[Insight-developers] Dashboard Cleanup: RogueResearch6 : itkStatisticsAlgorithmTest : LLVM compiler

[Luis Ibanez]

Hi Sean,

More information on this topic:

A) Running Valgrind on the standard gcc build (not the llvm)
     did not reported any problems.

B) When building the REVIEW_STATISTICS with llvm,
     the test itkStatisticsAlgorithmTest passes fine (0.3 seconds),
     but the itkStatisticsAlgorithmTest2 gets trapped in an infinite loop.

C) Running (B) in gdb uses a single core (out of the two that are
     available), gets in the infinite loop, and ... when interrumpting it,
     we land as well in the InsertSort() method.


To summarize:

The problem is still there in the REVIEW_STATISTICS code.

(and "may" be something different from uninitialized variables,
or out-of-bounds access...         not 100% sure on this though...).


It looks like we will have to just walk through the steps of the
InsertSort() algorithm....



            Luis



----------------------------------------------------------
On Tue, Jan 19, 2010 at 7:47 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
> Hi Sean,
>
> Thanks a lot for these additional details.
>
> There are indeed very helpful.
>
> I have managed to reproduce the problem in my laptop.
>
>    *   Ubuntu 9.10 64 bits
>    *   Compiler LLVM-gcc-4.2  (64 bits) (1) below
>    *   Building for Debug (enabled in ccmake)
>
>
> (1) [ /usr/lib/llvm/llvm/gcc-4.2/bin/x86_64-linux-gnu-llvm-g++ ]
>
>
> When running, the test takes all the CPU, (dual core).
> Stays in an infinite loop,  and when interrupting it,
> it lands in itkStatisticsAlgorithmTest.cxx:150.
>
>                         InsertSort()
>
> An important  observation is that in RogueResearch6, the
> REVIEW flags is OFF, and so is the REVIEW_STATISTICS.
> That's how I build it also in my laptop.
>
> I'm now testing it with REVIEW_STATISTICS   ON
> (with the hope that the new code may not have this problem..)   :-)
>
>
> I'll keep tracking it..
>
>
>
>     Thanks
>
>
>
> ---------------------------------------------------------------------------
> On Tue, Jan 19, 2010 at 10:30 AM, Sean McBride <sean at rogue-research.com> wrote:
>> On 1/18/10 7:11 PM, Luis Ibanez said:
>>
>>>Anything special about this machine/platform
>>>that could give us a clue about the potential
>>>source of the problem ?
>>>
>>>I see that it is 64 bits,...(something else?)
>>
>> It's Debug (it works in Release).  And it uses the llvm-gcc compiler.
>>
>>>An interesting first thing to try would be to
>>>run this test in a single thread.
>>>
>>>You can do this from the command line by
>>>adding --without-threads to the command
>>>that executes the test.
>>
>> No change. :(
>>
>> Anyone know if this test passes with valgrind?  (I'd try, but valgrind
>> supports only 10.5, not 10.6).
>>
>> --
>> ____________________________________________________________
>> Sean McBride, B. Eng                 sean at rogue-research.com
>> Rogue Research                        www.rogue-research.com
>> Mac Software Developer              Montréal, Québec, Canada
>>
>>
>>
>