Re: PATCH: Fix __WORDSIZE for ia64.

["H. J. Lu" <hjl at lucon dot org>]

On Thu, Jul 31, 2003 at 05:39:50PM -0700, Roland McGrath wrote:
> > On Thu, Jul 31, 2003 at 05:33:09PM -0700, Roland McGrath wrote:
> > > > On Thu, Jul 31, 2003 at 03:51:19PM -0700, Roland McGrath wrote:
> > > > > The meaning of this is to change the utmp file format on IA64.
> > > > > Which systems have used which of the two formats?
> > > > 
> > > > That is the purpose. Otherwise, ia32 binaries won't work right on
> > > > ia64.
> > > 
> > > Please answer the question.
> > 
> > As I said, ia64 is the same on this as x86-64, both are 64bit systems
> > which also run ia32 binaries.
> 
> That does not answer the question.  If the existing definition has been
> using on IA64 systems, then they have the incompatible file format already.
> x86-64 had this format settled early enough that we were not too concerned
> about existing systems with utmp files in the wrong format.  I understand
> the problems of this format not staying identical for 64-bit platforms and
> the 32-bit platforms they support emulation of.  But that does not remove
> the problem of this format changing from what it has heretofore been on
> IA64 systems, and we need to discuss the ramifications before sliding in a
> change with as little comment as you would have us do.

The problem is none of the ia32 programs which access utmp work right
on ia64 while x86-64 does the right thing. Yes, this patch will change
the format of utmp. That is the whole purpose. As it stands now, utmp
on ia64 doesn't work with ia32 programs. With this change, the only
loss is we can no longer access the old utmp file. I think it is the
price we should pay to fix the problem. If we really want, we can
provide a preload library to allow access the old utmp files.


H.J.