On Wed, Feb 10, 2016 at 7:06 AM, Ronny Meeus <ronny.me...@gmail.com> wrote: > On Tue, Feb 9, 2016 at 9:23 PM, Ronny Meeus <ronny.me...@gmail.com> wrote: > >> Philippe >> >> I have ported the patch to our version of Xenomai and started the test. >> It runs already for several minutes without issues so it looks like >> the problem is solved also with the patch below. >> I will let it run overnight and update you tomorrow morning. >> > > Test is still running, no issues seen. > So the conclusion is clear: your patch solves the issue. > Thanks > > Ronny
Philippe, you state in on of the comments in the patch: " * Whether a condition should be deemed satisfied if it is * signaled during the race window described above is * debatable, but this is a simple and straightforward way to * handle such grey area. " I understand this, it could also be possible to return a timeout in this case to the caller, but it is not acceptable that the message is actually lost. If you implement it in such a way that a timeout error is returned, the message that has been copied into the receiver's buffer should be put back in the queue so that it is received on the next call to q_receive. Ronny _______________________________________________ Xenomai mailing list Xenomai@xenomai.org http://xenomai.org/mailman/listinfo/xenomai