On Sat, Dec 02, 2006 at 02:14:09PM +0000, Ben Hutchings wrote: > Ludovic RESLINGER <[EMAIL PROTECTED]> wrote: > > If fmit works when you add --disable-alsa, it is probably > > because it is an alsa bug. > > Of course! The fact that it selects different capturing code in fmit > couldn't have anything to do with it! > > One or more of these obvious errors might be involved: > > 1. CaptureThread uses volatile bool flags for communication between > threads, but there's no guarantee that a volatile bool is or even can be > updated atomically. > 2. CaptureThread::m_packet_size needs to be consistent with > CaptureThread::m_values (I think) but it is written without the lock > held in CaptureThreadImplALSA::capture_loop() and > CaptureThreadImplJACK::jackProcess(). > 3. CaptureThreadImplALSA::capture_finished() uses delete to free an > array; it must use delete[] instead. > 4. CaptureThreadImplALSA::set_params() will continue if it can't set the > format as requested (however the logs posted to this bug report indicate > that this hasn't happened). The format could use 32-bit samples, which > would overflow the buffer in captureLoop(). > > Someone who cares about this program could try to fix them. > > Ben. >
Hi,
Thanks you for your help.
Yesterday, I contacted fmit upstream author about that.
He already corrected some of these errors.
When I will have his patch, I will apply it on debian package.
I don't know if it will solve segfault problems, but upstream author
had solved similar problems with theses changes.
If this patch doesn't solve the problem, I will probably reassign the bug
to alsa.
Regards,
--
.---. Ludovic RESLINGER
/ \
[EMAIL PROTECTED]@./ Trumpet Student in CNR
/`\_/`\ Free Software Developer
// )X( \\
| \ : )|_ _,'| .''`.
/`\_`> <_/ \ @======TTT=====::_ | : :' :
\__/'---'\__/ ((_<=HHH___)) `.| `. `'`
`---UUU---'=> `-
signature.asc
Description: Digital signature
