OK. That's useful information. There are some known thread-related problems in Ubuntu 8.10, but it looks like our OpenBTS problems are more general than that.
I know of someone running under Debian 4.0 who is not reporting these problems. There's also someone on this list running under Gentoo, but I don't know if he's having this bug.
The bottom line is that the stuff in CommonLibs/Threads needs to be patched to work under most Linux variants. I particularly suspect the pthread_cond_timedwait does not behave as documented.
At this time, the only USB 2.0 development machines we have handy at Kestrel are Macs, and they don't show these errors. Kestrel will need to fix this as part of its move to embedded systems, but it's not on our schedule for a few more weeks, and even when we do fix it, we can't put the fix into GNU Radio because of the injunction. So I would encourage anyone on the list with pthread experience under Linux to take a crack at it.
It would also be useful if the specific treading failures can be reproduced in a stand-alone fixture, without having to run the whole OpenBTS application or even connect a USRP. This could probably be done by adding some new operations to InterthreadTest.cpp.
On Jan 15, 2009, at 4:44 AM, Alexsander Loula wrote:
No way, same errors as Ubuntu!
Do you know someone running OpenBTS under GNU/Linux that is not getting this issue?
2009/1/14 Alexsander Loula <email@example.com>
I'll give a try with OpenSUSE 11 64 bits only to see if the behaviour will be the same. I let you know!
David A. Burgess
OpenBTS on the web: