Re: CVS commit: pkgsrc/emulators/qemu

[maya%NetBSD.org@localhost]

On Fri, Mar 06, 2020 at 03:18:18PM +0000, Roy Marples wrote:
> On 06/03/2020 13:18, Greg Troxel wrote:
> > "Kamil Rytarowski" <kamil%netbsd.org@localhost> writes:
> > 
> > > Module Name:	pkgsrc
> > > Committed By:	kamil
> > > Date:		Fri Mar  6 09:19:43 UTC 2020
> > > 
> > > Modified Files:
> > > 	pkgsrc/emulators/qemu: Makefile
> > > 
> > > Log Message:
> > > qemu: Conditionalize NVMM for NetBSD
> > > 
> > > Build NVMM support whenever there is /usr/include/nvmm.h available.
> > 
> > I am uncomfortable with this method of conditionalizing, as it seems to
> > move away from repeatable builds.
> > 
> > Would it be sufficient/reasonable to instead just enable the option by
> > default on NetBSD >= 9?
> 
> It's not that simple.
> What if NetBSD-9 was built without NVMM support?
> We can build NetBSD without YP support and have to deal with similar stuff
> .... see python about that for example.
> 
> Also see the curses framework is is very condition based on the platform.
> 
> Roy

I am still drawing a different conclusion.
My first priority is to have packages just do the correct thing when one
runs "./configure; make". It's a real treat when it works that way and I
don't need to spend 30 minutes reading the Makefile and patches to see
what I need to reproduce when building things outside of pkgsrc.

MKYP is extremely esoteric, so there were issues with it. That happens.
I don't think MKNVMM needs to exist. It's a kernel module and a small
library. And if someone added tunables for many things in NetBSD I don't
think it would look pretty.