Re: How to destroy _reent structure

[Corinna Vinschen <vinschen at redhat dot com>]

On Apr 24 09:51, Joel Sherrill wrote:
> On 4/24/2013 7:55 AM, Corinna Vinschen wrote:
> >On Apr 24 13:10, Sebastian Huber wrote:
> >>What is the purpose of _wrapup_reent()?  Its not called from within
> >>Newlib and I don't find a function that can register an atexit
> >>handler in a non-global _reent structure.  There is also no public
> >>header file for this function.
> >Sorry, I can't answer this question.  This code predates my involvement
> >in newlib and I never looked deeper into this.  Cygwin also uses
> >_reclaim_reent to get rid of the per-pthread reent structure.
> We are looking at ancient history code.  The RTEMS code that uses
> our extensions
> on thread create, delete, restart, and switch for newlib dates back
> to the 1992-3
> time frame.
> 
> My vague recollection is that newlib had leaks. Looking at the comments in
> reent.c, my impression is that the code and comments don't match.
> _wrapup_reent() has a comment being run at "task exit" but task is undefined
> in POSIX. For RTEMS task is synonymous with thread but this routine is
> invoking atexit() methods.
> 
> _reclaim_reent() frees a lot and also invokes the atexit() but has comments
> like this that indicate it is at process exit():
> 
>       /* Malloc memory not reclaimed; no good way to return memory
> anyway. */
> 
> Looking at this again, I still don't see a clear indication of what
> should be invoked
> at thread exit and what should be invoked at process exit.
> 
> What does Cygwin do at thread exit and process exit?

Nothing much.  It calls atexit at process exit and _reclaim_reent
at pthread exit.  reeent::_atexit is not used for pthreads.

> Any other insight? Except that the names don't add insight. :)

Not really, no.


Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat