(
Jump to the original submission)
Fri 17 Apr 2009 09:08:12 AM UTC, comment #7:
Samuel would eventually also hit this relro problem, but then also come up with a fix: <http://sourceware.org/ml/libc-alpha/2009-04/msg00003.html>
|
Sun 10 Jun 2007 03:39:03 PM UTC, comment #6:
I've just tried re-compiling hurd with a relro-enabled glibc, and it boots just fine. I tried to compile a package, no problem either.
|
Sun 10 Jun 2007 08:34:17 AM UTC, comment #5:
No, I haven't tried 2.6, precisely because it needs other changes. I preferred to start with a working -z relro -enabled 2.5 before moving to 2.6
|
Sat 09 Jun 2007 08:34:40 AM UTC, comment #4:
glibc 2.6 indeed needs (a lot) more work to make it usable on Hurd systems. The thing glibc 2.6 changed w.r.t. ``-z relro'' was ``only'' that it is no longer possible to switch it off via the command line ``libc_cv_z_relro=no''.
For me, the (only!) test case that illustrates the problem was to use a ``-z relro'' enabled glibc to build the Hurd servers, boot a system with those (e.g. a subhurd) and then one of the core Hurd servers (perhaps the root file system, but I don't remember at the moment) would segfault (also not sure about that, but that's what it was, I think). I once was spending some time working on debugging that, but didn't find the cause of the segfault so far. I think I took some notes about what I did, in case you are interested.
|
Fri 08 Jun 2007 09:06:25 PM UTC, comment #3:
If I'm not wrong 2.5 got released a couple of days before this change, my glibc-2.5 tarball doesn't seem to have the change Thomas referenced in comment #1.
Did you try with glibc-2.6 from experimental instead (though more work is probably needed there to get it built)?
|
Fri 08 Jun 2007 06:59:06 PM UTC, comment #2:
Which testcase makes the dynamic loader crash? I tried compiling 2.5 with z_relro, I am currently using it, and it seems to work fine. People can try a debian package from my repo at http://dept-info.labri.fr/~thibault/debian-hurd main unstable
|
Mon 06 Nov 2006 03:48:50 PM UTC, comment #1:
The hack of configuring glibc with ``libc_cv_z_relro=no'' to get an usable result will not work anymore with glibc trunk versions that are newer than 2006-10-28: <http://sourceware.org/ml/glibc-cvs/2006-q4/msg00202.html>.
|
Fri 08 Sep 2006 08:37:54 AM UTC, original submission:
Binutils 2.16 introduced ``-z relro'', which glibc uses if available and which makes the dynamic loader crash on GNU/Hurd, probably because it is somewhere being used too restrictively.
We currently work around that by specifying ``libc_cv_z_relro=no'' when configuring glibc.
|
(Note: upload size limit is set to 16384 kB, after insertion of the required
escape characters.)
Attach File(s):
Comment:
No files currently attached
Depends on the following items: None found
Items that depend on this one: None found
Follow 3 latest changes.