Re: egcs 1.0.1 miscompiles Linux 2.0.33

Jeffrey A Law (law@hurl.cygnus.com)
Thu, 09 Apr 1998 23:39:18 -0600


In message <199803102207.RAA29024@saturn.cs.uml.edu>you write:
> I don't see why the following is not acceptable:
>
> 1. gcc detects a clobbered input
> 2. gcc translates that into an earlyclobbered output with a matching input
Because from a design standpoint it is wrong.

You're asking the compiler to take a statement which is written
incorrectly and fix it. Nowhere else in the compiler do we do
this. Instead you have to fix the incorrect statement.

> If done as soon as the inline assembly is parsed, this wouldn't
> affect very much of the compiler at all. It would certainly not
> affect the way gcc allocates registers, since the constraints
> would be already converted.
Actually it does cause problems. It's a subtle technical issue, but
changing clobbers in that manner is not safe. This came up with the
ppc linux folks.

jeff

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu