[PATCHv2] build: allow some python variable overrides

[Gustavo Zacarias]

On 02/07/2014 03:03 PM, Andrew Bartlett wrote:

> I'm not entirely comfortable with this as a solution, as it isn't
> 'discoverable'.  That is, how will the next user attempting
> cross-compile of Samba find it?
> 
> I realise this is one of the most difficult challenges in this
> cross-compilation saga, and if this was a patch to just the wafsamba end
> of things, I might be willing to say OK, but I don't see how this would
> be acceptable upstream (meaning we have to rework it at some point
> anyway) and it just doesn't feel right, I have to ask that we try
> again. 

Cross compiling properly requires a certain amount of work, documenting
it completely/further would be a step in the right direction.
There are many stumbling blocks around, for instance cross-compiling
python is a PITA which requires a lot of patches that aren't upstream
yet (it's being worked on, slowly, but for python 3.x mostly).
And python is a requirement for samba4, so just to get to the part of "i
want to compile samba4" will require a great deal of work if people want
to follow the "linux form scratch" school of (cross) building.
That's where tools like buildroot, OE and others fill in to do the hard
work for people which would be the most common scenario for cross
compiling complex packages.
Right now i'm under contract to get samba4 building with buildroot, the
main aim being to cross-compile whichever means necessary and for it
work of course (with both being a success), and a secondary aim to
upstream all the necessary changes back upstream in acceptable form of
course :)
Problem is there's an impending deadline as usual with these things, and
an extension may or may not be possible since it's a secondary objective
(famous contribute back to the community that some people like myself
like to stick with).
Please don't take it as a pushy attitude, it's not, i'll still try to
get the necessary changes upstreamed in a proper way, but that will
happen on free personal time which will probably make it happen very slowly.
So with background info and PR out of the way, seems waf isn't helpful
with cross-python situations and it would require some big changes
uptream, what would be the best way to handle this? Other waf changes
would be required to support hostcc as well (the heimdal bits).
Purely python-config as i've suggested before this patch doesn't work
too well, and even if i can sort that out it still relies on a patched
python to get it cross-compiled, and a python-config backport i've
cooked up so it's far from vanilla.
Regards.