Re: [PATCH] Improve i386 prologue analyzer

[Mark Kettenis <kettenis@chello.nl>]

   Date: Sun, 08 Aug 2004 06:24:04 -0400
   Sender: mec.gnu@mindspring.com

   eliz> If so, could you please explain why you think it needs any more
   eliz> ``marinating'' in HEAD?

   It's a new design.

Yeah, it's a pretty invasive change.  It not only affects the
unwinder, but also prologue skipping in order to determine the first
line of real code for a function.

   And this feature interacts with many different external variations.
   For example, I haven't seen anyone test it with gcc 2.

I see no changes in the testsuite on i386-unknown-openbsd3.5 and
i386-unknown-freebsd4.7 which both use GCC 2.95.3/4 as their system
compiler.  I think these are fine since the code is never used.  I'm
more worried about GCC 3.1/3.2.  Those compilers are in pretty wide use.

   Or with a program that uses a lot of floating point.

That shouldn't be too relevant.  I mean, if they were broken before
the patch, they're probably still broken.  But that doesn't really
matter for the branch, does it?

Personally I'd like to wait for another two weeks, before deciding
it's ok.  I'd be willing to check it in on the branch *now* if we'd
agree we'd release 6.2.1 after that period.  That gives me the
oportunity to pull it out if any problems arise.

Mark