[iiwusynth-devel] Re: [Swami-devel] iiwusynth segfault

[Josh Green]

On Wed, 2003-01-22 at 08:52, Mark Knecht wrote:
> Josh,
>    This is not a Swami issue, specifically, but before I join yet
> another mailing list, have you seen or heard of this?
> 

Both of the FluidSynth developers are on this list as well (I'm cc'ing
the iiwusynth devel list), so they probably will get your message as is.

>    On my main DAW I've downloaded from CVS twice, built iiwusynth using
> 
> ./autogen.sh
> ./configure --disable-oss-support --enable-jack-support
> make 
> make install
> 
> 
>    Just trying to run it at the command line with the jack option it
> segfaults. It's also not happy about memory, apparently. I'm running as
> a user.
> 
> 
> address@hidden mark]$ iiwusynth Soundfonts/Accord1.SF2 
> iiwusynth: warning: Failed to pin the sample data to RAM; swapping is
> possible. 
> 
> Nothing happens. Ctl-C to break
> 

Hmmm. Strange. That was probably what was causing Swami to not start
before. The "Failed to pin the sample data to RAM" probably has to do
with the mprotect call that is used to ensure that the sample data
doesn't get swapped (causing overruns). It isn't really critical. If you
were running as root the mprotect call would probably succeed. As things
are, I wouldn't run it as root till you know it is working (SCHED_FIFO
could lock machine if 100% CPU is reached).

> 
> 
> address@hidden mark]$ iiwusynth -a jack Soundfonts/Accord1.SF2 
> iiwusynth: warning: Failed to pin the sample data to RAM; swapping is
> possible. 
> Segmentation fault 
> address@hidden mark]$ iiwusynth -a jack Soundfonts/elepiano.sf2 
> iiwusynth: warning: Failed to pin the sample data to RAM; swapping is
> possible. 
> Segmentation fault 
> address@hidden mark]$ 
> 

Did you also just try running it without loading a SoundFont?

iiwusynth -a jack

or

iiwusynth -a alsa

> 
> 
> If I stop Jack, then it knows that:
> 
> address@hidden mark]$ iiwusynth -a jack Soundfonts/elepiano.sf2
> iiwusynth: warning: Failed to pin the sample data to RAM; swapping is
> possible.
> cannot connect to jack server
> cannot connect to default JACK server
> iiwusynth: error: Jack server not running?
> Failed to create the audio driver
> address@hidden mark]$
> 
> Doesn't work with alsa either.
> 
> address@hidden mark]$ iiwusynth -a alsa Soundfonts/elepiano.sf2
> iiwusynth: warning: Failed to pin the sample data to RAM; swapping is
> possible.
> iiwusynth: warning: Attempted to use 16 audio periods got 0
> period size: 64
> iiwusynth: interval.c:159: snd_interval_refine: Assertion
> `!snd_interval_empty(i)' failed.
> Aborted
> address@hidden mark]$
> 

Hmm, sounds like something to do with the ALSA buffer size setting code.
Perhaps try running FluidSynth with different buffer sizes and counts:

iiwusynth -z 384 -c 3

Some buffer sizes to try 128, 384, 512, 1024. Buffer count should
probably be 2 or 3.

> 
> This is a machine that's run Alsa and JAck for months now. Lots of other
> jack apps work.
> 
> Let me know what you think. I can join iiwusynth-devel and submit this
> there.
> 
> Thanks,
> Mark
> 

What I would suggest is that you compile FluidSynth with debugging
enabled "./configure --enable-debug" and then get some back traces on
the lockups and segfaults by running FluidSynth through gdb:

> gdb iiwusynth

(gdb) set args -a jack
(gdb) r

wait for segfault

(gdb) bt

This will give you a stack trace of where the crash occured. Post it to
iiwusynth-devel (although you are welcome to post it here). In the case
of running iiwusynth without specifying the driver (where you had to hit
ctrl-c), you can do the same thing. i.e., hit control-c and the program
will probably stop in the location that is locked, do a backtrace (bt).

That will help us figure out whats going on. Cheers.
        Josh Green