Hi,

thanks for the feedback. Yes, I accidentally added the patch I made for mesa 24.2.1. Currently I have it like this, which should also work:

diff --git a/src/gallium/frontends/rusticl/meson.build b/src/gallium/frontends/rusticl/meson.build

index 84df0a9edb9..20459f73993 100644

--- a/src/gallium/frontends/rusticl/meson.build

+++ b/src/gallium/frontends/rusticl/meson.build

@@ -122,6 +122,7 @@ endif

rusticl_bindgen_c_args = [

'-fno-builtin-malloc',

+ '-I@include@', '-I@include_cpp@', '-I@include_cpp_target_sys@',

]

cl_c_args = [

One assumption on the shader issue. intel_clc is currently added only for x86-64. If you are building for another platform intel_clc is not installed for mesa-native, but required for 'iris'.

Changing this line:

GALLIUMDRIVERS:append ="${@bb.utils.contains_any('PACKAGECONFIG', 'opencl-rusticl opencl-clover', ',iris', '', d)}"

to:

GALLIUMDRIVERS:append:x86-64 ="${@bb.utils.contains_any('PACKAGECONFIG', 'opencl-rusticl opencl-clover', ',iris', '', d)}"

might remove the need to disable raytracing in EXTRA_OEMESON

Alternatively also removing x86-64 from these lines:

EXTRA_OEMESON:append:class-target:x86-64 = " ${@bb.utils.contains_any('PACKAGECONFIG', 'opencl-rusticl opencl-clover', '-Dintel-clc=system', '', d)}"

EXTRA_OEMESON:append:class-native:x86-64 = " ${@bb.utils.contains_any('PACKAGECONFIG', 'opencl-rusticl opencl-clover', '-Dintel-clc=enabled -Dinstall-intel-clc=true -Dgallium-rusticl=false', '', d)}"

might help, but I'm unsure if intel_clc would be of any use apart from x86-64

On Fri, Sep 13 2024 at 12:43:46 PM +02:00:00, Böszörményi Zoltán <zboszor@gmail.com> wrote:

2024. 09. 13. 12:40 keltezéssel, Zoltan Boszormenyi via lists.openembedded.org írta:
2024. 09. 09. 11:45 keltezéssel, Markus Volk via lists.openembedded.org írta:
On Mon, Sep 9 2024 at 10:46:59 AM +02:00:00, Alexander Kanavin <alex.kanavin@gmail.com> wrote:
This does need to be properly looked into, there's probably a better patch possible, or a build setting we're missing, or at least an upstream ticket to be filed. It's not a good patch when it says 'workaround' and doesn't explain the issue and the fix.
Fully agree. This is ugly as hell. Just a quick and dirty hack to get things working (which is always a good starting point) but by no means a solution for oe-core. Thats why I marked it RFC
Don't call it a "workaround" then. Call it "pass in recipe-sysroot include directories" because that's what this patch does combined with the mesa.inc changes.
The underlying problem seems to be that bindgen does not seem to pick up CFLAGS or CPPFLAGS automatically, so it needs the extra -I options passed in under Yocto to look for headers in the correct places.
Though it's not complete in my testing. I needed this patch against 24.2.2 so I could build Mesa Rusticl, plus EXTRA_OEMESON += "-Dintel-rt=disabled" in my bbappend because the iris driver failed to build otherwise, complaining for every generated file from *.cl about "shader not available, consider installing libclc" or something like that, eventually failing the build. diff --git a/src/gallium/frontends/rusticl/meson.build b/src/gallium/frontends/rusticl/meson.build index eef09d8f01e..efac4cab964 100644 --- a/src/gallium/frontends/rusticl/meson.build +++ b/src/gallium/frontends/rusticl/meson.build @@ -146,6 +146,7 @@ rusticl_opencl_bindings_rs = rust.bindgen( rusticl_bindgen_c_args, pre_args, cl_c_args, + '-I@include@', '-I@include_cpp@', '-I@include_cpp_target_sys@' ], args : [ rusticl_bindgen_args, @@ -190,6 +191,7 @@ rusticl_llvm_bindings_rs = rust.bindgen( c_args : [ rusticl_bindgen_c_args, pre_args, + '-I@include@', '-I@include_cpp@', '-I@include_cpp_target_sys@' ], dependencies : [ dep_clang,

-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#204478): https://lists.openembedded.org/g/openembedded-core/message/204478 Mute This Topic: https://lists.openembedded.org/mt/108340502/3617728 Group Owner: openembedded-core+owner@lists.openembedded.org Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [zboszor@gmail.com] -=-=-=-=-=-=-=-=-=-=-=-