|
From: Tom H. <th...@cy...> - 2004-07-05 07:53:04
|
In message <1088990029.6827.10.camel@localhost>
Jeremy Fitzhardinge <je...@go...> wrote:
> Hm. Current glibc (on FC2) defines AT_ entries up to 37, so I presume
> the limit has been removed, or at least increased (even SYSINFO_EHDR is
> 33). On the other hand, these entries won't appear on older systems, so
> there's no issue there. I guess we could just hijack some lower numbers
> and hope for the best (we could use different constants on different
> systems, but I'm not terribly keen on that).
I looked at the FC2 glibc before posting, and _dl_sysdep_start in
sysdeps/generic/dl-sysdep.c still seems to be using an unsigned long
to keep track of which entries it has seen.
Admittedly certain HAVE_xxx defines will stop it doing so, so it isn't
clear whether the actual configuration used on FC2 has a problem but
certainly the last time I disassembled that routine on a RH box it was
doing so.
> Is this causing real problems?
The only time I've seen it cause a problem was on an old RH system
with a 2.2 kernel last time I was trying to get valgrind running
on a 2.2 system. It came up again last week when somebody on the users
list was trying to get valgrind running on a 2.2 system.
Actually, that's another question - are we still interested in trying
to support 2.2 kernels? There are a few minor tweaks needed to make it
compile but it isn't clear that it runs properly afterwards - there is
some suggestion of problems with signal/syscall handling by the looks
of the reports on the users list.
Tom
--
Tom Hughes (th...@cy...)
Software Engineer, Cyberscience Corporation
http://www.cyberscience.com/
|