DRM: urgent v4.9-rc6 build regression: master build: 2 failures 1 warnings (v4.9-rc5-213-g961b708)

DRM: urgent v4.9-rc6 build regression: master build: 2 failures 1 warnings (v4.9-rc5-213-g961b708)

[Arnd Bergmann]

On Thursday, November 17, 2016 4:14:44 AM CET Build bot for Mark Brown wrote:
>         arm64-allmodconfig
> ../drivers/gpu/drm/mediatek/mtk_drm_ddp_comp.c:126:36: error: 'OD_CFG' undeclared (first use in this function)
> ../drivers/gpu/drm/mediatek/mtk_drm_drv.c:220:5: error: 'struct drm_device' has no member named 'vblank_disable_allowed'
> 
>         arm-allmodconfig
> ../drivers/gpu/drm/mediatek/mtk_drm_ddp_comp.c:126:36: error: 'OD_CFG' undeclared (first use in this function)
> ../drivers/gpu/drm/mediatek/mtk_drm_drv.c:220:5: error: 'struct drm_device' has no member named 'vblank_disable_allowed'

These two patches need to be reverted or fixed:

f752fff ("drm/mediatek: set vblank_disable_allowed to true")
83ba62b ("drm/mediatek: fix a typo of OD_CFG to OD_RELAYMODE")

Both of them seem to work fine on linux-next, but were accidentally sent
to Linus yesterday through:

    Merge tag 'drm-fixes-for-v4.9-rc6' of git://people.freedesktop.org/~airlied/linux
    Merge branch 'mediatek-drm-fixes-2016-11-11' of https://github.com/ckhu-mediatek/linux.git-tags into drm-fixes

Clearly that branch has never been build-tested by itself. Unfortunately
the dependency on "HAVE_ARM_SMCCC" caused the driver to not be no longer
included in x86 allmodconfig builds despite the COMPILE_TEST check,
so neither David Airlie nor Linus Torvalds caught it before it got merged.

	Arnd

Re: DRM: urgent v4.9-rc6 build regression: master build: 2 failures 1 warnings (v4.9-rc5-213-g961b708)

[Arnd Bergmann]

On Thursday, November 17, 2016 8:50:05 PM CET Dave Airlie wrote:
> 
> Arnd could you send a git pull with the two reverts, with my Acked-by on
> it? I won't be in a place to do it for 8-9hrs.

I don't think it's that urgent, as long as we make sure it's fixed in the
next -rc. I've sent out the reverts as patches with a little more information
in the changelog: it turns out that they are actually broken on linux-next too,
they had just not made it in there, and one of the two actually did build
on older kernels.

I think what happened here is that the fixes were tested on a v4.4 kernel
and blindly forward-ported. It probably makes sense to look at the
entire series again in case another one of them is broken.

Philipp, Hu, Bibby, could one of you have another look?

	Arnd

Re: DRM: urgent v4.9-rc6 build regression: master build: 2 failures 1 warnings (v4.9-rc5-213-g961b708)

[CK Hu]

Hi, Arnd:

I've made a mistake that I've tried to build these patches on v4.9-rc1,
but I does not set CONFIG_DRM_MEDIATEK=y, therefore I didn't find out
these build fails. Now I fix the config problem, and I think I should
build these patches on latest kernel version even though patch's owner
test on old kernel version. I wish this flow would make things better.

It's ok that you just revert these two patches. I've fixed build fail
and will request pull later.

Regards,
CK

On Thu, 2016-11-17 at 16:24 +0100, Arnd Bergmann wrote:
> On Thursday, November 17, 2016 8:50:05 PM CET Dave Airlie wrote:
> > 
> > Arnd could you send a git pull with the two reverts, with my Acked-by on
> > it? I won't be in a place to do it for 8-9hrs.
> 
> I don't think it's that urgent, as long as we make sure it's fixed in the
> next -rc. I've sent out the reverts as patches with a little more information
> in the changelog: it turns out that they are actually broken on linux-next too,
> they had just not made it in there, and one of the two actually did build
> on older kernels.
> 
> I think what happened here is that the fixes were tested on a v4.4 kernel
> and blindly forward-ported. It probably makes sense to look at the
> entire series again in case another one of them is broken.
> 
> Philipp, Hu, Bibby, could one of you have another look?
> 
> 	Arnd