> The glibc included with FC5 has a change that breaks the lwp code in 
> openafs.  The setjmp() function now mangles the stack pointer, and the 
> lwp code has to be able to do the same mangling to be able to modify the 
> pointer.

That may ring a bell. Lately, all my arlad crashed unless I compile
with --with-pthreads which in Arlas case makes lwp a wrapper around
pthreads. Pthreads on Linux on the other hand seems to be more stable
nowadays than the house of cards on a folding table in a runaway train
down sherman hill it used to be. So you might find the wrap-pthreads
stuff useful.

Harald.
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel

Reply via email to