Fixed by commits up to & including r944261
There is no information on bad and expected behavior and/or symptoms.
It is impossible to work on this BZ at the moment.
Could you possibly add the description, please?
To test, create a client that has one thread doing a fetch (with an infinite timeout), and has the other attempting to close the session after a short time. The second thread will be locked out by the first thread prior to this fix.
You could also replace the attempted close with an attempt to send on the second thread; again the infinite wait on fetch on the first thread will lock the sender out.
Frantisek, is this still something you want to test?
The issue has been fixed, tested on RHEL 5.7 / 6.1 i86 / x86_64 on packages: