Re: checksum argmax fix

[Thomas Klausner <wiz%gatalith.at@localhost>]

On Wed, Aug 07, 2024 at 10:44:10AM +0100, Jonathan Perkin wrote:
> * On 2024-08-07 at 10:40 BST, Thomas Klausner wrote:
> 
> > On Wed, Aug 07, 2024 at 10:32:13AM +0100, Jonathan Perkin wrote:
> > > * On 2024-08-07 at 10:22 BST, Martin Husemann wrote:
> > > 
> > > > On Wed, Aug 07, 2024 at 10:12:30AM +0100, Jonathan Perkin wrote:
> > > > > I'm happy with saying that if folks want to build newer grafana's they need
> > > > > a newer bmake.  I'm not aware of any other packages that are yet close to
> > > > > the argmax limit, so we could even add a dependency on the required bmake
> > > > > version in grafana itself.  If other packages start showing up then we can
> > > > > move the logic to mk.
> > > >
> > > > Can this easily be tested in the makefile and then marked BROKEN automagically
> > > > with old (b)make?
> > > 
> > > It can be tested reasonably easily enough, something like:
> > > 
> > >   $ cd wip/grafana
> > >   $ awk 'NR > 2 && /^BLAKE2s/ {l += length($2) - 1} END{print l}' < $(bmake -v DISTINFO_FILE)
> > >   274119
> > > 
> > > but this would add an unacceptable performance impact to every package,
> > > especially if we also wanted to exec sysctl to compare it to kern.argmax or
> > > wherever each OS stores it, so I'd prefer to keep it manual for now.
> > 
> > If bmake grew a '-V' or '--version' that showed a version number,
> > would that be enough? (Would require a bmake update in pkgsrc, but
> > that's probably rather easy now since it was synced recently.)
> 
> It already has one:
> 
>   $ bmake -V MAKE_VERSION
>   20200524
> 
> I don't know how far back that goes, though it doesn't really matter as if
> it doesn't have one then we know it's too old.

Hm, NetBSD's builtin doesn't provide this?

# make -V MAKE_VERSION
#
 Thomas