Re: 1.0.16rc2 -build issues

Re: 1.0.16rc2 -build issues

[Alan Horstmann]

Hi Takashi,

I tried out yesterdays tarball hg20080207 to confirm DSP2000 Midi stuff before 
pointing another DSP2000 user to it.  Build proceeded OK but I did notice the 
following warnings, which I might have missed previously, or they could be 
new.  Do you care about redefinitions?

pcm_oss.c:639:1: warning: "AFMT_S32_LE" redefined
In file included from pcm_oss.c:43:
/lib/modules/2.4.21-99-athlon/build/include/linux/soundcard.h:562:1: warning: 
this is the location of the previous definition
pcm_oss.c:640:1: warning: "AFMT_S32_BE" redefined
/lib/modules/2.4.21-99-athlon/build/include/linux/soundcard.h:563:1: warning: 
this is the location of the previous definition
pcm_oss.c:641:1: warning: "AFMT_S24_LE" redefined
/lib/modules/2.4.21-99-athlon/build/include/linux/soundcard.h:558:1: warning: 
this is the location of the previous definition
pcm_oss.c:642:1: warning: "AFMT_S24_BE" redefined
/lib/modules/2.4.21-99-athlon/build/include/linux/soundcard.h:559:1: warning: 
this is the location of the previous definition

Regards

Alan

Re: 1.0.16rc2 -build issues

[Clemens Ladisch]

Alan Horstmann wrote:
> I tried out yesterdays tarball hg20080207 to confirm DSP2000 Midi stuff before 
> pointing another DSP2000 user to it.  Build proceeded OK but I did notice the 
> following warnings, which I might have missed previously, or they could be 
> new.  Do you care about redefinitions?
> 
> pcm_oss.c:639:1: warning: "AFMT_S32_LE" redefined
> In file included from pcm_oss.c:43:
> /lib/modules/2.4.21-99-athlon/build/include/linux/soundcard.h:562:1: warning: 
> this is the location of the previous definition

This warnings are caused by harmless whitespace in the 2.4.x soundcard.h
file.


Regards,
Clemens

1.0.16rc2

[Jaroslav Kysela]

Hi,

	1.0.16rc2 alsa-driver, alsa-lib, pyalsa packages are available for 
download. Full changelog from 1.0.16rc1 is on:

http://www.alsa-project.org/main/index.php/Changes_v1.0.16rc1_v1.0.16rc2

						Jaroslav

-----
Jaroslav Kysela <perex@perex.cz>
Linux Kernel Sound Maintainer
ALSA Project, Red Hat, Inc.

Re: 1.0.16rc2

[Alan Horstmann]

On Wednesday 30 January 2008 11:21, Takashi wrote:
> At Wed, 30 Jan 2008 10:09:15 +0000,
>
> Alan Horstmann wrote:
> >
> > I get the following compile error:
> >
> > /usr/share/alsa1.0.16rc2/alsa-driver-1.0.16rc2/include/adriver.h:1461:
> > error: redefinition of `__ffs'
> > /lib/modules/2.4.21-99-athlon/build/include/asm/bitops.h:410: error:
> > `__ffs' previously defined here make[1]: *** [hwdep.o] Error 1
> > make[1]: Leaving directory
> > `/usr/share/alsa1.0.16rc2/alsa-driver-1.0.16rc2/acore' make: ***
> > [compile] Error 1

> Could you apply the patch below and use ./hgcompile to configure?
>
> thanks,
>
> Takashi

Patch applied OK.  Unfortunately, hgcompile complains that the Autoconf 
version is too old on this system -needs >2.59 (looks like I have 1.4o).  
What is the difference using that rather than straight configure?  Is there a 
way to work round it?

Alan



> diff -r e600ecb8fbd7 configure.in
> --- a/configure.in	Wed Jan 30 08:36:00 2008 +0100
> +++ b/configure.in	Wed Jan 30 12:22:56 2008 +0100
> @@ -2597,6 +2597,37 @@ if test "$CONFIG_HAVE_INIT_UTSNAME" = "1
>    AC_DEFINE(CONFIG_HAVE_INIT_UTSNAME)
>  fi
>
> +dnl __ffs support?
> +if test "$kversion.$kpatchlevel" = "2.6"; then
> +  CONFIG_HAVE_FFS=1
> +else
> +  AC_MSG_CHECKING(for __ffs)
> +  init_utsname="0"
> +  ac_save_CFLAGS="$CFLAGS"
> +  ac_save_CC="$CC"
> +  CFLAGS="$KERNEL_CHECK_CFLAGS"
> +  CC=$KCC
> +  AC_TRY_COMPILE([
> +#define __KERNEL__
> +#include <linux/config.h>
> +#include <linux/bitops.h>
> +],[
> +  unsigned long (*x)(unsigned long);
> +  x = __ffs;
> +],
> +    AC_MSG_RESULT(yes);have_ffs="1",
> +    AC_MSG_RESULT(no);have_ffs="0",
> +    AC_MSG_RESULT(unknown);have_ffs="0"
> +)
> +  CFLAGS=$ac_save_CFLAGS
> +  CC=$ac_save_CC
> +  CONFIG_HAVE_FFS=$have_ffs
> +fi
> +dnl AC_SUBST(CONFIG_HAVE_FFS)
> +if test "$CONFIG_HAVE_FFS" = "1"; then
> +  AC_DEFINE(CONFIG_HAVE_FFS)
> +fi
> +
>  dnl PnP supports suspend/resume?
>  if test "$CONFIG_PNP_KERNEL" = "y"; then
>    AC_MSG_CHECKING(for PnP suspend/resume)
> diff -r e600ecb8fbd7 include/adriver.h
> --- a/include/adriver.h	Wed Jan 30 08:36:00 2008 +0100
> +++ b/include/adriver.h	Wed Jan 30 12:22:56 2008 +0100
> @@ -1455,7 +1455,7 @@ static inline unsigned char snd_pci_revi
>  #endif
>  #endif
>
> -#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 5, 3)
> +#ifndef CONFIG_HAVE_FFS
>  #if defined(__i386__)
>  static inline unsigned long __ffs(unsigned long word)
>  {
> diff -r e600ecb8fbd7 include/config.h.in
> --- a/include/config.h.in	Wed Jan 30 08:36:00 2008 +0100
> +++ b/include/config.h.in	Wed Jan 30 12:22:56 2008 +0100
> @@ -44,6 +44,7 @@
>  #undef CONFIG_HAVE_TTY_COUNT_ATOMIC
>  #undef CONFIG_HAVE_VIDEO_GET_DRVDATA
>  #undef CONFIG_HAVE_DUMP_STACK
> +#undef CONFIG_HAVE_FFS
>
>  /* 2.2 kernels */
>  #undef CONFIG_OLD_KILL_FASYNC

Re: 1.0.16rc2

[Takashi Iwai]

At Wed, 30 Jan 2008 12:41:46 +0000,
Alan Horstmann wrote:
> 
> On Wednesday 30 January 2008 11:21, Takashi wrote:
> > At Wed, 30 Jan 2008 10:09:15 +0000,
> >
> > Alan Horstmann wrote:
> > >
> > > I get the following compile error:
> > >
> > > /usr/share/alsa1.0.16rc2/alsa-driver-1.0.16rc2/include/adriver.h:1461:
> > > error: redefinition of `__ffs'
> > > /lib/modules/2.4.21-99-athlon/build/include/asm/bitops.h:410: error:
> > > `__ffs' previously defined here make[1]: *** [hwdep.o] Error 1
> > > make[1]: Leaving directory
> > > `/usr/share/alsa1.0.16rc2/alsa-driver-1.0.16rc2/acore' make: ***
> > > [compile] Error 1
> 
> > Could you apply the patch below and use ./hgcompile to configure?
> >
> > thanks,
> >
> > Takashi
> 
> Patch applied OK.  Unfortunately, hgcompile complains that the Autoconf 
> version is too old on this system -needs >2.59 (looks like I have 1.4o).  
> What is the difference using that rather than straight configure?

You need to regenerate the configure script itself.

>  Is there a 
> way to work round it?

Remove the line "AC_PREREQ(2.59)" in configure.in.


Takashi

Re: 1.0.16rc2 -build issues

[Alan Horstmann]

On Wednesday 30 January 2008 13:03, you wrote:
> At Wed, 30 Jan 2008 12:41:46 +0000,
>
> Alan Horstmann wrote:
> > On Wednesday 30 January 2008 11:21, Takashi wrote:
> > >
> > > Could you apply the patch below and use ./hgcompile to configure?
> > >
> > > thanks,
> > >
> > > Takashi
> >
> > Patch applied OK.  Unfortunately, hgcompile complains that the Autoconf
> > version is too old on this system -needs >2.59 (looks like I have 1.4o).
> > What is the difference using that rather than straight configure?
>
> You need to regenerate the configure script itself.
>
> >  Is there a
> > way to work round it?
>
> Remove the line "AC_PREREQ(2.59)" in configure.in.

Thanks.  However, I could not get (the old) autoconf to produce a working configure script due to 'AS_HELP_STRING', so I added
	#define CONFIG_HAVE_FFS
to the patched adriver.h and used the original configure script, to proceed for now (which it did).

Unfortunately make didn't get much further, having more ??historic kernel?? trouble in memalloc:

memalloc.c: In function `snd_mem_proc_open':
memalloc.c:611: warning: implicit declaration of function `single_open'
memalloc.c: At top level:
memalloc.c:715: error: `single_release' undeclared here (not in a function)
memalloc.c:715: error: initializer element is not constant
memalloc.c:715: error: (near initialization for `snd_mem_proc_fops.release')
memalloc.c: In function `snd_mem_init':
memalloc.c:731: warning: assignment discards qualifiers from pointer target type

Sorry if this is causing a lot of extra trouble.

Regards  Alan

Re: 1.0.16rc2 -build issues

[Takashi Iwai]

At Wed, 30 Jan 2008 21:03:45 +0000,
Alan Horstmann wrote:
> 
> On Wednesday 30 January 2008 13:03, you wrote:
> > At Wed, 30 Jan 2008 12:41:46 +0000,
> >
> > Alan Horstmann wrote:
> > > On Wednesday 30 January 2008 11:21, Takashi wrote:
> > > >
> > > > Could you apply the patch below and use ./hgcompile to configure?
> > > >
> > > > thanks,
> > > >
> > > > Takashi
> > >
> > > Patch applied OK.  Unfortunately, hgcompile complains that the Autoconf
> > > version is too old on this system -needs >2.59 (looks like I have 1.4o).
> > > What is the difference using that rather than straight configure?
> >
> > You need to regenerate the configure script itself.
> >
> > >  Is there a
> > > way to work round it?
> >
> > Remove the line "AC_PREREQ(2.59)" in configure.in.
> 
> Thanks.  However, I could not get (the old) autoconf to produce a working configure script due to 'AS_HELP_STRING', so I added
> 	#define CONFIG_HAVE_FFS
> to the patched adriver.h and used the original configure script, to proceed for now (which it did).
> 
> Unfortunately make didn't get much further, having more ??historic kernel?? trouble in memalloc:
> 
> memalloc.c: In function `snd_mem_proc_open':
> memalloc.c:611: warning: implicit declaration of function `single_open'
> memalloc.c: At top level:
> memalloc.c:715: error: `single_release' undeclared here (not in a function)
> memalloc.c:715: error: initializer element is not constant
> memalloc.c:715: error: (near initialization for `snd_mem_proc_fops.release')
> memalloc.c: In function `snd_mem_init':
> memalloc.c:731: warning: assignment discards qualifiers from pointer target type

Then the patch below should work.  Give it a try.


Takashi

---

diff -r e600ecb8fbd7 acore/memalloc.patch
--- a/acore/memalloc.patch	Wed Jan 30 08:36:00 2008 +0100
+++ b/acore/memalloc.patch	Thu Jan 31 12:39:11 2008 +0100
@@ -189,7 +189,7 @@
  
  
  #ifdef CONFIG_PROC_FS
-+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
  /*
   * proc file interface
   */
@@ -205,7 +205,7 @@
  static int __init snd_mem_init(void)
  {
  #ifdef CONFIG_PROC_FS
-+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
  	snd_mem_proc = create_proc_entry(SND_MEM_PROC_FILE, 0644, NULL);
  	if (snd_mem_proc)
  		snd_mem_proc->proc_fops = &snd_mem_proc_fops;
@@ -216,7 +216,7 @@
  
  static void __exit snd_mem_exit(void)
  {
-+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
  	remove_proc_entry(SND_MEM_PROC_FILE, NULL);
 +#endif
  	free_all_reserved_pages();

Re: 1.0.16rc2 -build issues

[Alan Horstmann]

On Thursday 31 January 2008 11:38, you wrote:
> At Wed, 30 Jan 2008 21:03:45 +0000,
>
> Alan Horstmann wrote:
> > Unfortunately make didn't get much further, having more ??historic
> > kernel?? trouble in memalloc:
> >
> > memalloc.c: In function `snd_mem_proc_open':
> > memalloc.c:611: warning: implicit declaration of function `single_open'
> > memalloc.c: At top level:
> > memalloc.c:715: error: `single_release' undeclared here (not in a
> > function) memalloc.c:715: error: initializer element is not constant
> > memalloc.c:715: error: (near initialization for
> > `snd_mem_proc_fops.release') memalloc.c: In function `snd_mem_init':
> > memalloc.c:731: warning: assignment discards qualifiers from pointer
> > target type
>
> Then the patch below should work.  Give it a try.

Yup, that nailed it!  Make completes OK now, thanks.

I applied the memalloc.patch patch by hand.  If this is to be applied, notice 
that Line 200 #endif text string also presumably needs changing.

Alsa-lib was OK, but the configure was slightly messy in reporting 
'python-config' Command not found  before a more reasonable message and 
disabling python from the build.

If both the patches are to be applied, would you like me to retest the nightly 
build at some stage(which presumably won't need hgcompile)?  If so can you 
give the link again.

Thanks 

Alan

> ---
>
> diff -r e600ecb8fbd7 acore/memalloc.patch
> --- a/acore/memalloc.patch	Wed Jan 30 08:36:00 2008 +0100
> +++ b/acore/memalloc.patch	Thu Jan 31 12:39:11 2008 +0100
> @@ -189,7 +189,7 @@
>
>
>   #ifdef CONFIG_PROC_FS
> -+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
> ++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
>   /*
>    * proc file interface
>    */
> @@ -205,7 +205,7 @@
>   static int __init snd_mem_init(void)
>   {
>   #ifdef CONFIG_PROC_FS
> -+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
> ++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
>   	snd_mem_proc = create_proc_entry(SND_MEM_PROC_FILE, 0644, NULL);
>   	if (snd_mem_proc)
>   		snd_mem_proc->proc_fops = &snd_mem_proc_fops;
> @@ -216,7 +216,7 @@
>
>   static void __exit snd_mem_exit(void)
>   {
> -+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 4, 0)
> ++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
>   	remove_proc_entry(SND_MEM_PROC_FILE, NULL);
>  +#endif
>   	free_all_reserved_pages();