[Patches] getpath patch (repost)

[Greg Stein]

On Tue, 23 May 2000, Skip Montanaro wrote:
> >>>>> "M-A" == M -A Lemburg <mal@lemburg.com> writes:
> 
>     M-A> "Fred L. Drake" wrote:
>     >>
>     >> On Tue, 23 May 2000, M.-A. Lemburg wrote:
>     >> > The fatal error produces the core dump... so its not all that bad
>     >> > ;-)
>     >>
>     >> Sigh.  Perhaps we need a non-coring Py_FatalErrorXXX()?
> 
>     M-A> Hmm, the core dump is caused by abort() -- I don't think we can
>     M-A> change that behaviour since it is defined by the C lib and using
>     M-A> exit() is not an option since that would cause atexit() hooks to be
>     M-A> called.
> 
> Can't you use the _exit() system call or is that a Unix-only thing?
> According to my Linux man page _exit is part of the defined APIs for SVr4,
> SVID, POSIX, X/OPEN and BSD 4.3.  It also states:
> 
>     _exit does not call any functions registered with the ANSI C atexit
>     function and does not flush standard I/O buffers. To do these things,
>     use exit(3).
> 
> so you might have to flush all open stdio buffers manually.
> 
> Maybe call _exit()+stdio flushes where _exit() is available and abort()
> otherwise...


Um... how about a reset here: WHY do you want to avoid the atexit()
functions?

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/