On May 5, 2005, at 17:08, Milton Turley wrote:
> Following are 2 patches for kerberos 1.4.1 to build on AIX 5.2. The
> patches are for the problem of not being able to resolve the address
> for the kdc.

Except, apparently, not for Lamar Saxon...

> The patch for dsnglue.c is if "thread-support" is enabled. The patch
> adds a 1024 byte buffer after the _res_state structure. IBM AIX has
> a problem where 735+- bytes are overlaid when res_ninit is called.
> The 1024 bytes pads the storage to stop res_ninit from overlaying
> critical storage. Ken Raeburn had tried a similar patch with 72
> bytes.

Actually, I suggested the patch, but can't try it out, Lamar did that.

It would be nice if someone at IBM could check into this for us and let
us know why the __res_state struct we allocate isn't enough, when the
docs we've found all seem to suggest that it should be.

Um, how does this relate to thread support?

> The patch for locate_kdc.c is when "disable-thread-support" is set for
> configure. Again the #ifndef LANL and #ifdef LANL is a local compiler
> directive. This will need to be changed for local setting or -D LANL
> set for configure process.

>> + #ifndef _AIX
>> + #ifdef AI_NUMERICSERV
>> + hint.ai_flags = AI_NUMERICSERV;
>> + #endif
>> + #endif /* _AIX */

So, are you saying that AI_NUMERICSERV doesn't work for numeric service
strings on AIX 5.2? (And does this relate to thread support?)


Kerberos mailing list Kerberos@mit.edu