Re: Fwd: Bug#397489: gtimer: crashes at startup
(Please cc me; I don't subscribe to the list.)
Michael Schmitz <schmitz@mail.biophys.uni-duesseldorf.de> writes:
> Please check that all of your varargs usage conforms to the C standard.
> Using varargs on PowerPC with the wrong data types, or direct
> manipulation of pointers to the va_list usually leads to stack
> corruption rather quick.
This is a very good thought, and now that you mention it I've run into
this before with PowerPC, but gtimer doesn't use stdarg or varargs at all.
> Sorry if that was already obvious to you; I have no idea what has
> already been tried, or for how long the bug persists (if there are no
> changes to varargs related code since the last version known working,
> you can safely ignore this).
I actually don't know the last time someone tried to run it on PowerPC, so
I don't know if it's been working, but the upstream source hasn't changed
in several years. So either it's always been broken, or it's something in
a local patch, but it's hard to know.
All that's been tried so far is valgrind, rebuilding it without
optimization, and checking the build logs on PowerPC for any warnings
(there were none except for a warning about tmpnam, which I'd previously
checked and should be okay in the way in which it's used).
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: