[PATCH] Implement run-built-tests=no for make xcheck

[Carlos O'Donell]

On 9/20/24 6:20 AM, Florian Weimer wrote:
> * Andreas K. Huettel:
> 
>> Am Donnerstag, 19. September 2024, 15:53:23 MESZ schrieb Florian Weimer:
>>> Previously, the second occurrence of the xtests target
>>> expected all xtests to run (as the r esult of specifying
>>> $(xtests)), but these tests have not been run due to
>>> the the first xtests target is set up in run-built-tests=no:
>>> it only runs tests in $(xtests-special).
>>>
>>> To make run-built-tests=no actually useful, tweak the
>>> first xtests target to build tests, similar to what the
>>> parallel tests target does.
>>>
>>
>> More general question in this context:
>>
>> I have been assuming that "make xcheck" is deprecated and should not be 
>> used anymore. Following Carlos' lead, I have consequently also removed
>> mentions of xcheck from our wiki.
>>
>> * Is this incorrect and is xcheck still actively in use?
> 
> I've recently added xcheck tests.  I wasn't aware that there was an
> effort to remove them.  We have xtests that should be deleted
> (nss/bug-erange for example, because we have equivalent/better coverage
> in other tests), or that could be converted to one or two regular tests
> (with the special one automatically marked as unsupported).  Some
> previously destructive tests can be carried out in namespaces today
> (e.g., testing sethostname).

Rather than remove them, I want them to move to check.

I've written up my thoughts here:
https://inbox.sourceware.org/libc-alpha/ede8eb08-0cef-4073-955a-a071aa82e5bc@redhat.com/T/#u
 
>> * If not, shouldn't it rather fully go away then?
> 
> I don't think they can fully go away, ever.  If we have tests that take
> minutes to run even on fast/large systems, we don't want them to run
> during make check.  It's not useful to stretch testing time from less
> than 15 minutes to more than half an hour without a substantial increase
> in coverage.

It can go away. But in doing so we would remove a choice that developers get to make
about coverage.

You seem to argue that you would like to keep that choice?

-- 
Cheers,
Carlos.