test-verify.t failure after 8d44b5a2974f

[Adrian Buehlmann]

On 2012-05-07 11:48, Adrian Buehlmann wrote:
> On 2012-05-07 02:00, Mads Kiilerich wrote:
>> I have been trying to make brendan's windows buildbut pass cleanly. But 
>> starting with "the" 2.2.1 changeset test-verify.t has started to fail.
>>
>> On line 76 of test-verify.t "hg verify" will pause and show a message 
>> box saying "python.exe has stopped working" and APPCRASH.
>>
>> I have no clue what the exact problem is - but I also haven't dived into 
>> the code. A quick linux valgrind shows no problem on that 'hg verify'.
>>
>> I assume it can be reproduced by following 
>> http://mercurial.selenic.com/wiki/HackableMercurial#Running_the_test_suite_under_MSYS 
>> - or Brendan can give access to the buildbut where there is a 
>> w2k8/hg/verifyfail.cmd that can reproduce the problem.
> 
> Manually exercising that specific testcase (revlog corruption) using 91323a78aac2 runs fine on Windows 7 x64 here using an x64 Python (2.6.6) and Mercurial's C extensions compiled with Microsoft's C compiler for x64 target (build log excerpt [1]).
> 
> The test output is the same as test-verify.t expects it on line 77.

I just tried the same with x86 variant (32-bit): works as expected as well.