[OpenAFS] Re: 1.6.2 compilation notes, Sol10

Russ Allbery rra@stanford.edu
Thu, 28 Mar 2013 13:52:17 -0700

Andrew Deason <adeason@sinenomine.net> writes:
> Russ Allbery <rra@stanford.edu> wrote:

>> No, regardless of the underlying issue, the current manual library
>> probing for Heimdal libraries in rra-c-util is broken, and I'm happy to
>> fix that in rra-c-util by probing for the additional libraries.

> Well, are you treating "handle modern heimdal" and "pass enough libs to
> aklog to make it work" two different issues?

I believe that, for the purposes of the manual library probes, they will
amount to the same thing.

> For just handling heimdal, the rra-c-util krb5 autoconf logic in OpenAFS
> 1.6.2 seems to work fine for heimdal 1.2.1 (in this thread) and 1.5.3.

On ELF platforms with dynamic libraries, because the rest of the libraries
are pulled in by transitive library dependencies.

It would not (but is supposed to) work with static linkage or on platforms
where shared libraries cannot declare a dependency on other shared
libraries.  And if it did work there, I believe it would also work
properly in this particular case (albeit somewhat by accident).

> If the afsio thing is fixed, the only thing we break on is aklog. And I
> thought you were saying that we break on aklog just because we require
> the weird extra libraries. So, I assumed you didn't want to handle the
> aklog case in rra-c-util, because it's beyond the krb5 api; is that
> incorrect?

I think that OpenAFS should separately probe for the actual libraries it
uses because I think that's more robust and more correct, but I believe
that fixing the manual library probing so that it would work properly for
static linkage is valuable and, as it so happens, would also fix OpenAFS's
immediate problem.

Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>