Re: [PATCH v4] base.bbclass: Fix matching of MACHINEOVERRIDES in COMPATIBLE_MACHINE

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2013-04-09 at 11:05 +0100, Phil Blundell wrote:
> On Mon, 2013-04-08 at 18:42 -0300, Otavio Salvador wrote:
> > When a MACHINEOVERRIDES has more than one value, split by ':' as usual
> > OVERRIDES, this were not being properly checked in COMPATIBLE_MACHINE
> > matching as we need to iterate over each SoC family and check if it is
> > compatible or not.
> 
> [...]
> 
> >              import re
> > -            this_machine = d.getVar('MACHINE', True)
> > -            if this_machine and not re.match(need_machine, this_machine):
> > -                this_soc_family = d.getVar('SOC_FAMILY', True)
> > -                if (this_soc_family and not re.match(need_machine, this_soc_family)) or not this_soc_family:
> > -                    raise bb.parse.SkipPackage("incompatible with machine %s (not in COMPATIBLE_MACHINE)" % this_machine)
> > +            compat_machines = (d.getVar('MACHINEOVERRIDES', True) or "").split(":")
> > +            for m in compat_machines:
> > +                if re.match(need_machine, m):
> 
> The checkin comment above doesn't really capture the semantic changes
> going on in this patch.  The comment implies that there was previously
> an issue with MACHINEOVERRIDES not being split when it contains multiple
> entries, but this is misleading: prior to this patch, COMPATIBLE_MACHINE
> wasn't being matched against MACHINEOVERRIDES at all.  So, the effect of
> this is that anything in MACHINEOVERRIDES (which could potentially be
> quite a large set of strings) will be considered as a candidate for
> matching COMPATIBLE_MACHINE.
> 
> It's not abundantly clear to me whether this is a good thing or not, but
> it certainly ought to be accurately described in the commit log.

Agreed. We're getting close to the release so in this case I've
rewritten the commit message since I need to get things pulled together
for the -rc build. People do need to think more about commit messages
and how to convey the right information in them...

Cheers,

Richard