This is the mail archive of the libc-alpha@sourceware.cygnus.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: libc/1508: symbol conflict due to nss_db.so usage of /lib/libdb.so.3


   Date: Wed, 05 Jan 2000 01:54:35 +0100
   From: Michael Marxmeier <mike@msede.com>

   Yep. I re-built glibc-2.1.2 with your proposed modifications. Works.
   So it's most likely related to __db_jump.

   While db_jump could be fixed (eg. changing the name) don't we need
   to worry about the other internal symbols which are visible from
   the application?

   #    __bam_init_print; __bam_pgin; __bam_pgout;
   #    __db_dispatch;__db_dump; __db_err; __db_init_print;__db_jump;
   #    __db_omode;__db_prdbt;
   #    __ham_init_print; __ham_pgin; __ham_pgout;
   #    __lock_dump_region;
   #    __log_init_print;
   #    __memp_dump_region;
   #    __txn_init_print;

   The same issue may affect them as well?

Sure and even worse: some of the public interfaces like memp_register
might cause problems too.  Chances are that the implementations in the
two DB libs are compatible but there is no guarantee.

I don't think renaming interfaces and adding weak aliases is something
we want to spend time on since we're planning to drop Berkeley DB from
glibc completely.

The bad news is also that the new approach that we are implementing
for glibc 2.2 isn't going to work either.

What's really needed is an implementation of the RTLD_GROUP flag for
dlopen.  This is probably quite a lot of work and won't happen
overnight :-(.

Mark

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]