RE: ruby test failure, 1.6.6 roll on hold (was Re: 1.6.6 on Oct. 14)

Can't look until late tonight. I think the mime error is expected (still failing on trunk as I haven't adjusted the expectation). First i've heard of segfault in 1.6 tho.

--
Joe
From handheld.  Sorry for typos.
-----Original Message-----
From: Hyrum K. Wright <hyrum_at_hyrumwright.org>
Sent: Wednesday, October 14, 2009 11:54 AM
To: Mark Phippard <markphip_at_gmail.com>
Cc: Joe Swatosh <joe.swatosh_at_gmail.com>; dev_at_subversion.tigris.org
Subject: Re: ruby test failure, 1.6.6 roll on hold (was Re: 1.6.6 on Oct. 14)
On Oct 14, 2009, at 1:48 PM, Mark Phippard wrote:
> On Wed, Oct 14, 2009 at 2:32 PM, Hyrum K. Wright <hyrum_at_hyrumwright.org 
> > wrote:
>> On Oct 12, 2009, at 11:28 AM, Hyrum K. Wright wrote:
>>
>>> On Oct 7, 2009, at 7:03 AM, Hyrum K. Wright wrote:
>>>
>>>> On Sep 28, 2009, at 1:20 PM, Hyrum K. Wright wrote:
>>>>
>>>>> It's that time again: time for another great-tasting release of
>>>>> Subversion!  This one will be baked on Oct. 14 as Subversion  
>>>>> 1.6.6.
>>>>> Get your nominations and votes in, to help make this the best
>>>>> release
>>>>> yet.
>>>>>
>>>>> [ Note to translators, et. al.: I'm not sure what time on Oct. 14
>>>>> the
>>>>> 1.6.6 tarball will roll, but you can safely assume that it will be
>>>>> after 0000 UTC.  In other words, if your translations and other
>>>>> updates are on the branch by the end of the day on Oct. 13, you've
>>>>> nothing to fear about missing the deadline.  The same applies for
>>>>> votes and nominations in STATUS. ]
>>>>
>>>> T minus 6.5 days, and counting.
>>>
>>> One more reminder: I'll be rolling 1.6.6 sometime after 0000 UTC on
>>> Oct. 14.  It may be several hours later, but any changes approved in
>>> STATUS and translations on the 1.6.x branch are guaranteed to be in
>>> the release.
>>>
>>> Please nominate, review and vote on items in STATUS in the next  
>>> couple
>>> of days.
>>
>> Getting ready to roll the tarball, I ran 'make check-swig-rb' on the
>> 1.6.x branch.  I get the following segfault:
>>
>> test_merge(SvnClientTest): /home/hwright/dev/svn-1.6.x/subversion/
>> bindings/swig/ruby/test/my-assertions.rb:31: [BUG] Segmentation fault
>> ruby 1.8.7 (2008-08-11 patchlevel 72) [x86_64-linux]
>>
>> I'm planning on holding the roll until we determine the status of  
>> this
>> bug.
>
> I am not getting this problem.  But I did have a test failure.  I do
> not know that this same failure did not also happen with other 1.6.x
> releases:
>
> Started
> ..............................................................................................F 
> ..........................................................................................................................
> Finished in 535.915051 seconds.
>
>  1) Failure:
> test_mime_type_detect_with_type_map(SvnCoreTest)
> [/Users/mphippard/work/svn-1.6.x/subversion/bindings/swig/ruby/test/ 
> test_core.rb:631]:
> <Svn::Error::BadFilename> exception expected but none was thrown.
>
> 219 tests, 1483 assertions, 1 failures, 0 errors
Digging further, I see the following:
test_merge(SvnClientTest):
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fad06c586f0 (LWP 13356)]
0x00007fad0151cf46 in apr_hash_next (hi=0x1a01210) at tables/ 
apr_hash.c:128
128	        hi->this = hi->ht->array[hi->index++];
(gdb) backtrace
#0  0x00007fad0151cf46 in apr_hash_next (hi=0x1a01210) at tables/ 
apr_hash.c:128
#1  0x00007fad051449c9 in c2r_hash_with_key_convert (hash=0x1a01200,
     key_conv=0x7fad05142edc <c2r_string>, key_ctx=0x0,
     value_conv=0x7fad05142c0f <svn_swig_rb_from_swig_type>,
     value_ctx=0x7fad053a5348)
     at subversion/bindings/swig/ruby/libsvn_swig_ruby/swigutil_rb.c: 
1409
#2  0x00007fad05144a7c in c2r_hash (hash=0x1a01200,
     value_conv=0x7fad05142c0f <svn_swig_rb_from_swig_type>,  
ctx=0x7fad053a5348)
     at subversion/bindings/swig/ruby/libsvn_swig_ruby/swigutil_rb.c: 
1429
#3  0x00007fad05144ae8 in svn_swig_rb_apr_hash_to_hash_swig_type  
(hash=0x1a01200,
     type_name=0x7fad053a5348 "svn_log_changed_path_t *")
     at subversion/bindings/swig/ruby/libsvn_swig_ruby/swigutil_rb.c: 
1447
#4  0x00007fad05364a00 in _wrap_svn_log_entry_t_changed_paths_get  
(argc=0,
     argv=0x0, self=140381006173800) at subversion/bindings/swig/ruby/ 
core.c:5171
#5  0x00007fad06791c40 in rb_call0 (klass=140381119398840,  
recv=140381006173800,
     id=16729, oid=16729, argc=0, argv=0x0, body=0x7fad06c131e8,
     flags=<value optimized out>) at eval.c:5904
#6  0x00007fad06791dfa in rb_call (klass=140381119398840,  
recv=140381006173800,
     mid=16729, argc=0, argv=0x0, scope=0, self=140381005113880) at  
eval.c:6151
#7  0x00007fad0678bcf0 in rb_eval (self=140381005113880, n=<value  
optimized out>)
     at eval.c:3492
...
I can't reproduce this when running the test singly, though.
If others can't reproduce, I'll consider it an anomaly and continue to  
roll.
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=2407713