audio problems with -current/GCC4?

NAVIGATION

CATEGORIES

REFERRENCE

If you have some good idea, Cilck and mail to us please...

LINKS

Home » BSD

audio problems with -current/GCC4?
5 answers - 1444 bytes -
Cross-posted to each list because it seems relevant in each case
I've been tracking i386-current for some time and have been successfully
using 3.99.21 built with GCC4 since shortly after it was imported into
the tree.
Most, if not all of my packages, however, were built with GCC3.3.3 that
had been the standard compiler. They all played very happily together.
When pkgsrc tagged the pkgsrc-2006Q2 branch, I decided I would blow away
all my packages on one particular machine, update pkgsrc on that tag and
build everything from scratch. As far a building goes, I've only barely
started, but only had one glitch that I've worked around and filed a PR
against.
Then I re-built audio/xmms. The build itself went without error. Un-
fortunately, when I try to play any MP3 audio file or stream with it, all
I get is noise. It otherwise appears to be operating normally.
Since the previous incarnation of xmms was built with GCC3.3.3 and worked
great with my GCC4-built kernel/userland, perhaps there is some problem
with GCC4-built audio applications? I haven't had a chance to try any
others yet, but I did expect xmms to 'just work'. (I need to collect
some samples to try with the base userland 'audioplay' tool)
Has anyone else had problems with audio on i386 when switching from
GCC3.3.3 to GCC4.1.2 to build packages?
No.1 | | 1695 bytes | |
Sat, 8 Jul 2006 23:25:56 -0500 (CDT)
"John D. Baker" <jdbaker (AT) mylinuxisp (DOT) comwrote:

Cross-posted to each list because it seems relevant in each case

I've been tracking i386-current for some time and have been successfully
using 3.99.21 built with GCC4 since shortly after it was imported into
the tree.

Most, if not all of my packages, however, were built with GCC3.3.3 that
had been the standard compiler. They all played very happily together.

When pkgsrc tagged the pkgsrc-2006Q2 branch, I decided I would blow away
all my packages on one particular machine, update pkgsrc on that tag and
build everything from scratch. As far a building goes, I've only barely
started, but only had one glitch that I've worked around and filed a PR
against.

Then I re-built audio/xmms. The build itself went without error. Un-
fortunately, when I try to play any MP3 audio file or stream with it, all
I get is noise. It otherwise appears to be operating normally.

Since the previous incarnation of xmms was built with GCC3.3.3 and worked
great with my GCC4-built kernel/userland, perhaps there is some problem
with GCC4-built audio applications? I haven't had a chance to try any
others yet, but I did expect xmms to 'just work'. (I need to collect
some samples to try with the base userland 'audioplay' tool)

Has anyone else had problems with audio on i386 when switching from
GCC3.3.3 to GCC4.1.2 to build packages?

Yes, I had same problem with bmp (pkgsrc/audio/bmp) and GCC4.
But the problem dissapeared when I built it without -march.
No.2 | | 1676 bytes | |
Sat, Jul 08, 2006 at 11:25:56PM -0500, John D. Baker wrote:
Cross-posted to each list because it seems relevant in each case

I've been tracking i386-current for some time and have been successfully
using 3.99.21 built with GCC4 since shortly after it was imported into
the tree.

Most, if not all of my packages, however, were built with GCC3.3.3 that
had been the standard compiler. They all played very happily together.

When pkgsrc tagged the pkgsrc-2006Q2 branch, I decided I would blow away
all my packages on one particular machine, update pkgsrc on that tag and
build everything from scratch. As far a building goes, I've only barely
started, but only had one glitch that I've worked around and filed a PR
against.

Then I re-built audio/xmms. The build itself went without error. Un-
fortunately, when I try to play any MP3 audio file or stream with it, all
I get is noise. It otherwise appears to be operating normally.

Since the previous incarnation of xmms was built with GCC3.3.3 and worked
great with my GCC4-built kernel/userland, perhaps there is some problem
with GCC4-built audio applications? I haven't had a chance to try any
others yet, but I did expect xmms to 'just work'. (I need to collect
some samples to try with the base userland 'audioplay' tool)

Has anyone else had problems with audio on i386 when switching from
GCC3.3.3 to GCC4.1.2 to build packages?

Same problem here with audio/bmp, although mpg321 works just fine. I didn't
know building without -march could fix the problem, as xtraeme@ mentioned :)
No.3 | | 443 bytes | |
Sun, Jul 09, 2006 at 12:30:07PM +0000, Julio M. Merino Vidal wrote:
Same problem here with audio/bmp, although mpg321 works just fine. I
didn't know building without -march could fix the problem, as xtraeme@
mentioned :)

mpg321 uses fixed-point math (libmad), xmms (and i guess bmp too) uses
floating-point math (mpg123). this would suggest that GCC4 screws up
floating-point operations with -march.

regards,
No.4 | | 590 bytes | |
Sun, 9 Jul 2006 09:07:18 +0200
Juan RP <juan (AT) xtrarom (DOT) orgwrote:

Has anyone else had problems with audio on i386 when switching from
GCC3.3.3 to GCC4.1.2 to build packages?

Just now, also with bmp.

Yes, I had same problem with bmp (pkgsrc/audio/bmp) and GCC4.
But the problem dissapeared when I built it without -march.

Really? Is adding 'rm-optarg:-march rm-optarg:-mcpu' to BUILDLINK_TRANSFRM a right way to do that? It doesn't help for me. Rebuilding with pkgsrc's gcc3 works, though.

dieter
No.5 | | 1098 bytes | |
Sun, 9 Jul 2006, Julio M. Merino Vidal wrote:

Sat, Jul 08, 2006 at 11:25:56PM -0500, John D. Baker wrote:
Has anyone else had problems with audio on i386 when switching from
GCC3.3.3 to GCC4.1.2 to build packages?

Same problem here with audio/bmp, although mpg321 works just fine. I didn't
know building without -march could fix the problem, as xtraeme@ mentioned :)

In my case, I only used -march in my kernel config:

makeoptions CPUFLAGS+="-march=athlon -mmmx -msse"

(IIRC, I don't have access to the machine right now). It (or rather
'they' have a removable disk carrier that I can move between a number
of machines) are all older Athlon-based machines of various speeds.
The 550MHz and 1300MHz machines all have VIA ASICs/audio. The 1410MHz
machine has Acer ASICs and cmpci audio.

I didn't use -march anywhere else not in my /etc/mk.conf.

Does the pkgsrc machinery attempt to deduce what it thinks is an optimum
setting of "-march", or is it capable of pulling it out of the kernel
anywhere?

Re: audio problems with -current/GCC4?

EMSDN.COM

It's a question for 'audio problems with -current/GCC4?' under 'BSD'

BSD

audio problems with -current/GCC4?