On Sun, Aug 07, 2005 at 11:39:17AM -0500, Gerald (Jerry) Carter wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > NIGGEMYER Brant wrote: > > > > FYI updated kerberos libs from krb5-libs-1.4-3 to krb5-libs-1.4-5; > > problem still existed. > > > .... > > ==26396== Invalid free() / delete / delete[] > > ==26396== at 0x1B909743: free (vg_replace_malloc.c:152) > > ==26396== by 0x3DAABA: remove_error_table (in /lib/libcom_err.so.2.1) > > ==26396== by 0x4D98C3: (within /usr/lib/libkrb5.so.3.2) > > ==26396== by 0x4D95C6: (within /usr/lib/libkrb5.so.3.2) > > ==26396== by 0x52A9D9: (within /usr/lib/libkrb5.so.3.2) > > ==26396== by 0x1B8F2057: _dl_fini (in /lib/ld-2.3.5.so) > > ==26396== by 0x238C68: exit (in /lib/libc-2.3.5.so) > > ==26396== by 0x222DED: __libc_start_main (in /lib/libc-2.3.5.so) > > ==26396== by 0x2FD44: (within /usr/local/samba/bin/net) > > ==26396== Address 0x53BDB0 is not stack'd, malloc'd or (recently) > > free'd > > Jay, Can you give me some feedback on the krb issues > we are seeing in FC4? I've got 3 or 4 reports that all > appear to be bugs in the krb-1.4.1 code. > > I don't see any recent mention of problems on the MIT > mailing list either.
According to Nalin, who is our krb5 expert, it isn't actually a bug in our krb5, but in e2fsprogs, which is where the error_table handling code ended up. Apparently it isn't correctly distinguishing between a dynamically allocated error table (which it can free), and a statically allocated one (boom!). You can work around the problem by removing all the calls to initialize_krb5_error_table() (in source/libads/kerberos{_verify,_keytab,}.c), but I think that makes debugging krb5-related error significantly more difficult. -- JF -- To unsubscribe from this list go to the following URL and read the instructions: https://lists.samba.org/mailman/listinfo/samba