[Piglit] More JSON problems. :(

[Kenneth Graunke]

On 08/05/2011 01:57 PM, Chad Versace wrote:
> On 08/05/2011 01:14 PM, Ian Romanick wrote:
>> I was trying to test Paul's recent GLSL tests on AMD's closed-source
>> driver, and I encountered some difficulties that were unrelated to those
>> tests.  A couple tests, one of the FBO tests if I recall correctly,
>> generate output on failing runs that is not valid UTF-8.  Somewhere deep
>> in the JSON writer, it gets really angry (and throws an exception) when
>> it receives the illegal UTF-8 bits.  As soon as that happens, the whole
>> test run aborts.
>>
>> I'm not sure what the solution is, but we need to find one.
> 
> This needs to be fixed ASAP. Its unacceptable to have Piglit crashing due to
> silly UTF problems.
> 
> To debug this, it would help if I had the raw output of the grumpy test (by
> running the test binary stand-alone) and a Python backtrace.

Apparently Paul thinks he can fix it.

In the meantime you can apply this to work around the issue:

http://cgit.freedesktop.org/~kwg/piglit/patch/?id=1b6a9e14a98ae308ec4cb30efbb23363fde75e56

It's by no means ideal.