Re: [PATCH 0/4] Fix builds for MIPS64 N32

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

On Wed, 2015-07-29 at 18:00 +0300, Dmitry Eremin-Solenikov wrote:
> > Recent check-ins have changed some things.  The compile side was working for a
> > while, but not the compilation side (using it).
> >
> > I've been working with RP on IRC, and he's come up with some additional patches.
> >  (including this set of 4..)
> >
> > So I think the set is fine, it's up to the other pieces that need to mesh with
> > this to produce the working SDK.
> 
> Ok. So what would be the plan then? If you need any additional tests
> from my side,
> I can provide them.

Sorry for the delays on this. I found several issues and ended up with
some quite subtle gcc changes to make all this work. I've put both our
changes into master-next. The local.conf testing setup I've been using
is:

MACHINE = "qemumips64"
require conf/multilib.conf
DEFAULTTUNE = "mips64-n32"
MULTILIBS = "multilib:lib64 multilib:lib32"
DEFAULTTUNE_virtclass-multilib-lib64 = "mips64"
DEFAULTTUNE_virtclass-multilib-lib32 = "mips32r2"

then the build/test commands:

bitbake core-image-minimal -c populate_sdk
bitbake core-image-minimal -c testsdk

The nice thing is the latter automated test part now passes for this
configuration (and its tests all three multilibs).

What remains is to run these changes through the autobuilder and make
sure nothing else breaks as a result of this patchset. I'm about to do
some business travel which means my ability to debug any further issues
will be limited, compiling toolchains on the laptop isn't realistic. I
may therefore need help in ironing out any further wrinkles or we'll
have to wait until I get back in a couple of weeks.

Cheers,

Richard