make test fails when built with g++4.2

[shinaoka]

Here are error messages from Jenkins. https://alpscore.ci.cloudbees.com/job/cthyb-alpscore/220/console

When the CT-HYB is built with g++ (GCC) 4.7.2, one of unit tests fails with SegFault. The test, "unittest_fu" is exactly the same one with that at https://github.com/shinaoka/fastupdate. To trace back the cause of this crash, I built the same test with g++4.7.4 on my Mac. But, the test passed successfully.

How can I locate the point of SegFault? Can I get more information from Jenkins?

[egull]

I honestly don't know how to get more information from jenkins. I guess we could run it under valgrind? Alex, do you know?

[galexv]

From the log of the latest test run (https://alpscore.ci.cloudbees.com/view/ApplicationCodes/job/cthyb-alpscore/ws/build/Testing/Temporary/LastTest.log)

 [----------] 3 tests from DeterminantMatrixTypedTest/0, where TypeParam = alps::fastupdate::DeterminantMatrix<std::complex<double>, OffDiagonalG0<std::complex<double> >, creator, annihilator>
[ RUN      ] DeterminantMatrixTypedTest/0.CombinedUpdateRemoveRowsCols
[       OK ] DeterminantMatrixTypedTest/0.CombinedUpdateRemoveRowsCols (0 ms)
[ RUN      ] DeterminantMatrixTypedTest/0.SeparatedUpdateRemoveRowsCols
<end of output>

it seems that the crash is after the first call to DeterminantMatrixTypedTest / SeparatedUpdateRemoveRowsCols

If necessary, I will be able to look at this early the next week (after i am done with removing Boost libs from ALPSCore).

[galexv]

Actually, one simple way would be to add cout<< statements through the code and also compile with Boost assertions turned on, and then refer to the test log --- or run the test separately, outside "make test". This may give some idea where the crash occurs.

[shinaoka]

Thank you. The log messages help! I ran the test with valgrind (on my Mac) but any memory problem has been found yet.