Re: Creating Zombies with epicsThreadExitMain

[Andrew Johnson <anj@aps.anl.gov>]

Michael Abbott wrote:
> Ah, I misunderstood that -- however, it's a red herring in the current 
> case, as something is going horribly wrong on my machine when we do this.  

Which thread library is your machine using — is it the old LinuxThreads 
or NPTL?  According to man pthreads you can use this shell command to 
find out:
    getconf GNU_LIBPTHREAD_VERSION

> I need to dig a bit into the pthread specification, but I'm sure I 
> remember that _exit() is supposed to kill *everything* stone cold dead 
> without ceremony.

You are probably right.  However since we don't actually call _exit() in 
this circumstance it doesn't really matter — th stuff about _exit() 
seems to have been a Red Herring I introduced by not looking at our 
implementation of epicsThreadExitMain() — sorry about that.

We do have code in libCom/osi/os/posix that calls pthread_cancel() on 
all our threads, thus any call to epicsExit() will result in all our 
threads voluntarily shutting down.  We had to introduce the epicsExit() 
and epicsAtExit() API because Windows runs atexit() routines after 
unloading all DLLs, even when the routines are found in said DLLs...

> What exactly does epicsThreadExitMain() do to try and close the top level 
> thread without terminating the process?  When I've had to do this in my 
> own processes I've found it easiest to just created a semaphore and block 
> on it.

It calls pthread_exit(0).  This appears to be legal since the Posix 
description of pthread_exit() includes the following wording:
    The process shall exit with an exit status of 0 after the last
    thread has been terminated. The behavior shall be as if the
    implementation called exit() with a zero argument at thread
    termination time.
http://www.opengroup.org/onlinepubs/009695399/functions/pthread_exit.html

However we don't actually rely on the above behaviour ourselves.  Once 
epicsThreadExitMain() has been used the proper way to actually close an 
IOC down is for one of the remaining threads to call epicsExit(), which 
shuts down the other threads and calls exit().

HTH,

- Andrew
--
Talk is cheap. Show me the code. -- Linus Torvalds