[Openal] A bug in OpenAL, or am I in totally the wrong place?

[James Gregory]

I own copies of 3 Linux games that all suffer from similar (yet not
identical) problems. Two I know for certain use use OpenAL, the third
I'm only assuming it does because it suffers from similar problems to
the first two, and because I know it doesn't use SDL.

The first two games cause identical problems for many users, not just
myself. The third is so little known that I'm one of the only people to
have tried it, and it's disappeared off the internet altogether now.

My computer is as follows:

Mandrake 9.2/10.0 hybrid
OpenAL-0.0.6-13 Mandrake package, taken from OpenAL CVS Feb 10 2004
Cheapo Creative Labs sound card

Game 1:

Gish, a commercial game with a demo available at:
http://www.chroniclogic.com/gish/

The problem:

For many users, including myself, the game segfaults if music is enabled
(though it runs fine with sound). For other users, it runs fine with
both sound and music. Reports of this problem can be found in various
places, including the game's own forums at the above location, and also
in the comments section of the Linux Game Tome at:

http://www.happypenguin.org/show?Gish

What exactly happens (others have described identical symptoms):

If run with sound and music enabled then the Linux console displays:

[hash.c:395] failed read

a great number of times.

At some point it then says:

jmalloc silently ignoring 0 alloc from al_queue.c 442
jmalloc silently ignoring 0 alloc from al_queue.c 443
gish: al_queue.c:445: _alSourceUnqueueBuffers: Assertion `tempqueue'
failed.
Aborted

and aborts.

Game 2:

Trigger, an open source rally game at:
http://www.positro.net/trigger/

The problem:

For many users, including me, but not the game's author, the game
crashes when you hit enter at the start screen. The game uses OpenAL for
sound effects. It doesn't feature any music.

What exactly happens:

As the game is new I can't find anyone else giving full details of
exactly what happens before a crash, but this is what happens for me.

There is the gish-like symptom of:

[hash.c:395] failed read

being displayed loads of times before it segfaults.

I compiled a version with debugging information, and a backtrace when
the game segfaulted gave the following results:

(gdb) bt
#0  0x402395d9 in jhash_get () from /usr/lib/libopenal.so.0
#1  0x403bfee0 in _IO_2_1_stdout_ () from /lib/i686/libc.so.6

I assume the very small length of the backtrace is due to openal running
in its own thread?

Game 3:

RDWars, a very little know open source game which was only in the very
early stages of development and has now disappeared off the internet
entirely. I do know how to contact one of the original authors if really
required though.

The problem:

It never crashes. It uses wavs for sound effects, and has no music.

But it like the above two games it also displays:

[hash.c:395] failed read

a lot of times, and I remember finding out before that though it
happened on my system the game authors had never seen this happen on
their systems.

So, is all of this due to some sort of OpenAL bug?

	James