Re: gcc 3.2.2

[Reuben Thomas <rrt@...>]

> I agree. I think that this problem needs more careful investigation. It could
> point to a compiler bug in gcc 3.2,

This is possible.

> a rare heap corruption bug somewhere in
> lua,

This is unlikely. Running the same test under valgrind doesn't show the
problem. My understanding of valgrind is that since it's completely
transparent as far as user address space goes, it should have exactly the
same (user-space) memory allocation patterns as running natively, so
should show the same bugs.

> or even a bug in the dynamic linker used on those systems.

Again, possible. In particular, my investigations with glibc (which was
where the crash in my case actually ended up happening) turned up nothing,
so it was certainly a bizarre bug. I tried several memory error detectors
and all of them gave Lua a clean bill of health too.

> Correct me if I am wrong, but using PIC on the x86 architecture only affects
> the relocations that the dynamic linker needs to do. In our experience, many
> shared libraries actually run faster on Linux/x86 if they are compiled
> without the PIC flag.

That's what I thought: startup should speed up, execution may slow down.

-- 
http://www.mupsych.org/~rrt/ | impunity, n.  wealth (Bierce)