[OpenAFS-devel] openafs-1.4.1 build failure on RHEL3 -- aklog fails to build

Derek Atkins warlord@MIT.EDU
Wed, 19 Apr 2006 13:06:09 -0400


Quoting Ken Hornstein <kenh@cmf.nrl.navy.mil>:

>> In trying to build 1.4.1 on RHEL3 it appears the checks for
>> krb5/krb524 libraries are insufficient.   RHEL3 ships mit krb5
>> 1.2.7, which has a separate krb524 library.  I had a fix for
>> the krb5 migration kit that would test whether the krb524 functions
>> are in the standard krb5 libraries and, if not, attempt to add -lkrb524
>> and see if it worked.  Should I create a patch to openafs that
>> performs a similar series of tests?
>
> Sigh.  I really struggled with this.  Part of what made the aklog
> autoconf tests so difficult back in the migration kit days was the
> wacky handling I had to do for the seperate 524 library (especially
> since I was trying to go to the use of krb5-config).  Getting
> everything right ended up being a nightmare.  When we did the aklog
> integration for OpenAFS, we decided that Kerberos versions earlier than
> 1.3 were not going to be supported, for this very reason.  Considering
> how fragile the whole build process is with aklog right now, I am
> not crazy about modifying it further, especially to support four year
> old versions of Kerberos.

I see.  Well, I have two options:

1) Don't support RHEL3 or
2) Patch OpenAFS to check for libkrb524 if it can't find either of these
   symbols.

I suspect if we ask the user base they would prefer #2.  I don't think
it would be TOO hard to write a test that adds "-lkrb524" to LIBS and
checks again for appropriate symbol....

> I will note that the function in the MIT library is krb5_524_convert_creds();
> the symbol name krb524_convert_creds_kdc is for Heimdal.

That's interesting, as the latter symbol ALSO exists in mit's krb524 library,
as noted in my previous email but cut from your response.

Would a patch to aclocal.m4 that checks for libkrb524 in the absense of
either symbols from the krb5-config output be accepted by the openafs
team?  Or am I going to have to keep that patch private in perpetuity as
long as I continue to support RHEL3?

> --Ken

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available