[oe] bitbake does not fail when QA issues encountered

[Andreas Mueller]

On Friday 04 February 2011 14:10:33 Frans Meulenbroeks wrote:
> 2011/2/4 Otavio Salvador <otavio at ossystems.com.br>:
> > On Fri, Feb 4, 2011 at 09:57, Frans Meulenbroeks
> > <fransmeulenbroeks at gmail.com> wrote:
> >> Actually I did not react on the libtool forcing upon everyone, but on
> >> the QA errors (where this thread was about).
> >> That was also the cause of the ostrich remark (which was definitely
> >> not aimed at you in person, but more a general comment, triggered by
> >> the fact that I have some experience with projects being outsourced to
> >> an unnamed country where a bug report also could be "repaired" by
> >> removing the message or the the symptom instead of tackling the root
> >> cause).
> >>
> >> What I see is a lack of interest in fixing the existing QA errors.
> >> I've sent out a few error reports to the list a while ago, (on recipes
> >> that I feel not comfortable with), but unfortunately that did not
> >> result in any action :-(
> >
> > I think we ought to FAIL on QA errors. Distros can ignore the errors
> > if they want but OE as a whole cannot do that.
> 
> I tend to agree with this.
> Actually it would be nice if insane.bbclass would allow a way to
> specify which tests you want to do.
My suggestion: Run alI tests but make it configurable what causes just warnings 
and what causes errors (not too many options e.g. all QA cause errors or 
warning). There are different roles: OE-Users do not want the build to be 
broken - images with QA errors might run without errors. OE developers might
appreciate to get a hard error for the package currently working on or for test 
builds.

Andreas