OpenBSD/Perl 5.12.x/5.14.x module random test failures (Error 255)

[uG]

http://matrix.cpantesters.org/?dist=Text-Levenshtein-Damerau-XS-2.7;os=openbsd;reports=1

http://matrix.cpantesters.org/?dist=Text-Levenshtein-Damerau-XS-2.8;os=openbsd;reports=1

My module appears to fail randomly (notice they don't fail on the same test between versions, although it appears to be the same tests within the same version) when calling a function with Error 255. I'm guessing its a segmentation fault as it doesn't complete the rest of the tests once it fails.

I asked Bingos about it (as he is the smoke tester submitting the reports) and he commented that he gets asked about similar problems a lot, but wasn't aware of the cause of the problem (or the solution).

I'm going to bring up an OpenBSD box later today to test myself, but i'm not sure i'll even be able to replicate it (as I have many PASSes for the same configuration in the test matrix). Does anyone have any insight on where the problem might be?

EDIT: Just received a random failure on OpenBSD/Perl 5.8.9

http://www.cpantesters.org/cpan/report/d9a60682-6a7b-11e2-a2f1-82091a5d253f

[vsespb]

There is "Non-zero wait status: 139" - it's segfault. I think you definitely should try reproduce it.

[vsespb]

Another observation that failures are really random:

Parse errors: Bad plan.  You planned 17 tests but ran 2.
Parse errors: Bad plan.  You planned 17 tests but ran 4.
Parse errors: Bad plan.  You planned 17 tests but ran 16.

and there is no actual difference between versions.

[uG]

I just tried to replicate the problem on an OpenBSD 5.1 vm and Perl versions 5.12.x and couldn't get a failing test.

[choroba]

Check the environment. There should be something different. It also might be a bad memory on the smoker's machine.