Re: A patch for glibc 2.0/2.1

[hjl@lucon.org (H.J. Lu)]

> But we allow future versions which we don't know for sure will work,
> or which might require more special hacks like the C++ exceptions
> things.
> 

Many glibc developers are also egcs developers. We are confident that
all future egcs will work with glibc. At least, we can make them work
together.

> I meant, perhaps we should say 
> "glibc 2.1 should only be compiled with egcs 1.1.1".
> 
> > > I don't see why being "flooded with glibc bug reports associated with
> > > gcc" is a problem.  Simply forward them to the gcc maintainers, if
> > > they are a gcc problem, or fix them if not.
> > 
> > It is still a glibc problem, although it is caused by gcc. Forward
> > to the gcc maintainers doesn't solve the glibc problem.
> 
> Then shouldn't we fix the glibc problem?
> 

We have tried and some of us conclude the most appropriate solution is
disallow gcc 2.8.1. Unless someone comes up a patch and test it against
both gcc 2.8.1/egcs to build both Debian and RedHat to see if their
binaries can run on each other, I don't think we should allow gcc
2.8.1.

Please keep in mind that Linux is going mainstream. If a generic binary
compiled on one Linux system won't run on another Linuxn system based
on the same glibc, it does't help Linux at all. Disallow gcc 2.8.1 is
a small price to pay. With egcs 1.1.1/Linux, I don't see why someone
wants to insist to compile glibc with gcc 2.8.1.


H.J.