[Development] Removal of some of the blacklisted (non-working) autotests?

[Jedrzej Nowacki]

> Now: what should it do ?
Compile the test with additional debugging flags.

On Linux it would be looking like that:

for 10 min (minimum 3000 runs):
    Run the test
    Run the test with valgrind
    Use taskset to reduce CPU affinity to one CPU
    Put external load on the CPU to re-arrange threads schedule
    Use  other tools?

If it failed even once the change should be rejected. I'm not sure what should we do with warnings I guess we can not thread them as errors, but definitely we can print them out.

Cheers,
  Jędrek


________________________________________
From: Edward Welbourne
Sent: Wednesday, November 9, 2016 11:05:24 AM
To: Jedrzej Nowacki
Cc: development at qt-project.org
Subject: Re: [Development] Removal of some of the blacklisted (non-working) autotests?

Jedrzej Nowacki said:
> As you wrote, in the first iteration of the functionality I would go
> for an easy solution and I would detect just all modifications to
> tests/auto/*
> On the other hand I do not want to have the definition of stress
> testing in Coin code. Every one should be able to run it

Fair enough.

> therefore it should be a make target.  So everyone could just type:
>   make stress-test
> in a test folder and after 10 min get the result.

Makes sense - nice target, well worth adding to Makefiles.
Now: what should it do ?

        Eddy.