Re: Stack alignement with pthreads

[Anthony Mallet <anthony.mallet%useless-ficus.net@localhost>]

On Friday, at 13:39, Matthias Drochner wrote:
| it is generally up to the compilerto issue some runtime calculations

When running no thread, the code produced by gcc is correct and the same code
breaks as soon as it is executed from a thread. Thus I was wondering if
something somewhere in the netbsd libc/libpthread/kernel wouldn't be pushing an
int/pointer behind the back of gcc. gcc generates codes that makes no runtime
checks on the stack alignment: it just makes sure that main() is correct, so
this fails if some other code breaks the alignment behind the scene.

Note that I don't have this problem with gcc-4.1.2 running under an i386-linux
machine.

| I'd suggest to try with a newer compiler before taking any
| further actions.

Yes, makes sense. I'll do that soon. Recompiling libpthread with a new gcc will
be enough or should I recompile more userland? libc?