[CMake] test executable failing only when run in ctest

[Matthew Leotta]

Brad,

   Thanks, but I'm not sure if I completely understand your  
suggestion.  Let me clarify.  If I run the test executable

./bvxm_test_all

I get no errors and no segfault.  I know how to debug this, but there  
is nothing to debug.  If in the same directory I run

ctest

then the test will segfault.  Even stranger, on some platforms the  
test passes, but when I run in ctest it fails (produces the wrong  
values) but does not segfault.  On some other platforms everything  
seems to work both with and without ctest.

I'm actually running only the specific test that fails (the 5th one)  
with

./bvxm_test_all test_amp_processors

or

ctest -V -I 5,5

but the results are the same.  When I try to debug ctest I never hit  
breakpoints in bvxm_test_all because ctest spawns off a separate  
process for bvxm_test_all, and gdb is attached to ctest.  I'm not much  
of an expert on using gdb from the command line, I usually debug with  
an IDE front end (XCode or KDevelop).

--Matt


On Jan 26, 2009, at 11:49 AM, Brad King wrote:

> Matthew Leotta wrote:
>> I've come across a test on the VXL dashboard that seems to segfault  
>> only
>> on some test machines.
>>
>> http://www.cdash.org/CDash/testSummary.php?project=12&name=bvxm_test_apm_processors&date=2009-01-26
>>
>>
>> One of the failing machines is mine (Mac_OS_X-10.5.6_unix_make).   
>> When I
>> run the same test executable from the command line, all the tests  
>> pass
>> and there is no segfault.  Strangely, the segfault only occurs when  
>> the
>> executable is run by ctest.  I can reproduce the dashboard results by
>> running ctest in the executable's directory.  Has anyone seen  
>> anything
>> like this before?  Any suggestions on how to debug this?
>
> Hack the test command line to run gdb on the test executable.  Then  
> run
> the test inside it by hand.
>
> -Brad