* * * * *
                                        
                      It's never lupus nor a compiler bug
                                        
While testing “Project: Sippy-Cup [1],” one of the components I was using for
testing asserted [2]:

> tpp.c:63: __pthread_tpp_change_priority: Assertion `new_prio == -1 || (new_prio >= __sched_fifo_min_prio && new_prio <= __sched_fifo_max_prio)' failed.
> 

This is not good. This wasn't our code that called assert(), but pthreads
[3].

Sigh.

The hardest part of debugging this was getting it to happen consistently
(yes, it's one of those bugs [4]) to test out various hypotheses.

It wasn't neglecting to initialize a variable [5]. It wasn't a compiler bug
(It's never lupus!) [6]. It turned out to be a abuse of the mutex API
(Application Program Interface) [7], where one thread locks a mutex, and some
other thread attempts to unlock the mutex.

[1] gopher://gopher.conman.org/0Phlog:2014/03/05.1
[2] http://pubs.opengroup.org/onlinepubs/009695399/functions/assert.html
[3] https://en.wikipedia.org/wiki/POSIX_Threads
[4] https://en.wikipedia.org/wiki/Race_condition
[5] http://stackoverflow.com/questions/21825291/threading-issues
[6] https://gcc.gnu.org/ml/gcc-bugs/2006-10/msg00805.html
[7] http://stackoverflow.com/questions/1105745/pthread-mutex-assertion-error

Email Sean Conner at sean@conman.org

.