[petsc-dev] broken nightlybuilds (next vs next-tmp)

[Smith, Barry F.]

> On Nov 11, 2017, at 3:26 PM, Jed Brown <jed at jedbrown.org> wrote:
> 
> "Smith, Barry F." <bsmith at mcs.anl.gov> writes:
> 
>>   There is no reason to waste time protesting the attempt to change
>>   to the new model. The attempt will happen as soon as we have the
>>   new test harness fully working. So help out or get out of the way.
> 
> I can't help convert tests if I'm busy arguing against a change that I
> think will be disastrous to the quality of 'master'.

   You are arguing against a change in the abstract because you love next! You are making up stray men and attacking them. Wait until there is a real proposal then point out flaws and make suggestions on how to improve it. There is no reason to develop the new model until we have the test harness done so the new model would have any hope of working.

   Regardless of what next/.../... model you want we all benefit greatly from a much faster testing. Surely you cannot be opposed to that.


> 
> Let's table the discussion of removing 'next' until we have a test
> system capable of keeping 'next' clean.  It provides a ton of safety at
> the moment.  If an automated testing system is capable of keeping
> 'master' clean, it can also be used to keep 'next' clean.  So if it
> works, 'next' will be clean all the time and we can justify removing it
> as needless overhead.  If it's still catching problems, then the
> automated testing system clearly isn't sufficient.