[oe] [OE-core] State of bitbake world, Failed tasks 2016-03-09

[Trevor Woerner]

On 03/10/16 06:53, Martin Jansa wrote:
> I'll look at cef3 and chromium issue, my recent patch for meta-browser
> fixed the build for armv7a MACHINEs, so it probably needs to be extended
> a bit to cover warnings in x86 and then duplicate the same to cef3.
> This issue is there for so long, slowing down the builds significantly
> (and causing logs to be much bigger) - I don't use these recipes but I
> hate the failures even more.

Working with the chromium recipe is painful! Each build takes _hours_, 
of just this one recipe alone, even on beefy build hardware. There are 
several PACKAGECONFIG options which affect the build (component build, 
lost context, side painting). There are multiple MACHINEs. You can 
perform a 'Release' or a 'Debug' build. Plus there are two rendering 
systems to support (x11 and wayland). Simply switching between x11 and 
wayland isn't just a matter of setting some build flags, you have to add 
another entire repository to your build from which you have to apply 
dozens of huge patches in order to add an Ozone layer underneath 
chromium so it can talk to wayland/weston.

The matrix of builds to perform and test explodes, not to mention the 
number of days it would take to build all these permutations.

I think one way to help contain this situation would be to split the 
chromium recipe into two: one for x11 and one for wayland. Another thing 
we should consider is wether or not all those permutations are really 
necessary. Is anyone actively using/needing Debug builds? Do we really 
need a component build?