On Tue, 2002-06-18 at 12:34, Geoff Keating wrote:
Geoff or Ulrich have the final ok here. My personal opionion would be
the "more comprehensive" solution. Geoff, what do you think?
If I have a choice, I'd really like the more comprehensive solution!
I have not the slightest interest in adding more hacks. The code works
fine for the platforms so far supported. If some new processor
implementation comes along which needs the code changed this has to
happen without effecting the existing code. If the final solution
requires using the auxiliary vector values and therefore a memory load,
so be it. But no going back to the generic code and no disabling the
benefits for supported implementations.
In this context, i was wondering what do you do with processor bugs like
for example errata #77 on the IBM PPC405GP which have impacts on glibc?
What's the "official" glibc handling procedure in such a case? A
separate patch to fix it resulting in a specific glibc? Or integrating a
work around into glibc if performance doesn't suffer (too much) for
other targets?