Skip to contentSkip to conversation

broken alsa-lib

Buildroot Daniel Mack 3 participants 18 messages Feb 22, 2009 – Mar 2, 2009

lore
Share & tools
lore
#1 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Daniel" == Daniel Mack <[email protected]> writes:

Daniel> Hi,
Daniel> as pointed out in two threads on the alsa-dev ML, there seems to be Daniel> general problem with ALSA libraries built by buildroot:

Daniel> http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/014999.html
Daniel> http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/015005.html

Alsa-lib had an issue with !LARGEFILE support that we fixed recently and pushed upstream, but other than that I don't remember any bugreports.

You mentioned something about 1.0.19, but buildroot is still using 1.0.18, are you using some local changes?

Signature · 1 line
Bye, Peter Korsgaard
#2 lore raw
Quoted 9 lines · Peter Korsgaard

Daniel> as pointed out in two threads on the alsa-dev ML, there seems to be Daniel> general problem with ALSA libraries built by buildroot:

Daniel> http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/014999.html
Daniel> http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/015005.html

Alsa-lib had an issue with !LARGEFILE support that we fixed recently and pushed upstream, but other than that I don't remember any bugreports.

Hmm, what I see is certainly not a LARGEFILE issue.

Quoted 2 lines

You mentioned something about 1.0.19, but buildroot is still using 1.0.18, are you using some local changes?

Yes, true, forgot to mention that. We locally patched to use 1.0.19 but that was merely because I hoped to fix the problem by that - the same problem exists with 1.0.18. And I got this problem since a long time and always patched my way around it (by overriding the value with some hard-coded constant) as I thought it is a minor issue that will be fixed sooner or later. But as it turns out, the problem seems to be bigger and might also have other impact than just that single failing function call.

I'll build alsa-lib and alsa-utils with OE tomorrow and see whether that makes any difference.

Daniel

#3 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Daniel" == Daniel Mack <[email protected]> writes:

Hi,

Quoted 3 lines
Quoted 3 lines

Alsa-lib had an issue with !LARGEFILE support that we fixed recently and pushed upstream, but other than that I don't remember any bugreports.

Daniel> Hmm, what I see is certainly not a LARGEFILE issue.

Probably not, just mentioning the only bugs we've found in alsa lib recently.

Quoted 2 lines
Quoted 2 lines

You mentioned something about 1.0.19, but buildroot is still using 1.0.18, are you using some local changes?

Daniel> Yes, true, forgot to mention that. We locally patched to use Daniel> 1.0.19 but that was merely because I hoped to fix the problem Daniel> by that - the same problem exists with 1.0.18. And I got this Daniel> problem since a long time and always patched my way around it Daniel> (by overriding the value with some hard-coded constant) as I Daniel> thought it is a minor issue that will be fixed sooner or Daniel> later. But as it turns out, the problem seems to be bigger Daniel> and might also have other impact than just that single Daniel> failing function call.

Please report those kind of problems to us so we have a chance to fix them. This is the first I hear of it.

Daniel> I'll build alsa-lib and alsa-utils with OE tomorrow and see Daniel> whether that makes any difference.

Great, Please send us the config.log from OE and BR if it works under OE.

Signature · 1 line
Bye, Peter Korsgaard
#10 lore raw
In reply to Daniel Mack
Quoted 9 lines · Daniel Mack

Hi,

as pointed out in two threads on the alsa-dev ML, there seems to be general problem with ALSA libraries built by buildroot:

	http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/014999.html
	http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/015005.html

Any hints, anyone?

Not sure if this is the same as a problem I was seeing with alsa-lib. It was a while ago so not entirely sure. However I found my problem was to do with the complicated alsa versioning API meaning that the wrong version of the function was called. I fixed it by adding --with-versioned="no" to the alsa-lib configure (I also moved to 1.0.19 as that had a number of bug fixes for other things I needed).

Hope that helps,

Will

Signature · 5 lines
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7371 2032
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------
#11 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Will" == Will Wagner <[email protected]> writes:

Hi,

Will> Not sure if this is the same as a problem I was seeing with Will> alsa-lib. It was a while ago so not entirely sure. However I found my Will> problem was to do with the complicated alsa versioning API meaning Will> that the wrong version of the function was called. I fixed it by Will> adding --with-versioned="no" to the alsa-lib configure (I also moved Will> to 1.0.19 as that had a number of bug fixes for other things I Will> needed).

Will> Hope that helps,

Care to send a patch for both?

Signature · 1 line
Bye, Peter Korsgaard
#17 lore raw
In reply to Will Wagner
Quoted 6 lines · Will Wagner
Quoted 6 lines

as pointed out in two threads on the alsa-dev ML, there seems to be general problem with ALSA libraries built by buildroot:

	http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/014999.html
	http://mailman.alsa-project.org/pipermail/alsa-devel/2009-February/015005.html
Quoted 8 lines
Quoted 1 line

Any hints, anyone?

Not sure if this is the same as a problem I was seeing with alsa-lib. It was a while ago so not entirely sure. However I found my problem was to do with the complicated alsa versioning API meaning that the wrong version of the function was called. I fixed it by adding --with-versioned="no" to the alsa-lib configure (I also moved to 1.0.19 as that had a number of bug fixes for other things I needed).

Aha, that fixed it for me, too. Thanks :)

Daniel

#12 lore raw
In reply to Peter Korsgaard
Quoted 3 lines

Care to send a patch for both?

I will, might take me a while to get it into shape as we have made a number of changes.

Signature · 5 lines
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7371 2032
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------
#14 lore raw
In reply to Peter Korsgaard
Quoted 11 lines · Peter Korsgaard

Will> Not sure if this is the same as a problem I was seeing with Will> alsa-lib. It was a while ago so not entirely sure. However I found my Will> problem was to do with the complicated alsa versioning API meaning Will> that the wrong version of the function was called. I fixed it by Will> adding --with-versioned="no" to the alsa-lib configure (I also moved Will> to 1.0.19 as that had a number of bug fixes for other things I Will> needed).

Will> Hope that helps,

Care to send a patch for both?

We'd need to find out if the avr-inline patch is still needed for 1.0.19 or if that's been fixed upstream now. Apart from that, I rebased the fix-ott_t-patch.

Can anyone with a AVR32 environment try to build and use 1.0.19 without that patch?

Daniel

#13 lore raw
In reply to Will Wagner
Quoted 5 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Will" == Will Wagner <[email protected]> writes:

Care to send a patch for both?

Will> I will, might take me a while to get it into shape as we have made a number of changes.

Ok, thanks.

Signature · 1 line
Bye, Peter Korsgaard
#15 lore raw
In reply to Daniel Mack
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Daniel" == Daniel Mack <[email protected]> writes:

Hi,

Daniel> We'd need to find out if the avr-inline patch is still needed Daniel> for 1.0.19 or if that's been fixed upstream now. Apart from Daniel> that, I rebased the fix-ott_t-patch.

That was afaik committed upstream - Didn't it make it for 1.0.19?

Daniel> Can anyone with a AVR32 environment try to build and use Daniel> 1.0.19 without that patch?

Yes please.

Signature · 1 line
Bye, Peter Korsgaard
#16 lore raw
Quoted 6 lines

Daniel> We'd need to find out if the avr-inline patch is still needed Daniel> for 1.0.19 or if that's been fixed upstream now. Apart from Daniel> that, I rebased the fix-ott_t-patch.

That was afaik committed upstream - Didn't it make it for 1.0.19?

I don't believe it made it for 1.0.19. I also submitted a couple of others things that waiting on 1.0.20

Signature · 5 lines
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7371 2032
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------
#4 lore raw
In reply to Peter Korsgaard
Quoted 10 lines · Peter Korsgaard

Daniel> Yes, true, forgot to mention that. We locally patched to use Daniel> 1.0.19 but that was merely because I hoped to fix the problem Daniel> by that - the same problem exists with 1.0.18. And I got this Daniel> problem since a long time and always patched my way around it Daniel> (by overriding the value with some hard-coded constant) as I Daniel> thought it is a minor issue that will be fixed sooner or Daniel> later. But as it turns out, the problem seems to be bigger Daniel> and might also have other impact than just that single Daniel> failing function call.

Quoted 2 lines

Please report those kind of problems to us so we have a chance to fix them. This is the first I hear of it.

I'm playing ping-pong between the two lists now, and over there they pointed out that this was indeed reported quite a while ago:

	http://lists.busybox.net/pipermail/buildroot/2008-July/009884.html

Anyway, let's make sure to apply this patch asap :)

Daniel

#5 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Daniel" == Daniel Mack <[email protected]> writes:

Hi,

Daniel> I'm playing ping-pong between the two lists now, and over Daniel> there they pointed out that this was indeed reported quite a Daniel> while ago:

Daniel> http://lists.busybox.net/pipermail/buildroot/2008-July/009884.html

Ahh, nothing happened to that mail as the submitter never replied to Daniel Laird's mail.

Daniel> Anyway, let's make sure to apply this patch asap :)

So what exactly needs to get committed? The --with-versioned=no is easy, but should the:

ifeq ($(BR2_arm),y)
ALSA_LIB_CFLAGS+=-mabi=aapcs-linux
endif

Be in an EABI conditional?

Signature · 1 line
Bye, Peter Korsgaard
#6 lore raw
Quoted 6 lines · Peter Korsgaard

Daniel> I'm playing ping-pong between the two lists now, and over Daniel> there they pointed out that this was indeed reported quite a Daniel> while ago:

Daniel> http://lists.busybox.net/pipermail/buildroot/2008-July/009884.html

Quoted 7 lines

Ahh, nothing happened to that mail as the submitter never replied to Daniel Laird's mail.

Daniel> Anyway, let's make sure to apply this patch asap :)

So what exactly needs to get committed? The --with-versioned=no is easy, but should the:

This one is certainly needed.

Quoted 5 lines

ifeq ($(BR2_arm),y)
ALSA_LIB_CFLAGS+=-mabi=aapcs-linux
endif

Be in an EABI conditional?

I'm not aware of any ABI problem with that, so I can't say.

Daniel

#7 lore raw
In reply to Peter Korsgaard
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Peter" == Peter Korsgaard <[email protected]> writes:

Hi,

Peter> So what exactly needs to get committed? The --with-versioned=no is Peter> easy, but should the:

Peter> ifeq ($(BR2_arm),y)
Peter> ALSA_LIB_CFLAGS+=-mabi=aapcs-linux
Peter> endif

Peter> Be in an EABI conditional?

That hunk just seems wrong to me. We always add -mabi=<> on ARM, so it's a noop on EABI, and who knows what happens on OABI?

Ulf, why did you add it in the first place?

Signature · 1 line
Bye, Peter Korsgaard
#8 lore raw
Quoted 11 lines · Peter Korsgaard

Peter> So what exactly needs to get committed? The --with-versioned=no is Peter> easy, but should the:

Peter> ifeq ($(BR2_arm),y)
Peter> ALSA_LIB_CFLAGS+=-mabi=aapcs-linux
Peter> endif

Peter> Be in an EABI conditional?

That hunk just seems wrong to me. We always add -mabi=<> on ARM, so it's a noop on EABI, and who knows what happens on OABI?

This one wasn't commited, right? With --with-versioned=no things are fine now for EABI, but I don't know about the legacy case.

Thanks,
Daniel

#9 lore raw
Quoted 4 lines
Quoted 4 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Daniel" == Daniel Mack <[email protected]> writes:

That hunk just seems wrong to me. We always add -mabi=<> on ARM, so it's a noop on EABI, and who knows what happens on OABI?

Daniel> This one wasn't commited, right? With --with-versioned=no things are Daniel> fine now for EABI, but I don't know about the legacy case.

No it wasn't committed. And yeah, it should be a noop with eabi as we already add -mabi=aapcs-linux to CFLAGS for eabi builds, but it seems very wrong for oabi builds.

Signature · 1 line
Bye, Peter Korsgaard