Failing ASN.1 tests with PKINIT on HP-UX

[Osipov, Michael (IN IT IN)]

On 2025-09-27 16:40, Ken Hornstein wrote:
>>> 771 #ifndef DISABLE_PKINIT
>>> 772     /****************************************************************/
>>> 773     /* encode_krb5_pa_pk_as_req */
>>> 774     {
>>> 775         krb5_pa_pk_as_req req;
>>> 776         ktest_make_sample_pa_pk_as_req(&req);
>>> 777         encode_run(req, "pa_pk_as_req", "", acc.encode_krb5_pa_pk_as_req);
> 
> It would be interesting to drill down into the value of "req".  I can see
> two possibilities:
> 
> - There's a bug somewhere in the ktest_make_sample_pa_pk_as_req() code path
>    that is making an invalid krb5_pa_pk_as_req structure.
> - There's a bug in the ASN.1 encoder somewhere.
> 
> Either way, assuming you want PKINIT to work, I don't think it's
> something you should ignore.  I'd start with looking at "req" and then
> figuring out what part of req it is trying to encode when you get this
> core dump.  It looks like you omitted part of the stack trace?
> 

I'll try to look into that if I can with my humble C knowledge. I had a 
feeling that it could be an endianess issue since HP-UX on IA64 is big 
endian. I have seen weird stuff like 
https://github.com/cr-marcstevens/sha1collisiondetection/commit/855827c583bc30645ba427885caa40c5b81764d2.

Michael