This is the mail archive of the
libc-alpha@sources.redhat.com
mailing list for the glibc project.
Re: PATCH: Make _errno/errno available in GLIBC_2_0.
On Tue, Sep 16, 2003 at 02:15:12PM -0700, H. J. Lu wrote:
> On Tue, Sep 16, 2003 at 01:49:15PM -0700, Roland McGrath wrote:
> > Your patch gives errno@GLIBC_2.0 users an `errno' that is always zero and
> > never set by libc functions. That is worse than refusing to load, IMHO.
> > I am not convinced that such ancient binaries really need to be able to run
> > with NPTL, so as to warrant the effort of trying to do this right.
>
> I am not against that. But it should be consistent between linuxthreads
> and nptl.
Why?
linuxthreads --without-__thread can very well serve as a workaround
for the buggy binaries/libraries and the only way to run glibc 2.0
binaries.
linuxthreads --with-tls --with-__thread doesn't export TLS either.
BTW: Your patch actually did worse thing than exporting always zero
errno never set by libc. You exported errno as STT_TLS while programs
wanting errno@GLIBC_2.0 were looking for STT_OBJECT. Which is a big no no.
> FWIW, the "ancient" binary is portmap from RHAS 2.1.
It is about ancient and buggy binaries/libraries.
Portmap in RHAS 2.1 was the latter category.
Jakub