El Miércoles, 20 de Julio de 2005 17:33, Eugene Konev escribió: > Hello David. [...] > It seems the libvgahw.a problem can be just a part of the problem, > though well noticeable one. Many drivers use macros like this: > > #define INREG(addr) *(volatile CARD32 *)(MMIOBase + (addr)) > > And using such macros just to read the mmapped registers without using > returned values later can lead to the same "optimized out" code.
I know, I know. The original poster of the patch said that libvgahw.a
was
probably only 1 of 100 cases where volatile was used along the whole X.Org
sources. So yes, your patch is only a band-aid for a bigger problem.
Let's hope that the GCC boys commit soon a fix to the compiler
(supposing
that they finally got consensus about this) and GCC maintainers add it to the
current GCC suite in Debian.
Best regards,
Ender.
--
- Kyle, all those times I said you were a dumb, stupid Jew,
well, I was wrong, you're not a Jew.
- Cartman, I *am* Jewish!
- There, there, don't be hard on yourself, Kyle.
-- Cartman & Kyle (South Park).
--
Debian developer
pgpzuZfIFWjZe.pgp
Description: PGP signature
