcibly against the repository
>> root URL (which might not be readable by the user per authz
>> policy). While this is not a full solution to issue #3242, it
>> should get us back to Subversion 1.4-level behavior which has
>> pretty consistently been deemed acceptable (enough) by our user
>> base.
>> Justification:
>> Issue #3242 is a hot issue. Hot topic, hot tempers, etc.
>> Votes:
>> +1: cmpilato, ivan, rhuijben
>
> This is also the revision that is causing the failure with
> basic_tests.py 37 "use folders with names like 'c:hi'".
>
> With 1.6.x-issue-3242-partial in place, basic_tests.py 37 now calls
> svn_client__open_ra_session_internal(), in
> libsvn_client/delete.c:delete_urls(), on the specific URL to be
> deleted:
>
> "file:///C%3A/SVN/src-branch-1.6.x/Debug/subversion/tests/cmdline/svn-test-work/repositories/basic_tests-37/c:hi"
>
> rather than the root of the reposisotry:
>
> "file:///C%3A/SVN/src-branch-1.6.x/Debug/subversion/tests/cmdline/svn-test-work/repositories/basic_tests-37"
>
> Note the trailing part now in the URL: "/c:hi". This ultimately
> causes svn_io_check_path() in libsvn_repos/repos.c:check_repos_path()
> to fail because it can't handle the ':'
>
> libsvn_repos-1.dll!check_repos_path() Line 1265 C
> libsvn_repos-1.dll!svn_repos_find_root_path() Line 1366 C
> libsvn_ra-1.dll!svn_ra_local__split_URL() Line 131 C
> libsvn_ra-1.dll!svn_ra_local__open() Line 454 C
> libsvn_ra-1.dll!svn_ra_open3() Line 481 C
> libsvn_client-1.dll!svn_client__open_ra_session_internal() Line 296 C
> libsvn_client-1.dll!delete_urls() Line 187 C
> libsvn_client-1.dll!svn_client_delete3() Line 270 C
> svn.exe!svn_cl__delete() Line 83 C
> svn.exe!main() Line 2119 C
> svn.exe!__tmainCRTStartup() Line 582 C
> svn.exe!mainCRTStartup() Line 399 C
>
> There are a slew of fixes in this space on trunk that will probably
> need to be backported,
>
> http://svn.apache.org/viewvc?view=revision&revision=920292
>
> and
>
> http://svn.apache.org/viewvc?view=revision&revision=876414
>
> look to be the first two, but so far all they do is move the problem around :-)
>
> I'll dig some more tomorrow.
>
> Bert - CCed you specifically since you've done a lot of work with the
> svn_[dirent|uri]_* APIs as well as the "c:hi"-type fixes, so if you
> think I'm on a fool's errand here (i.e. going to require 1.7 APIs to
> fix) please let me know!
Appears it was simpler than I first thought: The only change needed to
fix basic_tests.py 37 over ra_local on Windows is
http://svn.apache.org/viewvc?view=revision&revision=877014
Running the [ra_local x fsfs] tests in full right now. Assuming there
are no unexpected problems I'll nominate r877014 for backport when
they are finished.
Paul
Received on 2010-04-02 17:46:49 CEST