This is the mail archive of the
libc-hacker@sourceware.cygnus.com
mailing list for the glibc project.
Re: glibc-2.0.99 & linux 2.1.125/2.1.126
- To: zack@rabi.columbia.edu
- Subject: Re: glibc-2.0.99 & linux 2.1.125/2.1.126
- From: John Kennedy <jk@csuchico.edu>
- Date: Sun, 25 Oct 1998 16:13:12 -0800 (PST)
- Cc: glibc <libc-alpha@cygnus.com>
10/25/98 @ 04:12:58 PM (Sunday)
[Zack Weinberg]
> Those tests are the most interesting I think.
> The change in binfmt_elf.c is to unconditionally honor the p_vaddr
> field of the elf_phdr struct. If there's a binutils bug such that
> this field sometimes contains garbage, that'd explain the problem.
> I don't see any immediately relevant changes in the kernel headers.
Given the other comments, it seems pretty clear that it has everything
to do with the running kernel and not the kernel includes. I happen
to be using binutils-2.9.1.0.15 (+egcs-1.1b) and I saw one other person
mention that they were using binutils-2.9 (and presumably something).
kern-run kern-hdr glibc stat
2.1.125 2.1.125 2.0.98 OK
2.1.125 2.1.125 2.0.99 pending #2 (ok?)
2.1.125 2.1.126 2.0.98 OK
2.1.125 2.1.126 2.0.99 pending next (ok?)
2.1.126 2.1.125 2.0.98 FAIL
2.1.126 2.1.125 2.0.99 FAIL
2.1.126 2.1.126 2.0.99 FAIL
2.1.126 2.1.126 2.0.98 FAIL
Anything marked as "ok?" sounded like someone on the list had already
confirmed it to work.
This next "pending" ought to get me a bootstrap glibc-2.0.99 that I
can throw and my (homebrew) distribution-builder. In earlier testing,
some applications certainly don't have a problem. Before I (re)build
glibc, I (re)compile these apps:
bzip2-0.9.0b perl-5.005.01 bison-1.25
flex-2.5.4a gettext-0.10.35
The "seed" applications (compilers, glibc, etc) that made up the
original binaries were glibc-2.0.96, egcs-1.1b and I'm not positive
about which version of binutils I was using (based on dates, I'd say
binutils-2.9.1.0.13). When they were dropped into the chrooted build
environment, they were using glibc-2.0.98 in my tests.
Tom Rini pointed out that ftp.yggdrasil.org also has the patch, so
maybe I'll grab the patch there and reverse the binfmt_elf.c version.
This list is a little slow (human moderated?), so it is possible that
there is a success story in the wings.
--- john