Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: HEADS UP: I will be merging christos-time_t by the end of the week

On Mon, Jan 05, 2009 at 07:31:08AM +0000, David Holland wrote:
> On Sun, Jan 04, 2009 at 09:10:21PM -0600, Eric Haszlakiewicz wrote:
>  > Does this mean that, as pkgsrc currently stands, a package built against
>  > the new system (w/ 64-bit time_t) will use a package built against an old
>  >  system but fail at runtime?  yuck.
> Not if we do the mass revbump.
> A process image that's all old stuff (old executable and old libs)
> will work fine. A process image that's all new stuff (new executable
> and new libs) will also work fine. If you mix old and new stuff in the
> same process, e.g. if you recompile a program but not its libs, or
> recompile some libs but not every program that uses them, it may blow
> up at runtime.
> The mass revbump insures that newly compiled stuff will be all new.
> There's no better or less drastic way to do it, because there's no way
> to encode the necessary dependence information in ELF program images.

A mass revbump only solves the problem for those people tracking pkgsrc-HEAD
and NetBSD-current at the same time (and in the right order).  People who are
using 5.0_BETA or older now and upgrade to current (or 6.0) later, while
having updated their pkgsrc in the meanwhile (most likely), will still face
the problem.  For them the mass revision bumping in pkgsrc will only be an
extra annoyance in the meanwhile.

Can't we just yell loudly in UPDATING, on current-users and in the 6.0 release


Home | Main Index | Thread Index | Old Index