On Thu, Sep 22, 2005 at 01:33:49PM +0200, Alfred M. Szmidt wrote:
> Silly me... Why didn't someone hit me in the head?
Well, we tried to, I'd say.
Often it helps if one carefully (!) reads (!) and (!) thinks (!) about
mails before answering to them...
Regards,
Thomas
That's, however, not the end of the story: ext2fs and ext2fs.static
still don't work.
And ufs. Something in init-main.c:40.
___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd
On Thu, Sep 22, 2005 at 06:29:50PM +0200, Alfred M. Szmidt wrote:
> The patch that I sent is quite wrong. The proper place to define
> DL_ARGV_NOT_RELRO is not in sysdeps/mach/hurd/i386/dl-machine.h, but
> in sysdeps/mach/hurd/dl-sysdeps.h (it needs to be included by
> ldosdefs.h).
Yes, that's al
Well, we tried to, I'd say.
Often it helps if one carefully (!) reads (!) and (!) thinks (!) about
mails before answering to them...
Hit harder.
___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd
The patch that I sent is quite wrong. The proper place to define
DL_ARGV_NOT_RELRO is not in sysdeps/mach/hurd/i386/dl-machine.h, but
in sysdeps/mach/hurd/dl-sysdeps.h (it needs to be included by
ldosdefs.h). Now _dl_argv ends up in .bss, and not in .rel.ro; as it
should be.
2005-09-22 Alfred M
Silly me... Why didn't someone hit me in the head? Obviously, we
should define DL_ARGV_NOT_RELRO since we are frobbing _dl_argv.
2005-09-22 Alfred M. Szmidt <[EMAIL PROTECTED]>
* sysdeps/mach/hurd/i386/dl-machine.h (DL_ARGV_NOT_RELRO): Define
macro. Reported by Thomas Schwing
Judging from the bits posted by Thomas, I'm pretty sure this macro
(DL_ARGV_NOT_RELRO) should be defined for hurd-i386, like it is for
sparc, ia64 and alpha.
This isn't a reason why it should be defined. You are confusing code
that is specific for a operating system and a arch, and code
Alfred M. Szmidt wrote:
Hey,
and so has GNU/Linux on x86.
I'm quite confident that GNU/Linux on x86 doesn't use
sysdeps/mach/hurd/i386/init-first.c. Right, and that is where the
problem is.
I don't think so. The attribute_relro attribute for _dl_argv was added
only last year. As Thomas note
On Thu, Sep 22, 2005 at 10:11:28AM +0200, Alfred M. Szmidt wrote:
> I don't think that this is the whole problem.
I also can't say that for sure, yet.
Do you know what else needs to be fixed?
> We've been using -z reloc for ages,
I was writing about `-z relro', which was introduced on 2004/05/11
Do you know what else needs to be fixed?
Not exactly, I'm a bit busy with other stuff, but will look into.
> We've been using -z reloc for ages,
I was writing about `-z relro', which was introduced on 2004/05/11.
Typo on my side.
[...] because of binutils-2.16's test suite failing
I don't think that this is the whole problem. We've been using -z
reloc for ages, and so has GNU/Linux on x86.
___
Bug-hurd mailing list
Bug-hurd@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-hurd
Hi!
I may have found what is causing the recent glibc breakage with respect
to dynamically linking a simple `int main(void) { return 0; }' with
`-lthreads' and only getting EXC_BAD_ACCESS on the following line of code:
`_dl_argv = (void *) (newsp + 1);', sysdeps/mach/hurd/i386/init-first.c.
If th
12 matches
Mail list logo
