Cortex-A9 bot unstable in Clang

Hi Galina, folks,

Does anyone have an idea why the A9 bots are unstable on Clang? They
all seem to be the same problem:

1. <eof> parser at end of file
Segmentation fault

http://lab.llvm.org:8011/builders/clang-native-arm-cortex-a9/builds/26073/steps/check-all/logs/Clang%3A%3Adead-stores.c

http://lab.llvm.org:8011/builders/clang-native-arm-cortex-a9/builds/26068/steps/check-all/logs/Clang%3A%3ACFDateGC.m

http://lab.llvm.org:8011/builders/clang-native-arm-cortex-a9/builds/26068/steps/check-all/logs/Clang%3A%3Ainline.cpp

This happens on more than one buildslave, so I don't think it's a
board related problem, but some instability either on the system
(libraries, tools?) or specific to A9. This behaviour seems to be
pretty consistent over the last several few months to be coincidence.

cheers,
--renato

Renato Golin <renato.golin@linaro.org> writes:

Does anyone have an idea why the A9 bots are unstable on Clang? They
all seem to be the same problem:

1. <eof> parser at end of file
Segmentation fault

Could it be something as simple as running out of memory (or some other
resource) and getting killed by the host OS? That kind of thing tends to
look something like this.

Hi Justin,

Yes, that could be it. Panda boards have 1GB of RAM, and Clang objects
are getting big.

Galina, do you have some kind of swap on that board? I know swapping
on USB is not great, but at least it stops the bot from dying every
few builds.

cheers,
--renato

Hi Renato,

There is swap on each board and I’m not aware of it’s going anywhere close to the end.

KiB Mem: 1003948 total, 637868 used, 366080 free, 9788 buffers
KiB Swap: 524284 total, 13836 used, 510448 free. 183144 cached Mem

$ swapon -s
Filename Type Size Used Priority
/swapfile file 524284 13836 -1

I’ll take one of the boards off and see if I could get the issue in some reproducible form.

Thanks

Galina

There is swap on each board and I'm not aware of it's going anywhere close
to the end.

Doesn't seem like it... :confused:

I'll take one of the boards off and see if I could get the issue in some
reproducible form.

Thanks!

--renato