[Debichem-devel] Bug#889859: Bug#889859: chemps2: FTBFS on mipsel: test5 segfaults

[Michael Banck]

Hi,

On Fri, Feb 09, 2018 at 09:26:25AM +0100, Sebastian Wouters wrote:
> First one:
> 
> Test project /<<PKGBUILDDIR>>/obj-mipsel-linux-gnu
>     Start 3: test3
> 1/3 Test #3: test3 ............................   Passed    8.10 sec
>     Start 5: test5
> 2/3 Test #5: test5 ............................***Exception: SegFault  9.37
> sec
>     Start 8: test8
> 3/3 Test #8: test8 ............................***Exception: SegFault  6.62
> sec
> 
> 33% tests passed, 2 tests failed out of 3
> 
> Total Test time (real) =  24.11 sec
> 
> The following tests FAILED:
>   5 - test5 (SEGFAULT)
>   8 - test8 (SEGFAULT)
> 
> 
> 
> Second one:
> 
> Test project /<<PKGBUILDDIR>>/obj-mipsel-linux-gnu
>     Start 3: test3
> 1/3 Test #3: test3 ............................   Passed    8.38 sec
>     Start 5: test5
> 2/3 Test #5: test5 ............................   Passed   19.40 sec
>     Start 8: test8
> 3/3 Test #8: test8 ............................***Exception: SegFault  6.53
> sec
> 
> 67% tests passed, 1 tests failed out of 3
> 
> Total Test time (real) =  34.35 sec
> 
> The following tests FAILED:
>   8 - test8 (SEGFAULT)
> 
> 
> 
> Third one:
> 
> Test project /<<PKGBUILDDIR>>/obj-mipsel-linux-gnu
>     Start 3: test3
> 1/3 Test #3: test3 ............................   Passed   19.01 sec
>     Start 5: test5
> 2/3 Test #5: test5 ............................***Exception: SegFault  1.75
> sec
>     Start 8: test8
> 3/3 Test #8: test8 ............................   Passed   14.19 sec
> 
> 67% tests passed, 1 tests failed out of 3
> 
> Total Test time (real) =  34.99 sec
> 
> The following tests FAILED:
>   5 - test5 (SEGFAULT)
> 
> 
> It seems that passing or seg fault seems to occur randomly.

Yeah, unfortunately, mipsel is not the most stable architecture. I am
not sure what we should do here, we can either agressively retry to
build on mipsel until all tests pass or alternatively ignore test
failures. I don't think it would be easy to just ignore them on mipsel,
so that'd mean we might not catch issues going forward on other arches.

We do ignore testsuite failures for several other packages, but I think
it is way better to fail on testsuite failures and to diagnose them. In
this case, there's probably just not much to diagnose.

The final alternative would be to remove the previous mipsel binaries so
the missing build here will not impact chemps2's testing migration.


Michael