> On Mon, 14 Sep 1998, Florian La Roche wrote:
>> I am not so pleased with Alan blocking all egcs-related patches from kernel
>> 2.0.x. I am using development snatshopts and now egcs 1.1 on many machines
>> and haven't had any failure that could be blamed on egcs (having the known
>> kernel-bugs fixed of course).
> With all due respect, egcs is an experimental (read: unstable, development)
> compiler set, the 2.0 Linux kernel is a stable release. And while it may
> compile some user-space applications fine, there are certain areas of the
> 2.0 kernel that it generates bad code for, and you _will_ experience
> failures that can easily be blamed on this false egcs-born code.
Fixing the bugs in linux 2.0.x, I run egcs-compiled kernels on
several machines for months without any problems. I know that
many other people do the same. (I used to have development egcs-compilers
and now the egcs-1.1-release.)
I like egcs 1.1 much better than gcc 2.7.x and I want to use it as my
main C-compiler. (I do have some other gcc/egcs-versions around and
I could compile the kernel with another compiler, but I don't *want* to.)
> This is why kernel developers such as Alan Cox have stated they will ignore
> all bug reports of 2.0 kernels compiled with egcs. The issues are known,
> the bugs are because of egcs, and the upcoming 2.2 release kernels have
> fixed or worked around the problems. IMHO the kernel developers right now
> should be working towards the 2.2.0 goal, and not worrying about
> backporting fixes to a soon-to-be-obselete stable kernel series.
I hope that Alan is updating the linux 2.0.x kernel for another year, so
that people have an option to choose between a 2.0 and a 2.2 kernel.
Having this option between a linux-2.0 and a linux-2.[12] kernel is
doable on a libc5 based system and that's what I want to have.
(I am waiting for glibc 2.2 for a glibc-change.)
You want to force people to use a very untested kernel 2.2, but you
don't want to see people using egcs 1.1-release. Maybe you should
found another church for that...
> It's not too much difficulty to have gcc-2.7.2.3 and egcs-1.1 cohabitating,
> and using egcs if you so wish for your user-space programs and gcc for your
> kernel-space programs.
And it is not too difficult to fix the bugs in the 2.0.x kernel. Then people
have the *option* to use the compiler they like. Why should testing the
linux-2.0.x kernel with a very current egcs not be done?
Linux is about enabling people to do things and not disabling.
Sorry for being a bit harsh, but I have seen so many postings that
repeat Alans words without re-thinking about them...
(Though I can agree with Alan, that he himself doesn't want to bother
if someone uses an egcs-compiler. It is ok to refuse bug-reports, if it
might be due to an untested compiler. But as you already said, those people
can so easily have two compilers that this is no problem for them
and for Alan... :-)
Florian La Roche
-
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/