Hamish Moffatt <[EMAIL PROTECTED]> writes:
> On Mon, Feb 11, 2002 at 07:01:09PM +0100, Kjetil Torgrim Homme wrote:
> > Aha, I didn't realize there was that kind of black magic in ld.so
> > (documented in ldconfig). Well, then I'd venture that ld.so is
> > imperfect. If it knows to ignore certain paths in ld.so.conf, it
>
> When does it ignore paths in ld.so.conf?
ldconfig tags the directories in the cache, so that ld.so knows which
paths should be consider for a given program.
>From ldconfig(1):
ldconfig will attempt to deduce the type of ELF libs (ie. libc
5.x or libc 6.x (glibc)) based on what C libraries if any the
library was linked against, [...]
Some existing libs do not contain enough information to allow
the deduction of their type, therefore the /etc/ld.so.conf file
format allows the specification of an expected type. This is
only used for those ELF libs which we can not work out. The
format is like this "dirname=TYPE", where type can be libc4,
libc5 or libc6.
> If your binary contains a hard path to /lib/libc.so.6,
What do you mean by a hard path? dlopen()? Sure, it won't work.
Usually, /lib/ld-linux.so.2 is hard coded, libc is not.
> and then we move that library to /usr/i386-glibc2-linux/lib instead
> because libc.so.7 is available, your binary will no longer work.
If I understand the above excerpt correctly, ldconfig would need to be
extended to know about "libc7", but then it'd work.
Kjetil T.
