Re: [PATCH] new makefile for fbcon

Dominik Kubla (dominik.kubla@uni-mainz.de)
Sat, 13 Nov 1999 11:40:50 +0100


On Sat, Nov 13, 1999 at 12:40:26AM +0000, Alan Cox wrote:

> Only one bunch of people have ever moved Linux to a non gcc compiler.
> Irrespective of stuff like makefiles its _very_ hard. I know, Im the nutter
> who started Linux 8086, thankfully Alistair Riddoch took it over and is
> even crazier.

I knew i shouldn't have touched that subject! (SIGH!) I am well aware that
it is (at the momement and possibly always) impossible to change this. I
was merely pointing out that the reasoning "we use gcc, so why shouldn't we
use gmake as well" is not flawless.

> Linux requires inline assembler, and compiler ways of specifying alignment
> rules on a per struct field basis. It makes use of a whole pile of
> preprocessor and other stuff that quite frankly is simply essential and not
> portable across compilers.

I am not wholly convinced that inline assembler is really needed: after all
the first C compilers didn't have it (AFAIK) and they were used to implement
Unix (IIRC). But i guess that is leading us nowhere fast...

> They won't make much difference if any. All the kernel hot spots are hand
> tuned. We in places tune the C to get the code we want out of gcc - things
> like branch order

And we have had problems there in the past: each time the compiler folks
thought it wise to change optimization strategies...

Again i am not trying to push for a change of the used compiler (realizing well
that this would not fly) but simply for keeping the build environment as simple
as necessary (opposed to "as simple as possible") by avoiding excessive use of
non-standard features of the utilities. (Working from the assumption that the
standard features are usually better understood than some obscure specialities
with ANSI-C preprocessor string functions being the exception to prove the
rule :-))

Dominik

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/