[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: GCC vs. /bsd]
On Fri, Jun 21, 2002 at 03:28:25AM +0200, Marc Espie wrote:
> On Fri, Jun 21, 2002 at 12:42:22AM +0000, Miod Vallat wrote:
> > > Did you run every program on your system and validated all the options?
> > > Or did it *appear* to work because it booted?
> > > Did you measure all the performance gains you got?
> > Well, experience has shown that gcc bugs in this case are so subtle that
> > they do not show themselves until you push your kernel very hard (lots
> > of network traffic, long uptime, etc).
> I'm not at all certain all these ARE gcc bugs. Some probably are. A few are
> probably just kernel bugs.
OpenBSD kernel has many ways to debug thanks to all the developers thought
and also allows custom additions and modifications as needed.
> In some cases, timing issues can be involved in seeing some bugs, and
> optimizing the kernel further may rearrange the relative timings of
> some sequences, and exhibit buggy behavior from the kernel.
> So, you will say, if such bugs do exist, why don't they show up with
> gcc and the standard options ?
> Natural evolution: such bugs DO show up with gcc and the standard options,
> but they don't survive long, since they are usually quickly noticed and
> I'm quite willing to dump quite a lot on gcc. But, hey, a bug is a bug.
> And until you've fixed it, you can't be certain whose fault it is.
Did you ever find a compiler bug or a linker bug?