[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-tar] Core dumps in tar-1.26 on Solaris SPARC
From: |
Paul Eggert |
Subject: |
Re: [Bug-tar] Core dumps in tar-1.26 on Solaris SPARC |
Date: |
Wed, 13 Apr 2011 12:09:44 -0700 |
User-agent: |
Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.15) Gecko/20110307 Fedora/3.1.9-0.39.b3pre.fc14 Thunderbird/3.1.9 |
On 04/13/2011 01:40 AM, Joerg Schilling wrote:
> Why do you believe this is a ELF problem?
>
> From the bug desciption, it seems that GNU tar just made an illegal
> assumption
> or can you verify that the link order was violated in this case?
As I understand it, it's too easy, with ELF, for a program
to define a symbol FOO, that inadvertantly usurps the same
symbol FOO that (unbeknownst to the program) is used in a
library. Sometimes the ELF behavior is wanted, but often
(as is the care here) it is not. On some non-ELF systems
a program can't redefine a symbol exported by a library,
and on such systems this problem wouldn't have occurred.
In this particular case, FOO was introduced in a later
version of Solaris, so it is a real upwards-compatibility
problem: a program builds and runs fine on Solaris X,
but the same executable crashes on Solaris X+1. This is
an inherent property to the way ELF link order is implemented.
It's not a bug with the linker per se: the bug is in the
overall ELF design.