Re: "random" jamin crashes

On 29 Apr 2008, at 00:54, Dana Simmons wrote:
>    Well to add to the frustration, jamin apears to be camera shy,  
> while
> running trace jamin is reluctant to crash, however, running without  
> the
> trace continues to generate crashes. I am running jackd 0.109.2, so  
> the
> divide by zero error should not apply to me. However the symptoms do  
> fit
> perfectly. Would there be any mathematical difference (divide by zero
> "protection" for instance) in a trace run of jamin? I'm sure I sound
> very ignorant, but all this help is very appreciated.  I did manage to
> get jamin to crash once during a trace, but I believe it has more to  
> do
> with a jackd zombie as ardour also crashed. In any case the last bit  
> of
> the trace log is below.

...
> ioctl(3, FIONREAD, [0])                 = 0
> gettimeofday({1209400924, 127442}, NULL) = 0
> poll([{fd=3, events=POLLIN}], 1, 95)    = ? ERESTART_RESTARTBLOCK  
> (To be
> restarted)
> +++ killed by SIGABRT +++

If it's harder to make it crash under strace that implies that it's a  
timing issue, those can be difficult to debug.

At least we know why Jamin is quitting, it's getting a SIGABRT. I  
wonder if that's what JACK sends if Jamin has missed a deadline? Can  
you try running it with jackd in its soft timeout mode (I don't  
remember the flag offhand), and see if it happens then too?

If Jamin or libjack was getting a divide by zero error (possible) then  
I would expect it to get SIGFPE.

- Steve

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone