[OE-core] Multiconfig and native recipes

[richard.purdie at linuxfoundation.org]

On Thu, 2018-12-13 at 17:41 +0100, Ricardo Ribalda Delgado wrote:
> Hi Richard
> 
> Thanks for your fast response
> On Thu, Dec 13, 2018 at 5:19 PM Richard Purdie
> <richard.purdie at linuxfoundation.org> wrote:
> > On Thu, 2018-12-13 at 16:19 +0100, Ricardo Ribalda Delgado wrote:
> > You haven't said how your configurations were different.
> > 
> > With multiconfig you have to use it responsibly and a little
> > caution.
> > 
> > It sounds like you're using the same TMPDIR for these builds but
> > the
> > builds are differently configured. That wouldn't be expected to
> > work.
> > The builds can only share TMPDIR if the configurations are
> > compatible
> > with each other (same DISTRO type config, different MACHINE for
> > example).
> > 
> 
> That is exactly my usecase. I wan to build for all the machines with
> exactly the same distro:
> 
> ricardo at neopili:~/curro/yocto/europa/build$ for a in
> conf/multiconfig/*; do echo $a; cat $a; done
> conf/multiconfig/inteli7.conf
> MACHINE = "inteli7"
> conf/multiconfig/qt5022.conf
> MACHINE = "qt5022"
> conf/multiconfig/qt5506.conf
> MACHINE = "qt5506"
> 
> I am now trying to hack bitbake/lib/bb/cooker.py to convert all tasks
> multiconfig:*:AAAAA-native into AAAAA-native
> it might not be the final sollution, but maybe we can use it to
> discuss further options.

Rather than hacking this, I'd suggest you look at why it thinks it
needs to run this native target three times.

In a normal build, if you select one of the machines, "bitbake erlang-
native", then change machines and run the same command, does it
rebuild?

If so, the task hashes are different. The next thing to do would be to
run bitbake-diffsigs and find out why they're different.

There are signature tests in oe-selftest which will find similar
problems like this, assuming I'm guessing at the problem correctly.

Cheers,

Richard