Re: Build regressions/improvements in v3.0-rc7 (sound/isa/)

[Takashi Iwai <tiwai@suse.de>]

At Thu, 28 Jul 2011 13:06:19 -0400,
Arnaud Lacombe wrote:
> 
> Hi,
> 
> On Thu, Jul 28, 2011 at 12:44 PM, Randy Dunlap <rdunlap@xenotime.net> wrote:
> > On Thu, 28 Jul 2011 12:36:27 -0400 Arnaud Lacombe wrote:
> >
> >> Hi,
> >>
> >> On Thu, Jul 28, 2011 at 12:21 PM, Randy Dunlap <rdunlap@xenotime.net> wrote:
> >> > On Thu, 28 Jul 2011 11:17:09 +1000 Stephen Rothwell wrote:
> >> >
> >> >> Hi Randy,
> >> >>
> >> >> On Wed, 27 Jul 2011 12:13:20 -0700 Randy Dunlap <rdunlap@xenotime.net> wrote:
> >> >> >
> >> >> > On Mon, 25 Jul 2011 09:10:03 -0700 Randy Dunlap wrote:
> >> >> >
> >> >> > > On Mon, 25 Jul 2011 09:37:53 +0200 Takashi Iwai wrote:
> >> >> > >
> >> >> > > > At Sun, 24 Jul 2011 12:26:24 -0700,
> >> >> > > > Randy Dunlap wrote:
> >> >> > > > >
> >> >> > > > > On Wed, 13 Jul 2011 21:33:30 +0200 Geert Uytterhoeven wrote:
> >> >> > > > >
> >> >> > > > > >   + src/sound/isa/ad1816a/ad1816a_lib.c: error: implicit declaration of function 'snd_dma_disable':  => 544
> >> >> > > > > >   + src/sound/isa/ad1816a/ad1816a_lib.c: error: implicit declaration of function 'snd_dma_pointer':  => 302
> >> >> > > > > >   + src/sound/isa/ad1816a/ad1816a_lib.c: error: implicit declaration of function 'snd_dma_program':  => 244
> >> >> > > > > >   + src/sound/isa/es1688/es1688_lib.c: error: implicit declaration of function 'snd_dma_pointer':  => 509
> >> >> > > > > >   + src/sound/isa/es1688/es1688_lib.c: error: implicit declaration of function 'snd_dma_program':  => 417
> >> >> > > > > >   + src/sound/isa/gus/gus_dma.c: error: implicit declaration of function 'snd_dma_disable':  => 177
> >> >> > > > > >   + src/sound/isa/gus/gus_dma.c: error: implicit declaration of function 'snd_dma_program':  => 79
> >> >> > > > > >   + src/sound/isa/gus/gus_pcm.c: error: implicit declaration of function 'snd_dma_pointer':  => 619
> >> >> > > > > >   + src/sound/isa/gus/gus_pcm.c: error: implicit declaration of function 'snd_dma_program':  => 591
> >> >> > > > > >   + src/sound/isa/sb/sb16_main.c: error: implicit declaration of function 'snd_dma_pointer':  => 456
> >> >> > > > > >   + src/sound/isa/sb/sb16_main.c: error: implicit declaration of function 'snd_dma_program':  => 276
> >> >> > > > > >   + src/sound/isa/sb/sb8_main.c: error: implicit declaration of function 'snd_dma_pointer':  => 425
> >> >> > > > > >   + src/sound/isa/sb/sb8_main.c: error: implicit declaration of function 'snd_dma_program':  => 172
> >> >> > > > > >   + src/sound/isa/sscape.c: error: implicit declaration of function 'snd_dma_program':  => 481
> >> >> > > > > >   + src/sound/isa/wss/wss_lib.c: error: implicit declaration of function 'snd_dma_disable':  => 1695
> >> >> > > > > >   + src/sound/isa/wss/wss_lib.c: error: implicit declaration of function 'snd_dma_pointer':  => 1160
> >> >> > > > > >   + src/sound/isa/wss/wss_lib.c: error: implicit declaration of function 'snd_dma_program':  => 1025
> >> >> > > > >
> >> >> > > > > All of these source files do #include <sound/core.h>,
> >> >> > > > > but these missing/implicit functions are conditional:
> >> >> > > > >
> >> >> > > > > #ifdef CONFIG_ISA_DMA_API
> >> >> > > > > #define DMA_MODE_NO_ENABLE    0x0100
> >> >> > > > >
> >> >> > > > > void snd_dma_program(unsigned long dma, unsigned long addr, unsigned int size, unsigned short mode);
> >> >> > > > > void snd_dma_disable(unsigned long dma);
> >> >> > > > > unsigned int snd_dma_pointer(unsigned long dma, unsigned int size);
> >> >> > > > > #endif
> >> >> > > > >
> >> >> > > > > so should these drivers all depend on ISA_DMA_API ?
> >> >> > > >
> >> >> > > > Yes, these are ISA sound cards, so they must have CONFIG_ISA and
> >> >> > > > CONFIG_ISA_DMA.  Although CONFIG_SND_SB8_DSP, CONFIG_SND_SB16_DSP
> >> >> > > > and CONFIG_SND_WSS_LIB have no direct dependency, but the drivers
> >> >> > > > selecting these are all in "if SND_ISA" in sound/isa/Kconfig.
> >> >> > >
> >> >> > > Hm, yes, they are.
> >> >> > >
> >> >> > > > > or should only pieces of each driver depend on ISA_DMA_API ?
> >> >> > > >
> >> >> > > > I wonder which configuration makes it possible.
> >> >> > > > Did you see a similar problem with randconfig?
> >> >> > >
> >> >> > > No.  This parisc allmodconfig file from linux-next builds has
> >> >> > > CONFIG_SND_ISA=y but it does not have CONFIG_ISA_DMA_API enabled.  Weird.
> >> >> > >
> >> >> > > Geert, do you know how this happens?
> >> >> > >
> >> >> > > parisc config:  http://kisskb.ellerman.id.au/kisskb/buildresult/4323458/config/
> >> >> > > from this failed build:  http://kisskb.ellerman.id.au/kisskb/buildresult/4323458/
> >> >> >
> >> >> > How is CONFIG_ISA_DMA_API enabled in parisc allmodconfig?
> >> >> > I don't see how that happens, and it's causing build errors...
> >> >>
> >> >> Sorry, but I don't understand the question.  The problam is that
> >> >> CONFIG_ISA_DMA_API is not enabled, right?  On parisc, it is just (from
> >> >> arch/parisc/Kconfig)
> >> >>
> >> >> config ISA_DMA_API
> >> >>       bool
> >> >>
> >> >> and nothing selects it.  However, I would have expected an allmodconfig
> >> >> to set it to y.  Clearly I don't understan how allmodconfig works any
> >> >> more :-(
> >> >>
> >> >> Michal, is my understanding correct?  Is this a bug in allmodconfig?
> >> >
> >> > When I do:
> >> > $ make ARCH=parisc O=PAR allmodconfig
> >> >
> >> > I don't get CONFIG_ISA_DMA_API enabled, so the question is simply
> >> > how is the linux-next build system ending up with CONFIG_ISA_DMA_API enabled?
> >> >
> >> Where do you see that "linux-next build system ending up with
> >> CONFIG_ISA_DMA_API enabled" ?
> >>
> >> From what I understood of the problem, the issue is that
> >> CONFIG_ISA_DMA_API does _not_ ends up enabled.
> >
> > Thanks for catching me.
> >
> > The problem (second try) is that ISA SND drivers are being built (on parisc)
> > even though ISA_DMA_API is not enabled -- and should not be enabled,
> > so the ISA SND drivers should not be built.  I think I have it right this time.
> >
> > From the parisc .config file listed above:
> >
> > CONFIG_SND_ISA=y
> > CONFIG_SND_ADLIB=m
> > CONFIG_SND_AD1816A=m
> > CONFIG_SND_AD1848=m
> > CONFIG_SND_ALS100=m
> > CONFIG_SND_AZT1605=m
> > CONFIG_SND_AZT2316=m
> > CONFIG_SND_AZT2320=m
> > CONFIG_SND_CMI8330=m
> >
> >
> > etc.  But none of these should be enabled.  AFAIK.
> > So how does this happen?
> >
> warning: (RADIO_MIROPCM20) selects SND_ISA which has unmet direct
> dependencies (SOUND && !M68K && SND && ISA && ISA_DMA_API)

Hm, I thought this was fixed by
commit 4756fc64ecd628efa6f8296711d7dbcefabf25dd 
Author: Ralf Baechle <ralf@linux-mips.org>
Date:   Fri Jun 24 10:30:09 2011 -0300

    [media] MEDIA: Fix non-ISA_DMA_API link failure of sound code
    
OK, it landed after 3.0-rc7, so this problem must be fixed now?
I'm not sure at which point this thread is focused on.


Takashi