[Bug binutils/4035] BFD failure on solaris 10 with very large shared library

[laurent at oget dot net]

------- Additional Comments From laurent at oget dot net  2007-02-27 15:59 
-------
Subject: Re:  BFD failure on solaris 10 with very large shared library

27 Feb 2007 14:48:14 -0000, nickc at redhat dot com
<address@hidden>:
>
> ------- Additional Comments From nickc at redhat dot com  2007-02-27 14:48 
> -------
> Hi Laurent,
>
>   Does this problem still exist with a version of objdump built from the 
> current
> mainline binutils sources from the CVS repository ?
>
>   Is it possible to reproduce the problem as a small test case ?  (I suspect
> not, but I have to ask).
>
>   If you look at the components of the libBloxArray.so library individually do
> they throw up any invalid string offset errors ?
>
>   If you build libBloxArray.so using a newer version of gcc than 3.4.3 (say 
> 4.2)
> then does the problem still exist ?
>
> Cheers
>   Nick
>
>

Hi Nick,

The problem only happens with that one huge shared library, so, no I
do not have a small test case, saddly enough, and even if you were
interested in rebuilding that huge source tree to reproduce the bug,
we would probably both be dead before we figure out who needs to sign
what NDA...

I got things to work after rebuilding the whole toolchain with the gnu
ld instead of the solaris ld, so it sounds like this might be a
solaris ld issue. The mangled entry points for those heavily templated
classes probably have long names and i seem to remember the solaris ld
had issues with that.

I can not jump to newer gcc yet, because we use jace to build proxy
objects between c++ and java, and jace was not building with gcc 4
last time i tried.....

Do you think I should file a bug with sun about this? Are there any
known issues with sun's ld?

Thanks for your help,

Laurent


-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=4035

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.