[SeaBIOS] [Qemu-devel] [QEMU RFC PATCH 6/7] i386: topology & APIC ID utility functions

[Blue Swirl]

On Mon, Jul 23, 2012 at 8:14 PM, Eduardo Habkost <ehabkost at redhat.com> wrote:
> On Mon, Jul 23, 2012 at 07:44:44PM +0000, Blue Swirl wrote:
>> On Mon, Jul 23, 2012 at 7:28 PM, Eduardo Habkost <ehabkost at redhat.com> wrote:
>> > On Mon, Jul 23, 2012 at 07:11:11PM +0000, Blue Swirl wrote:
>> >> On Mon, Jul 23, 2012 at 6:59 PM, Eduardo Habkost <ehabkost at redhat.com> wrote:
>> >> > On Mon, Jul 23, 2012 at 04:49:07PM +0000, Blue Swirl wrote:
>> >> >> On Mon, Jul 16, 2012 at 5:42 PM, Eduardo Habkost <ehabkost at redhat.com> wrote:
>> >> >> > On Sat, Jul 14, 2012 at 09:14:30AM +0000, Blue Swirl wrote:
>> >> >> > [...]
>> >> >> >> >> > diff --git a/tests/Makefile b/tests/Makefile
>> >> >> >> >> > index b605e14..89bd890 100644
>> >> >> >> >> > --- a/tests/Makefile
>> >> >> >> >> > +++ b/tests/Makefile
>> >> >> >> >> > @@ -15,6 +15,7 @@ check-unit-y += tests/test-string-output-visitor$(EXESUF)
>> >> >> >> >> >  check-unit-y += tests/test-coroutine$(EXESUF)
>> >> >> >> >> >  check-unit-y += tests/test-visitor-serialization$(EXESUF)
>> >> >> >> >> >  check-unit-y += tests/test-iov$(EXESUF)
>> >> >> >> >> > +check-unit-y += tests/test-x86-cpuid$(EXESUF)
>> >> >> >> >>
>> >> >> >> >> This probably tries to build the cpuid test also for non-x86 targets
>> >> >> >> >> and break them all.
>> >> >> >> >
>> >> >> >> > I don't think there's any concept of "targets" for the check-unit tests.
>> >> >> >>
>> >> >> >> How about:
>> >> >> >> check-qtest-i386-y = tests/test-x86-cpuid$(EXESUF)
>> >> >> >
>> >> >> > test-x86-cpuid is not a qtest test case.
>> >> >>
>> >> >> Why not? I don't think it is a unit test either, judging from what the
>> >> >> other unit tests do.
>> >> >
>> >> > It is absolutely a unit test. I don't know why you don't think so. It
>> >> > simply checks the results of the APIC ID calculation functions.
>> >>
>> >> Yes, but the other 'unit tests' (the names used here are very
>> >> confusing, btw) check supporting functions like coroutines, not
>> >> hardware. Hardware tests (qtest) need to bind to an architecture, in
>> >> this case x86.
>> >
>> > test-x86-cpuid doesn't check hardware, either. It just checks the simple
>> > functions that calculate APIC IDs (to make sure the math is correct).
>> > Those functions aren't even used by any hardware emulation code until
>> > later in the patch series (when the CPU initialization code gets changed
>> > to use the new function).
>>
>> By that time the function is clearly x86 HW and the check is an x86
>> hardware check. QEMU as whole consists of simple functions that
>> calculate something.
>
> It's useful onily for a specific architecture, yes, but it surely
> doesn't make libqtest necessary.
>
> That's the difference between the unit tests and qtest test cases: unit
> tests simply test small parts of the code (that may or may not be
> hardware-specific), and qtest tests hardware emulation after starting up
> a whole qemu process. It's a different kind of testing, with different
> sets of goals.
>

Thanks for the clarification. I agree now that this is not a qtest. I
think this is a new category of tests compared to those we have now:
supporting function unit tests and hardware tests at device I/O
boundary level.

> I suppose you are not arguing that no function inside target-* would be
> ever allowed to have a unit test written for it. That would be a very
> silly constraint for people writing tests.

Of course. What I really want is that if x86 targets are not built,
this test is skipped like qtests. This could be achieved with
something like:

check-unit-arch-i386-y = tests/test-x86-cpuid$(EXESUF)
check-unit-y += $(foreach TARGET,$(TARGETS), $(check-unit-arch-$(TARGET)-y))

>
> --
> Eduardo