> >No, you have to convince Alan Cox, that it should. Seems he doesn't like
> >to fix anything for gcc-280.
> If you can get people to run kernels with gcc 2.8.0 and this patch for a few
> weeks and convince everyone that nothing breaks, it can probably go into
> 2.0.34. Putting it in otherwise is implicitly saying "gcc 2.8.0 is ok", and
> that's a dangerous thing to do.
You don't want to fix a bug in the kernel because somebody using a
compiler which does (correctly) fail on it has the chance to fall into
other problems because of (other) compiler bugs? Strange idea!
I'd rather fix everything I could and risk to have 2.0.35 and even .36 if
necessary to cope with compiler problems.
Note, that there is not only gcc-280 but also the egcs compiler (in which
bugs are fixed were quickly) but which also (correctly) fails on the
Kurt Garloff, Dortmund