Bad rever-select of CONFIG_SND_HDA_I915

Bad rever-select of CONFIG_SND_HDA_I915

[Takashi Iwai]

Hi,

while build testing, I noticed the following warning:
  warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)

And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
driver side.  This is the condition depending on DRM i915 driver.

Fixing this doesn't look too trivial, though.  Since
SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
"depends on".  So, the top-level configs that select
SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
modify the driver to be built without i915 binding.


Takashi

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Vinod Koul]

On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> Hi,
> 
> while build testing, I noticed the following warning:
>   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)

Thanks Takashi for reporting,

> And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> driver side.  This is the condition depending on DRM i915 driver.

But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
to select it. I didn't see anyone select it, either from Gfx side or HDA
side?

> 
> Fixing this doesn't look too trivial, though.  Since
> SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> "depends on".  So, the top-level configs that select
> SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> modify the driver to be built without i915 binding.

can you please share the config where you saw this...

-- 
~Vinod

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Takashi Iwai]

On Fri, 18 Mar 2016 09:06:49 +0100,
Vinod Koul wrote:
> 
> On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> > Hi,
> > 
> > while build testing, I noticed the following warning:
> >   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)
> 
> Thanks Takashi for reporting,
> 
> > And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> > driver side.  This is the condition depending on DRM i915 driver.
> 
> But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
> to select it. I didn't see anyone select it, either from Gfx side or HDA
> side?

No, it's not a kconfig to be selected manually by others, but it
adjusts itself automatically depending on DRM_I915.  That is, it's a
read-only value.

> > Fixing this doesn't look too trivial, though.  Since
> > SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> > "depends on".  So, the top-level configs that select
> > SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> > modify the driver to be built without i915 binding.
> 
> can you please share the config where you saw this...

Just set DRM_I915=n and run config again on your machine.


Takashi

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Vinod Koul]

On Fri, Mar 18, 2016 at 10:22:45AM +0100, Takashi Iwai wrote:
> On Fri, 18 Mar 2016 09:06:49 +0100,
> Vinod Koul wrote:
> > 
> > On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> > > Hi,
> > > 
> > > while build testing, I noticed the following warning:
> > >   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)
> > 
> > Thanks Takashi for reporting,
> > 
> > > And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> > > driver side.  This is the condition depending on DRM i915 driver.
> > 
> > But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
> > to select it. I didn't see anyone select it, either from Gfx side or HDA
> > side?
> 
> No, it's not a kconfig to be selected manually by others, but it
> adjusts itself automatically depending on DRM_I915.  That is, it's a
> read-only value.

Okay then

> > > Fixing this doesn't look too trivial, though.  Since
> > > SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> > > "depends on".  So, the top-level configs that select
> > > SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> > > modify the driver to be built without i915 binding.
> > 
> > can you please share the config where you saw this...
> 
> Just set DRM_I915=n and run config again on your machine.

Yes saw this now. I made depends on DRM_I915 to make it disappear.

I will check again for compilation fix and get back on proper patch either
which depends on DRM or ensure we compile clean

Thanks
-- 
~Vinod

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Takashi Iwai]

On Fri, 18 Mar 2016 11:28:06 +0100,
Vinod Koul wrote:
> 
> On Fri, Mar 18, 2016 at 10:22:45AM +0100, Takashi Iwai wrote:
> > On Fri, 18 Mar 2016 09:06:49 +0100,
> > Vinod Koul wrote:
> > > 
> > > On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> > > > Hi,
> > > > 
> > > > while build testing, I noticed the following warning:
> > > >   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)
> > > 
> > > Thanks Takashi for reporting,
> > > 
> > > > And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> > > > driver side.  This is the condition depending on DRM i915 driver.
> > > 
> > > But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
> > > to select it. I didn't see anyone select it, either from Gfx side or HDA
> > > side?
> > 
> > No, it's not a kconfig to be selected manually by others, but it
> > adjusts itself automatically depending on DRM_I915.  That is, it's a
> > read-only value.
> 
> Okay then
> 
> > > > Fixing this doesn't look too trivial, though.  Since
> > > > SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> > > > "depends on".  So, the top-level configs that select
> > > > SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> > > > modify the driver to be built without i915 binding.
> > > 
> > > can you please share the config where you saw this...
> > 
> > Just set DRM_I915=n and run config again on your machine.
> 
> Yes saw this now. I made depends on DRM_I915 to make it disappear.
 
OTOH, this makes impossible to use the driver without graphics, even
if you want only Realtek codec.

> I will check again for compilation fix and get back on proper patch either
> which depends on DRM or ensure we compile clean

I guess the build failure won't happen even if you rip off
SND_HDA_I915 dependency from SND_SOC_INTEL_SKYLAKE.  Let's check.


Takashi

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Vinod Koul]

On Fri, Mar 18, 2016 at 11:33:38AM +0100, Takashi Iwai wrote:
> On Fri, 18 Mar 2016 11:28:06 +0100,
> Vinod Koul wrote:
> > 
> > On Fri, Mar 18, 2016 at 10:22:45AM +0100, Takashi Iwai wrote:
> > > On Fri, 18 Mar 2016 09:06:49 +0100,
> > > Vinod Koul wrote:
> > > > 
> > > > On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> > > > > Hi,
> > > > > 
> > > > > while build testing, I noticed the following warning:
> > > > >   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)
> > > > 
> > > > Thanks Takashi for reporting,
> > > > 
> > > > > And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> > > > > driver side.  This is the condition depending on DRM i915 driver.
> > > > 
> > > > But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
> > > > to select it. I didn't see anyone select it, either from Gfx side or HDA
> > > > side?
> > > 
> > > No, it's not a kconfig to be selected manually by others, but it
> > > adjusts itself automatically depending on DRM_I915.  That is, it's a
> > > read-only value.
> > 
> > Okay then
> > 
> > > > > Fixing this doesn't look too trivial, though.  Since
> > > > > SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> > > > > "depends on".  So, the top-level configs that select
> > > > > SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> > > > > modify the driver to be built without i915 binding.
> > > > 
> > > > can you please share the config where you saw this...
> > > 
> > > Just set DRM_I915=n and run config again on your machine.
> > 
> > Yes saw this now. I made depends on DRM_I915 to make it disappear.
>  
> OTOH, this makes impossible to use the driver without graphics, even
> if you want only Realtek codec.
> 
> > I will check again for compilation fix and get back on proper patch either
> > which depends on DRM or ensure we compile clean
> 
> I guess the build failure won't happen even if you rip off
> SND_HDA_I915 dependency from SND_SOC_INTEL_SKYLAKE.  Let's check.

It doesn't. So both work well :)

Depends on DRM_I915 as well as removing current select.

FWIW, the removing the DRM will not work as HDMI will fail so dailinks wont
be there and sound card will not get registered :)

So which one to do in above?

-- 
~Vinod

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Takashi Iwai]

On Fri, 18 Mar 2016 11:54:07 +0100,
Vinod Koul wrote:
> 
> On Fri, Mar 18, 2016 at 11:33:38AM +0100, Takashi Iwai wrote:
> > On Fri, 18 Mar 2016 11:28:06 +0100,
> > Vinod Koul wrote:
> > > 
> > > On Fri, Mar 18, 2016 at 10:22:45AM +0100, Takashi Iwai wrote:
> > > > On Fri, 18 Mar 2016 09:06:49 +0100,
> > > > Vinod Koul wrote:
> > > > > 
> > > > > On Fri, Mar 18, 2016 at 08:34:49AM +0100, Takashi Iwai wrote:
> > > > > > Hi,
> > > > > > 
> > > > > > while build testing, I noticed the following warning:
> > > > > >   warning: (SND_SOC_INTEL_SKYLAKE) selects SND_HDA_I915 which has unmet direct dependencies (SOUND && !M68K && !UML && SND && DRM_I915 && SND_HDA_CORE)
> > > > > 
> > > > > Thanks Takashi for reporting,
> > > > > 
> > > > > > And it's indeed broken: you can't select CONFIG_SND_HDA_I915 from the
> > > > > > driver side.  This is the condition depending on DRM i915 driver.
> > > > > 
> > > > > But who selects CONFIG_SND_HDA_I915. It is a bool symbol, so someone needs
> > > > > to select it. I didn't see anyone select it, either from Gfx side or HDA
> > > > > side?
> > > > 
> > > > No, it's not a kconfig to be selected manually by others, but it
> > > > adjusts itself automatically depending on DRM_I915.  That is, it's a
> > > > read-only value.
> > > 
> > > Okay then
> > > 
> > > > > > Fixing this doesn't look too trivial, though.  Since
> > > > > > SND_SOC_INTEL_SKYLAKE is the item to be selected, it can't have
> > > > > > "depends on".  So, the top-level configs that select
> > > > > > SND_SOC_INTEL_SKYLAKE need to have "depends on DRM_I915" lines, or
> > > > > > modify the driver to be built without i915 binding.
> > > > > 
> > > > > can you please share the config where you saw this...
> > > > 
> > > > Just set DRM_I915=n and run config again on your machine.
> > > 
> > > Yes saw this now. I made depends on DRM_I915 to make it disappear.
> >  
> > OTOH, this makes impossible to use the driver without graphics, even
> > if you want only Realtek codec.
> > 
> > > I will check again for compilation fix and get back on proper patch either
> > > which depends on DRM or ensure we compile clean
> > 
> > I guess the build failure won't happen even if you rip off
> > SND_HDA_I915 dependency from SND_SOC_INTEL_SKYLAKE.  Let's check.
> 
> It doesn't. So both work well :)
> 
> Depends on DRM_I915 as well as removing current select.
> 
> FWIW, the removing the DRM will not work as HDMI will fail so dailinks wont
> be there and sound card will not get registered :)
> 
> So which one to do in above?

HDMI would never work without graphics, so it shouldn't be a big
problem.  I'm for dropping the dependency, for now.


Takashi

Re: Bad rever-select of CONFIG_SND_HDA_I915

[Vinod Koul]

On Fri, Mar 18, 2016 at 12:23:54PM +0100, Takashi Iwai wrote:
> > > I guess the build failure won't happen even if you rip off
> > > SND_HDA_I915 dependency from SND_SOC_INTEL_SKYLAKE.  Let's check.
> > 
> > It doesn't. So both work well :)
> > 
> > Depends on DRM_I915 as well as removing current select.
> > 
> > FWIW, the removing the DRM will not work as HDMI will fail so dailinks wont
> > be there and sound card will not get registered :)
> > 
> > So which one to do in above?
> 
> HDMI would never work without graphics, so it shouldn't be a big
> problem.  I'm for dropping the dependency, for now.

Okay sent you the patch for that

-- 
~Vinod