[Intel-gfx] [PATCH] drm/i915: rename header test build commands to avoid conflicts

[yamada.masahiro at socionext.com]

Hi Jani,

> -----Original Message-----
> From: Jani Nikula [mailto:jani.nikula at intel.com]
> Sent: Thursday, June 06, 2019 4:25 PM
> To: Yamada, Masahiro/山田 真弘 <yamada.masahiro at socionext.com>;
> intel-gfx at lists.freedesktop.org
> Cc: lkp at intel.com; chris at chris-wilson.co.uk; sam at ravnborg.org;
> masahiroy at kernel.org
> Subject: RE: [PATCH] drm/i915: rename header test build commands to avoid
> conflicts
> 
> On Thu, 06 Jun 2019, <yamada.masahiro at socionext.com> wrote:
> > Hi,
> >
> >> -----Original Message-----
> >> From: Jani Nikula [mailto:jani.nikula at intel.com]
> >> Sent: Wednesday, June 05, 2019 10:22 PM
> >> To: intel-gfx at lists.freedesktop.org
> >> Cc: jani.nikula at intel.com; kbuild test robot <lkp at intel.com>; Chris
> Wilson
> >> <chris at chris-wilson.co.uk>; Yamada, Masahiro/山田 真弘
> >> <yamada.masahiro at socionext.com>; Sam Ravnborg <sam at ravnborg.org>
> >> Subject: [PATCH] drm/i915: rename header test build commands to avoid
> >> conflicts
> >>
> >> We have a local hack to test if headers are self-contained, while
> >> upstreaming a proper generic solution in kbuild [1]. Now that both have
> >> found themselves in linux-next, the identical cmd_header_test build
> >> commands conflict, leading to errors such as:
> >>
> >> >> drivers/gpu/drm/i915/header_test_intel_audio.c:1:10: fatal error:
> >> >> drivers/gpu/drm/i915/intel_audio.h: No such file or directory
> >>     #include "drivers/gpu/drm/i915/intel_audio.h"
> >> 	     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >>
> >> Rename the i915 local build command until the proper kbuild solution
> >> finds its way to Linus' master and gets backported to our tree, and we
> >> can finally switch over.
> >>
> >> Note that since the kbuild header test requires CONFIG_HEADER_TEST=y,
> >> and our hack requires our debug option CONFIG_DRM_I915_WERROR=y, this
> is
> >> likely hit only by build test bots.
> >>
> >> [1] http://marc.info/?i=20190604124248.5564-1-jani.nikula@intel.com
> >>
> >> Reported-by: kbuild test robot <lkp at intel.com>
> >> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> >> Cc: Masahiro Yamada <yamada.masahiro at socionext.com>
> >> Cc: Sam Ravnborg <sam at ravnborg.org>
> >> Signed-off-by: Jani Nikula <jani.nikula at intel.com>
> >> ---
> >
> >
> > This is not really queued up yet.
> >
> > So, we can squash fix-up to avoid 0day bot report.
> 
> Except I don't think your linux-kbuild.git baseline has the files you're
> patching below. The problem only exists at the merge of our trees,
> currently only linux-next, and not "for real" until the v5.3 merge
> window. So I think the sane option is to patch it up in our tree.


I do not understand.

This is a _real_ problem since
  drivers/gpu/drm/i915/Makefile.header-test
exists in Linus' tree.


The 0-day bot reported the build error against my tree,
so I must fix it in my tree.


> I want to use our local hack until we can get the backmerge from
> v5.3-rc1, because it's 7-8 weeks away, and I want to retain our own
> pre-merge build test coverage until then rather than relying on 0day
> post-merge testing on linux-next.


Neither of my patches breaks your test coverage.
CONFIG_DRM_I915_WERROR still works in linux-next too.

What am I missing?



Thanks.

Masahiro Yamada