Re: glibc 2.1.95 (5th test release)

[Geoff Keating <geoffk at cygnus dot com>]

> Date: Sat, 14 Oct 2000 18:45:48 -0400
> From: Daniel Jacobowitz <dmj+@andrew.cmu.edu>
> Content-Disposition: inline
> User-Agent: Mutt/1.1.9i
> 
> On Sat, Oct 14, 2000 at 03:42:20PM -0700, Geoff Keating wrote:
> > No, because previous applications used the old __gmon_start__
> > mechanism which didn't involve weak symbols.
> 
> That's what I thought.  Thanks.
> 
> > > All of a sudden I'm glad we haven't built debian/ppc packages of the
> > > new libc yet...
> > 
> > There are a few more changes to come, too, related to .protected and
> > .hidden.  I hope that very soon ppc glibc will build cleanly; x86 is
> > now clean too, and so we may actually get a release!
> 
> How severe are these?  Worth sticking with 2.1.3 a bit longer for?

I think that unless you have to make a release next week, you probably
want to wait for the patch.  It might not help just to use glibc
2.1.3, unless you also use gcc 2.95.*.  (By comparison, if you use the
new glibc, you should also use new gcc and new binutils.)

I'm now building glibc with a Red Hat custom toolchain which branched
2000-09-12 (it would be related to gcc a few days earlier).  There're
two bugs in later gcc involving reload inheritance and register
preferences which together cause pow() to be miscompiled.

It strikes me that it might be a good idea to set up a branch on
sourceware gcc to mark the date of the C++ ABI that glibc 2.2-based
ppc linux distributions use, if y'all want to share the same ABI.  My
suggested date would be the snapshot before 20000905T000500Z.

-- 
- Geoffrey Keating <geoffk@cygnus.com>