Re: [PATCH] base.bbclass: Use bb.warn instead of bb.error for deprecation notification.

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

On Tue, 2015-02-03 at 11:02 -0200, Otavio Salvador wrote:
> On Tue, Feb 3, 2015 at 11:00 AM, Burton, Ross <ross.burton@intel.com> wrote:
> >
> > On 3 February 2015 at 12:53, Otavio Salvador <otavio@ossystems.com.br>
> > wrote:
> >>
> >> > last time I checked (some days ago) it was not necessarily fatal. I was
> >> > able to
> >> > build complete image collecting a pile of PR_INC errors. That's what
> >> > confused me...
> >> > I'll check if I still have the log on another machine later.
> >>
> >> Florian has a point here. We should use 'bb.fatal' and abort all build
> >> as soon as possible.
> >
> >
> > Its an error in a specific recipe, so why should that abandon the build
> > immediately?
> 
> It ends changing the content of the recipe and might raise a rebuild
> of several things. Which ends being invalid as PRINC would make an
> error in the end. So why not fail soon and ensure it is act upon it
> ASAP?

It errors during parsing which stops the build going any further. It
doesn't really matter whether its a bb.fatal or bb.error, both would
have the same effect here which is to stop the build at parse time.

It used to be a bb.warn(), people ignored it. Personally I'd like to
drop the thing entirely but people wanted an error so we have that now
for this release. We're not changing it back to a warn().

Cheers,

Richard