Re: egcs, glibc 2.0, and __register_frame_info

[Daniel Quinlan <quinlan@transmeta.com>]

Paul Kimoto <kimoto@lightlink.com> writes:

> I am curious about whether LSB will (or should) take up the question of the
> __register_frame_info symbol (and friends) that egcs puts into some of the
> glibc-2.0 shared libraries.  
>
> (Briefly, egcs-compiled executables linked against these libraries expect
> this symbol to be defined in the runtime environment, so these executables
> can't be run on systems with gcc-compiled glibc-2.0.  There is a nice
> description on the answer to FAQ#2.8 in the glibc-2.1.0 distribution.)

I haven't seen the FAQ, but I think we will avoid the question by
basing the LSB specification on glibc 2.1.  glibc 2.1 doesn't have a
public interface for that symbol, right?

All of the distributions I have talked to on the libc version issue
(Caldera, Debian (in the persons of Ian Jackson and Dale Scheetz, but
not the debian-policy group), Red Hat, and SuSE) are in favor of
basing the specification on glibc 2.1.  The glibc people are also
pushing for us to use glibc 2.1.

That wouldn't mean that distributions won't have to ship glibc 2.0 for
compatibility with some currently shipping applications, of course.

- Dan