On Thu, 7 Dec 2000, Kenneth E. Lussier wrote:
> Rich Payne wrote:
> > At the risk of starting a flaim war....that's not fair. We can argue about
> > it's releasability (not sure that's really a real word!), but the fact
> > remains that the numbers of bugs fixed in 2.96 is massive, and the
> > kernel not compiling properly is something the kernel guys need to (and
> > have for 2.2.18pre) fix. You also need to remember that Linux!=i386, and
> > the other archs (Sparc and especially Alpha) benefit greatly from this
> > change to the compiler.
>
> I have to wholeheartedly disagree with this. RH took the development
> code (from the CVS tree, mind you) and released it as the primary
> compiler on RH7. The gcc team has stated time and time again that the
> code that was shipped by RH is unstable, and completely incompatible.
> Anything compiled with 2.96 is incompatible with the egcs 1.1.2 libs,
> and it will be incompatible with the upcomming 3.0 as well because there
> are already major changes in the code base that they planned on making.
> No good can come of it.
The decision to ship 2.96 was not taken lightly. RH looked at both fixing
all the bugs in 1.1.2 that needed to be fixed, or patching up the current
CVS build. The later was actually less work, and it was not just pulled
from CVS and put into the dist but rather went through quite a bit of
testing. The major complaint that most people have is about the kernel
compilng (in fact I have yet to hear of a package that fails with RH's
2.96 except the kernel)...and that is a fairly simple fix that the kernels
guys would need to do anyway (and are doing).
The advantages for other arches is huge (it maybe for x86 as well,
but i don't play with x86 machines that much). We now have some decent
optimizations for Alpha*, KDE (actually QT) now gets compiled correctly,
etc...etc..... Now, for the sake of argument lets say RH held back 7.0 for
gcc 3.0, then we would all be critizing them for running with the first
release and not waiting for the first set of bug fixes. It's a no win
situation, there is never a good time to update major components like
this, so they did want they always do...roll the big changes into the .0
release and then clean up the mess with .1 and .2.
* In fact, under cetain circumstances 2.96 compiled code has outperformed
code compiled with Compaq's own Alpha compiler!
--rdp
--
Rich Payne
[EMAIL PROTECTED] www.alphalinux.org
**********************************************************
To unsubscribe from this list, send mail to
[EMAIL PROTECTED] with the following text in the
*body* (*not* the subject line) of the letter:
unsubscribe gnhlug
**********************************************************
