Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 18 May 1996 21:43:20 -0600
From:      Nate Williams <nate@sri.MT.net>
To:        Peter Mutsaers <plm@simplex.nl>
Cc:        terry@lambert.org, freebsd-hackers@FreeBSD.org
Subject:   Re: internal compiler error
Message-ID:  <199605190343.VAA02628@rocky.sri.MT.net>
In-Reply-To: <199605180724.JAA15835@plm.simplex.nl>
References:  <199605172004.NAA20510@phaeton.artisoft.com> <199605180724.JAA15835@plm.simplex.nl>

next in thread | previous in thread | raw e-mail | index | archive | help
>   TL> I guarantee that I can make any code you can name run 50% faster.
> 
>   TL> Provided it does not have to produce correct output (a condition you
>   TL> seem to be willing to accept).
...

> I had my own "distribution" that was completely compiled with 2.7.2
> (plus some patches made by H.J. Lu) and -O2. There were no problems
> (especially not the last months after H.J. Lu's latest patches) caused
> by compiler bugs.

You must realize that gcc 2.7.2 as released by H.J. is *not* the stock
2.7.2, but often contains post-release patches to fix bugs in things.
Also, his *only* job in Linux is to do gcc/libc, so he can easily follow
the folks doing GCC maintenance, while the rest of us do gcc 'upgrades'
as an afterthought.

> Who says 2.6.3 is bug free? There is always a small chance for
> compiler bugs. In those rare cases a workaround may be needed. I
> doubt that the latest release of 2.7.2 (2.7.2.9) has more bugs than
> 2.6.3 had.

There *are* some workaround for gcc 2.6.3 in the tree, but they are
*known* bugs, where the bugs in gcc 2.7.2 are unknown, or known well
enough to know they aren't worth the effort to workaround.

Can I count on you to compile *EVERYTHING* you get your hands on when
gcc 2.7.3 is released?  This means compiling the entire world, XFree86,
and all other programs you use on an everyday basis to make sure nothing
breaks?  That relatively simple task (lots of time, but not alot of
expertise required) is the reason 2.7.2 didn't go into the tree.  Things
were broken, and we don't have the time nor desire to go and find out
why when the current compiler works for all the source code distributed.


Nate



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199605190343.VAA02628>