Re: time_t incompatibility (was Re: Tracking -current + modules)

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Sun, May 17, 2009 at 03:29:45PM -0500, David Young wrote:
> On Sun, May 17, 2009 at 08:30:46PM +0100, Iain Hibbert wrote:
> > I tried a -current kernel a couple of months ago and found some problems
> > but nearly everything is working fine here now and I haven't updated
> > packages yet. (only Pine dumped core but I rebuilt it and it was better)
> 
> Pine may have dumped core because of the time_t changes.  After a
> -current -> -current update that straddled the transition to 64-bit
> time_t, I had to rebuild a few packages that started dumping core:
> openslp-1.2.1nb1.tgz, tcsh-6.14.00.tgz, and (I think) perl-5.8.7nb3.tgz
> were affected.  In a private discussion, people seemed to disagree
> whether those particular packages should have broken or not, however,
> they seemed to agree that *some* packages will.

Did you only update world or rebuild some packages as well?
It is supposed to just work as long as you have binaries consistently
from before *or* after the transistion. Just mixing it is not supposed
to work as soon as interface changes hit 3rd party libraries.

Joerg