[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [libunwind] "make install" installs

From: Matthieu Delahaye
Subject: Re: [libunwind] "make install" installs
Date: Thu, 07 Oct 2004 10:25:10 -0500

On Thu, 2004-10-07 at 10:01, David Mosberger wrote:
> >>>>> On Thu, 07 Oct 2004 09:49:13 -0500, Matthieu Delahaye <address@hidden> 
> >>>>> said:
>   Matthieu> David, If you are thinking about re-rolling a tarball
>   Matthieu> soon, would it be possible to add a symbol versioning
>   Matthieu> script at link time. I think the major version of the
>   Matthieu> librarie (0 for libunwind-setjmp and 7 for libunwind)
>   Matthieu> should be ok.
> Yes, I'm planning to re-roll the tarball soon.  I'm not sure what you
> want me to do with the symbol version script.  At the moment the major
> versions are already at 0 and 7 for libunwind-setjmp and libunwind,
> respectively.  If you tell me exactly what you want/need, I'll be
> happy to do so.

The need:

Preventing to have 2 different libunwind with same symbols at the same

Imagine we have an application which link against libunwind7 and
libunwind-setjmp0. And libunwind-setjmp0 is an "old" version which has
been compiled with libunwind0 and thus has a depedency against
libunwind0. At the load time we may have libunwind7 and libunwind0 both
loaded. The fact libunwind0 and libunwind7 are exporting the same
symbols may confuse the loader (usually nice segfault). This is purely
theorical and didn't have a test case which cause it.

Proposed solution:
"Attach" to each symbol a version. In other word, instead of having "I
need the symbol unw_step", you have "I need the symbol
address@hidden". This create uniq symbols and prevent clashes.
Sorry for the vulgarisation.

Here is some litteratur found around: (for glibc 2.1
and later)
For Debian

I would really like provide the patch. However my manager is not here
for a while and I need his ok for the sign-off. 

>       --david

Attachment: signature.asc
Description: This is a digitally signed message part

reply via email to

[Prev in Thread] Current Thread [Next in Thread]