On Tue, May 08, 2001 at 11:51:38AM -0400, mcmahill@mtl.mit.edu wrote:
> 
> On Tue, 8 May 2001, David Brownlee wrote:
> 
> > On Tue, 8 May 2001, Alistair Crooks wrote:
> > 
> > > On Tue, May 08, 2001 at 11:02:51AM +0100, David Brownlee wrote:
> > > > 	citrix_ica has a distinfo.i386 and distinfo.sparc, while navigator
> > > > 	keeps different architectures in the same distinfo file (and needs
> > > > 	more care when rebuilding distinfos)
> > > >
> > > > --
> > > > 		David/absolute		-- www.netbsd.org: No hype required --
> > > >
> > >
> > > I'd much prefer it if we had one distinfo file, without the arch
> > > suffix.  If this means we have to be more careful when upgrading
> > > packages, then so be it.
> > 
> > 	OK - Packages.txt updated to match.
> > 
> > 	Does it make sense to have a variable set in the Makefile
> > 	for packages with multiarch distinfos? If set a 'make mdi'
> > 	could display a warning then echo the new info to stdout
> > 	rather than rewrite distinfo.
> 
> maybe.  i don't like how often i see commits by others and myself like
> "restore solaris cksum".  
> 
> Whats considered the right way to handle patches when we have arch
> specific distfiles and end up with patches which are then also arch
> specific?  Note I'm speaking of cases where the patch is for a file which
> doesn't even exist in the distfile for another arch so simply
> 
> #ifdef (__alpha__)
> 
> or whatever won't work.
> 
> -Dan
> 

I've not been aware of any packages with this behaviour before now
(which doesn't prove anything except I'm growing old and forgetful).

So I guess we get to choose what to do.

Options that spring to mind:

1. arch-specific distinfo files (I'm not a fan of this because I'd like
to get away from specialised customised things).

2. special handling within distinfo itself. (I loathe this idea, included
only for completeness)

3. Perhaps split the package into two architecture-specific packages?

4. Something else I've not thought about.

Regards,
Al