ralf@uni-koblenz.de wrote:
> Now that I've taken myself two minutes to browse your attached sources -
> the patches to the NSPR thread routines look suspicious. The are
> playing games with the frame pointer which at least on the first look
> don't make sense as gcc automatically enables -fomit-frame-pointer when
> optimizing.
Turns out it's there only to save the FP so that you can use gdb to look
at the NSPR thread stacks during debugging. So, the fact that gcc
removes it when building optimized isn't a problem.
For the record, NSPR threads aren't clone() threads; they're
setjmp/longjmp things.
Mike
--
517588.35 442847.89
|