[Intel-gfx] [PATCH] drm/i915: fix header test with GCOV

[Jani Nikula]

On Fri, 21 Feb 2020, Randy Dunlap <rdunlap at infradead.org> wrote:
> On 2/21/20 9:49 PM, Masahiro Yamada wrote:
>> On Sat, Feb 22, 2020 at 2:25 PM Randy Dunlap <rdunlap at infradead.org> wrote:
>>>
>>> On 2/21/20 8:53 PM, Masahiro Yamada wrote:
>>>> On Sat, Feb 22, 2020 at 1:43 PM Masahiro Yamada <masahiroy at kernel.org> wrote:
>>>>>
>>>>> Hi Jani,
>>>>>
>>>>> On Fri, Feb 21, 2020 at 7:54 PM Jani Nikula <jani.nikula at intel.com> wrote:
>>>>>>
>>>>>> $(CC) with $(CFLAGS_GCOV) assumes the output filename with .gcno suffix
>>>>>> appended is writable. This is not the case when the output filename is
>>>>>> /dev/null:
>>>>>>
>>>>>>   HDRTEST drivers/gpu/drm/i915/display/intel_frontbuffer.h
>>>>>> /dev/null:1:0: error: cannot open /dev/null.gcno
>>>>>>   HDRTEST drivers/gpu/drm/i915/display/intel_ddi.h
>>>>>> /dev/null:1:0: error: cannot open /dev/null.gcno
>>>>>> make[5]: *** [../drivers/gpu/drm/i915/Makefile:307:
>>>>>> drivers/gpu/drm/i915/display/intel_ddi.hdrtest] Error 1
>>>>>> make[5]: *** Waiting for unfinished jobs....
>>>>>> make[5]: *** [../drivers/gpu/drm/i915/Makefile:307:
>>>>>> drivers/gpu/drm/i915/display/intel_frontbuffer.hdrtest] Error 1
>>>>>>
>>>>>> Filter out $(CFLAGS_GVOC) from the header test $(c_flags) as they don't
>>>>>> make sense here anyway.
>>>>>
>>>>>
>>>>>
>>>>> Is GCOV the only case that produces a separate file?
>>>>>
>>>>> Could you also test CONFIG_DEBUG_INFO_SPLIT, please ?
>>>>>
>>>>>
>>>>> The GCC manual says this:
>>>>>
>>>>> -gsplit-dwarf
>>>>>
>>>>>    Separate as much DWARF debugging information as possible into a
>>>>> separate output
>>>>>    file with the extension .dwo. This option allows the build system
>>>>> to avoid linking
>>>>>    files with debug information. To be useful, this option requires a debugger
>>>>>    capable of reading .dwo files.
>>>>>
>>>>
>>>>
>>>> I just tested it.
>>>>
>>>> This is not a problem for header test
>>>> because cmd_hdrtest uses '-S' instead of '-c'.
>>>>
>>>> If '-c' were used, we would see a similar error.
>>>>
>>>>
>>>> So, gsplit-dwarf is OK.
>>>
>>> Yes, works for me also.
>>>
>>> (I think you have a faster build machine that I do. I began the build
>>> almost immediately after reading your email. :)
>> 
>> 
>> I use a reasonable PC for my development.
>> (core-i9 with 4 physical cores)
>> 
>> 
>> I just compiled under i915/ with this command:
>> 
>> make drivers/gpu/drm/i915/
>
> Good trick.  I did a full build.

Thanks for the testing.

I also checked with some other options, did not see any failures. I'll
go with this for now, let's revisit as needed.

Another related thing is we try to hide this a bit from build testing
outside of our CI with "depends on !COMPILE_TEST" but I guess there's no
way to ensure something is not enabled on a randconfig build...

BR,
Jani.


-- 
Jani Nikula, Intel Open Source Graphics Center