On 2020-02-20 17:44, Chris Wilson wrote:
> Quoting Martin Peres (2020-02-20 15:32:09)
>> +###############################################################################
>> +# Perf tests are for people using performance counters to get details about
>> +# how the execution is performed (Observability Architecture). As such, the
>> +# audience is very limited (game developers, driver developers), and it does
>> +# not justify the overall execution time:
>> +#
>> +# - shard-skl: 0%
>> +# - shard-kbl: 0%
>> +# - shard-apl: 0%
>> +# - shard-glk: 0%
>> +# - shard-icl: 0%
>> +# - shard-tgl: 1.7% (~3.5 minutes)
>> +#
>> +# Data acquired on 2020-02-20 by Martin Peres
>> +###############################################################################
>> +igt@perf@gen12-mi-rpc
> 
> Nak. That is just a straightforward BUG. The test is doing it's job in
> reporting the interface is snafu. Then the test screws up by deadlocking
> due to unclean exit handling after detecting the error.

I raised the severity, but this has been opened for 3 weeks without much
work being done on it.

That being said, I don't mind dropping this hunk.

Martin