Re: [Qemu-devel] Slow tests on "make check" (was Re: [PULL 6/9] Convert error_report() to warn_report())

[Alistair Francis]

On Tue, Jul 25, 2017 at 4:24 PM, Alistair Francis
<address@hidden> wrote:
> On Tue, Jul 25, 2017 at 12:08 PM, Eduardo Habkost <address@hidden> wrote:
>> On Mon, Jul 24, 2017 at 09:30:06AM +0200, Paolo Bonzini wrote:
>>> On 24/07/2017 08:30, Markus Armbruster wrote:
>>> > Markus Armbruster <address@hidden> writes:
>>> >
>>> >> Kevin Wolf <address@hidden> writes:
>>> >>
>>> >>> Am 13.07.2017 um 15:27 hat Markus Armbruster geschrieben:
>>> >>>> From: Alistair Francis <address@hidden>
>>> >>>>
>>> >>>> Convert all uses of error_report("warning:"... to use warn_report()
>>> >>>> instead. This helps standardise on a single method of printing warnings
>>> >>>> to the user.
>>> >>>>
>>> >>>> All of the warnings were changed using these two commands:
>>> >>>>     find ./* -type f -exec sed -i \
>>> >>>>       's|error_report(".*warning[,:] |warn_report("|Ig' {} +
>>> >>>>
>>> >>>> Indentation fixed up manually afterwards.
>>> >>>>
>>> >>>> The test-qdev-global-props test case was manually updated to ensure 
>>> >>>> that
>>> >>>> this patch passes make check (as the test cases are case sensitive).
>>> >>>
>>> >>> This patch broke qemu-iotests 051 because it neglected to update the
>>> >>> reference output. Not sure if a change of the message was even intended,
>>> >>> but with a error location prefix, the order changes:
>>> >>>
>>> >>> -(qemu) QEMU_PROG: -drive if=scsi,media=cdrom: warning: bus=0,unit=0 is 
>>> >>> deprecated with this machine type
>>> >>> +(qemu) warning: qemu-system-x86_64: -drive if=scsi,media=cdrom: 
>>> >>> bus=0,unit=0 is deprecated with this machine type
>>> >>>
>>> >>> Personally, I would expect the error location or at least the program
>>> >>> name to come first even for warnings.
>>> >>
>>> >> I'll fix it.
>>> >>
>>> >> While focusing on something other than block, I forget qemu-iotests
>>> >> exist.  My fault, but it's a pretty common fault.  I reiterate my plea
>>> >> to include (a sensible subset of) it in "make check".
>>> >
>>> > Oh, 051 isn't run by "make check-block".  It could just as well not
>>> > exist then.
>>> >
>>> > What's the recommended way to run all iotests a build of QEMU can run?
>>>
>>> I use "path-to-srcdir/tests/check-block.sh".  It takes about 20 minutes.
>>>
>>> There are two issues with check-block:
>>>
>>> * tests that require QEMU are skipped (rationale: you might be building
>>> without the x86_64-softmmu target).
>>>
>>> * it only runs the "quick" group, but I think we should reverse the
>>> direction and mark slow tests specifically (that would cut about 10
>>> minutes out of the 20).
>>
>> About slow tests: I never included device-crash-test on "make
>> check" because it would take too long to run.  Also, "make
>> check-help" lies when it says "make check" runs all tests,
>> because it doesn't run check-block.
>>
>> Should we add a new Makefile target (check-all?) that includes
>> slower tests?
>
> Or a make check-slow?

Actually, what about to avoid duplication we can add a make check-slow
that runs all the slow tests that make check doesn't run. Then make
check-all can run both make check and make check-slow.

That way you can run make check for a quick sanity test and then run
make check-slow to test everything, without duplicating tests.

>
> I think that is a great idea, having tests but not running them isn't
> helping anyone.
>
> Thanks,
> Alistair
>
>>
>> --
>> Eduardo
>>