having llvm available for gallium

I think this might be a casualty of RSS, but when trying to build mesa for the gizmo2 I get:

	| checking for wayland-scanner... no
	| checking for XCB_DRI2... yes
	| checking for x86_64-oe-linux-llvm-config... no
	| checking for llvm-config... no
	| checking for RADEON... yes
	| configure: error: LLVM is required to build Gallium R300 on x86 and x86_64

There already is a dependency on llvm${MESA_LLVM_RELEASE} so I'm not sure what magic is required to make llvm-config available.

Check this:

http://lists.openembedded.org/pipermail/openembedded-core/2017-April/135050.html

Excellent! That works for me :-)

Is this a fix, or a work-around? Is this something that's going to be applied, or is a better fix needed?

Depends on what exactly works for you :).

In my builds I have khem's change reverted for oe-core mesa to find llvm from meta-oe.

I don't use meta-clang to check where does it install the llvm-config scripts if it's different location than what we have in meta-oe or not, but I assume khem was testing his mesa change with meta-clang's llvm.

If it's different than it would be nice to unify that first as discussed in:
http://lists.openembedded.org/pipermail/openembedded-core/2017-April/135468.html
and then use the same path from mesa configuration.

Regards,

Martin 'JaMa' Jansa     jabber: [email protected]

Trevor, what layer were you getting llvm from? Khem's fix looks good from a "stop host contamination" point of view, so we either need to fix meta-something or mesa very soon as oe-core is very close to the final RC builds.

Ross

meta-amd
http://git.yoctoproject.org/cgit/cgit.cgi/meta-amd/tree/common/recipes-core/llvm
MACHINE: steppeeagle

The mentor people fixed things up in meta-amd, so it looks like this was a localized thing:
http://git.yoctoproject.org/cgit/cgit.cgi/meta-amd/diff/common/recipes-graphics/mesa/mesa_git.bbappend