ubuntu-phone team mailing list archive

[Ricardo Salveti de Araujo <ricardo.salveti@xxxxxxxxxxxxx>]

On Tue, Feb 25, 2014 at 6:11 AM, Łukasz 'sil2100' Zemczak
<lukasz.zemczak@xxxxxxxxxxxxx> wrote:
> Hello,
>
> Sadly, regarding unity8 test flakyness, we don't have any good news. I
> have asked Saviq to help me out tracing down the underlying problem - it
> seems that sometimes during some random tests unity8 fails to start,
> causing the given tests to fail. It's reproducible locally. It seem only
> rare unity8 test-suite-runs pass completely - in every other run there
> is at least a few failures caused by this problem.
>
> The bad news is that we have still not identified the underlying
> problem. The crash seems to be happening due to the new Android bits,
> but debugging these kind of failures is a bit more time-consuming as we
> don't have direct access to the underlying android symbols. Michał is
> currently working on getting useful information from the stacktrace, but
> we suspect that the fix for this problem might have to be done elsewere.

The interesting thing is that the crash is not happening in the
Android libraries itself (at least with the crash that Saviq gave me).
Looking at proc/maps, it actually crashed in the heap, so it's either
a normal crash or tls related.

Will try to reproduce locally and play with TLS a bit to see.

Cheers,
-- 
Ricardo Salveti de Araujo