This is the mail archive of the
libc-hacker@cygnus.com
mailing list for the glibc project.
Re: ld.so version map
- To: Zack Weinberg <zack@rabi.columbia.edu>
- Subject: Re: ld.so version map
- From: Roland McGrath <roland@frob.com>
- Date: Thu, 11 Feb 1999 19:10:16 -0500
- Cc: Mark Kettenis <kettenis@wins.uva.nl>, libc-hacker@cygnus.com
> The original change was provoked by a bug-report: when you override
> one specific function, namely strcmp(), you get infinite recursion in
> the dynamic linker. Roland said that in this case the linker should
> be using its own routines.
I said this for the routines used in lazy binding symbol resolution (which
I believe to be just strcmp), not ld.so's "own routines" in general.
> I had a patch that copied libc.map and removed that one function,
> would this be preferable in your opinion?
That would be correct, and so would suffice to forestall vehement objection
from me. I still tend to think it might be more maintainable to use a
private version of the function (_dl_strcmp), but that is debatable. In
either case the important thing is to be paranoid about functions in the
symbol resolution code path and make sure any non-_dl_* functions are
handled specially in whichever way.