[vorbis] Short block test

[Nils Philippsen]

On Fri, 22 Dec 2000, Nils Philippsen wrote:

> On Fri, 22 Dec 2000, Segher Boessenkool wrote:
>
> >
> >
> > Nils Philippsen wrote:
> > >
> > > On Thu, 21 Dec 2000, Monty wrote:
> > >
> > > > > > I couldn't use mode -b 350 because oggenc always hangs at 98% with
> > > > > > this clip and the file produced has annoying distorsions...
> > > >
> > > > Ahhhh!  Someone else reported this bug and I could not reproduce it.
> > > > Is this with Frank's test sample?  Windows oggenc or unix?  If I can
> > > > get things to hang on Linux where I can watch it, I'll be very happy.
> > >
> > > I can reproduce it here:
> > >
> > > - libogg, libvorbis, vorbis-tools from CVS 20001220
> > > - Red Hat Linux 7/IA32 (heavily updated)
> > > - glibc-2.2
> > > - kernel 2.4.0-test12 (if that matters, it shouldn't)
> > >
> > > It runs fine until 98.3% where it hangs -- apparently in an endless loop
> > > because I can't strace any system calls being executed. The resulting ogg
> > > file has some really nasty artifacts...
> >
> > Can you compile it without math function inlining, and then ltrace it? That
> > would be very slow, of course, but would tell us about where the bug is
> > (if we're a bit lucky).
> >
> > Or connect to the running encoder with gdb, and get a backtrace.
>
> Okie:
>
> #0  0x40038f1f in Newton_Raphson_Maehly () from /usr/lib/libvorbis.so.0
> #1  0x40039159 in vorbis_lpc_to_lsp () from /usr/lib/libvorbis.so.0
> #2  0x4003e02e in floor0_forward () from /usr/lib/libvorbis.so.0
> #3  0x4003fe28 in mapping0_forward () from /usr/lib/libvorbis.so.0
> #4  0x4003963c in vorbis_analysis () from /usr/lib/libvorbis.so.0
> #5  0x804a870 in strcpy () at ../sysdeps/generic/strcpy.c:31
> #6  0x80497a5 in strcpy () at ../sysdeps/generic/strcpy.c:31
> #7  0x400cbfd1 in __libc_start_main (main=0x8049140 <strcpy+280>, argc=4,
>     ubp_av=0xbffff83c, init=0x8048cf0 <_init>, fini=0x804ae2c <_fini>,
>     rtld_fini=0x4000e254 <_dl_fini>, stack_end=0xbffff834)
>     at ../sysdeps/generic/libc-start.c:118
>
> This is against the original libs, if I compile libvorbis as you suggested
> (with -fno-inline or make debug), it doesn't show this error, so this
> might as well be a compiler bug regarding inlined functions. I run
> gcc-2.96-69 (AKA "the compiler that spoilt world peace" :-)) for that

That should have been "the compiler that gave us back world peace after
2.96-54 spoilt it". The -69 release is quite good as far as I can see.

Before I forget it: Merry Christmas to you all!

Nils

-- 
         Nils Philippsen / +49.711.96437.250 / nils at redhat.de
       Red Hat GmbH / Hauptstätter Straße 58 / D70178 Stuttgart
The use of COBOL cripples the mind; its teaching should, therefore, be
regarded as a criminal offence.                  -- Edsger W. Dijkstra

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.