On Mon, Feb 14, 2005 at 04:06:51PM +0000, Maciej W. Rozycki wrote:
> > Bulletproofing 2.4 against newer tools is something that only makes limited
> > sense, especially wrt. to gcc 3.4 and newer. Chances for any such changes
> > to be accepted upstream are slim - and the kernel has traditionally been
> > easily affected by overoptimization, so I recommend against gcc 3.4. The
> > recommended compiler for 2.4 is still gcc 2.95.3 but except gcc 3.0 upto
> > gcc 3.3 is reasonable and can be considered well tested.
> I do agree in general, but given that the construct I've used has been
> supported by gas since 1995, there is no point in keeping our code broken.
> And binutils actually quite rarely trigger problems with Linux, while
> they've got improved significantly with the last few releases; unlike with
> GCC it's normally a good idea to use the latest version of binutils.
I wasn't objecting to your patch; it's just that I expect some users to
upgrade to a recent binutils and gcc at the same time and that has good
chances to end up in a nice firework ;-)