Michael Smith wrote:
> On Sun, 09 Jan 2005 18:15:26 +0100, Frits Letteboer
> <graver@graver.xs4all.nl> wrote:
>
>>Hi,
>>
>>I'm currently migrating from Shoutcast/MP3 to Icecast/OGG/Vorbis
using ices.
>>
>>Unfortunately, when initialising the encoder, ices crashes with a bus
error.
>
>
> It looks from the gdb output that it's crashing in libvorbis. Can you
> compile libvorbis AND ices with debugging symbols, and try again?
Program received signal SIGBUS, Bus error.
[Switching to Thread 1 (LWP 100163)]
0x2819e486 in _vp_couple (blobno=7, g=0x80adb34, p=0x80c8334, vi=0x80b0000,
res=0x805c960, mag_memo=0x8105930, mag_sort=0x8105960,
ifloor=0x8105870,
nonzero=0xbfaedc70, sliding_lowpass=574) at psy.c:1167
1167 for(k=0;k<partition;k++){
>
> Since you appear to be using x86, this is a surprising error (bus
> faults are much harder to trigger on x86 than on many other
> architectures).
> Does changing the quality level change anything?
No, I tried several, all crashed.
On Thu, 13 Jan 2005 14:43:01 +0100, Frits Letteboer <graver@graver.xs4all.nl> wrote:> Michael Smith wrote: > > On Sun, 09 Jan 2005 18:15:26 +0100, Frits Letteboer > > <graver@graver.xs4all.nl> wrote: > > > >>Hi, > >> > >>I'm currently migrating from Shoutcast/MP3 to Icecast/OGG/Vorbis > using ices. > >> > >>Unfortunately, when initialising the encoder, ices crashes with a bus > error. > > > > > > It looks from the gdb output that it's crashing in libvorbis. Can you > > compile libvorbis AND ices with debugging symbols, and try again? > > Program received signal SIGBUS, Bus error. > [Switching to Thread 1 (LWP 100163)] > 0x2819e486 in _vp_couple (blobno=7, g=0x80adb34, p=0x80c8334, vi=0x80b0000, > res=0x805c960, mag_memo=0x8105930, mag_sort=0x8105960, > ifloor=0x8105870, > nonzero=0xbfaedc70, sliding_lowpass=574) at psy.c:1167 > 1167 for(k=0;k<partition;k++){That's very, very strange. I'd be very tempted to say that that line couldn't possibly cause a bus fault (at least on x86)... but you seem to have evidence to the contrary. Can you reproduce this with any other vorbis encoders, using the same version of libvorbis? Can you try rebuilding libvorbis and ices with the optimisation level reduced (or even turned off entirely)? I'd expect that to "fix" the problem, but it might just be hiding something seriously wrong. Sorry I can't give any more helpful advice - this is something I've never seen before, and I really don't have much idea what could have gone wrong. Mike
Michael Smith wrote:> On Thu, 13 Jan 2005 14:43:01 +0100, Frits Letteboer > <graver@graver.xs4all.nl> wrote: > >>Michael Smith wrote: >> > On Sun, 09 Jan 2005 18:15:26 +0100, Frits Letteboer >> > <graver@graver.xs4all.nl> wrote: >> > >> >>Hi, >> >> >> >>I'm currently migrating from Shoutcast/MP3 to Icecast/OGG/Vorbis >>using ices. >> >> >> >>Unfortunately, when initialising the encoder, ices crashes with a bus >>error. >> > >> > >> > It looks from the gdb output that it's crashing in libvorbis. Can you >> > compile libvorbis AND ices with debugging symbols, and try again? >> >>Program received signal SIGBUS, Bus error. >>[Switching to Thread 1 (LWP 100163)] >>0x2819e486 in _vp_couple (blobno=7, g=0x80adb34, p=0x80c8334, vi=0x80b0000, >> res=0x805c960, mag_memo=0x8105930, mag_sort=0x8105960, >>ifloor=0x8105870, >> nonzero=0xbfaedc70, sliding_lowpass=574) at psy.c:1167 >>1167 for(k=0;k<partition;k++){ > > > > That's very, very strange. I'd be very tempted to say that that line > couldn't possibly cause a bus fault (at least on x86)... but you seem > to have evidence to the contrary. > > Can you reproduce this with any other vorbis encoders, using the same > version of libvorbis? > > Can you try rebuilding libvorbis and ices with the optimisation level > reduced (or even turned off entirely)? I'd expect that to "fix" the > problem, but it might just be hiding something seriously wrong. > > Sorry I can't give any more helpful advice - this is something I've > never seen before, and I really don't have much idea what could have > gone wrong. > > MikeAfter discussing things off-list with Karl, things seems to be caused by libvorbis combined with -march=pentium4 and -O and gcc (GCC) 3.4.2 [FreeBSD] 20040728