[v4,0/9] mesa: upgrade to 25.0.2

Message ID	20250331-mesa-25-v4-0-e62fc54a6efc@oss.qualcomm.com
Headers	show Return-Path: <dmitry.baryshkov@oss.qualcomm.com> ip: 205.220.168.131, mailfrom: dmitry.baryshkov@oss.qualcomm.com) From: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com> Subject: [PATCH v4 0/9] mesa: upgrade to 25.0.2 Date: Mon, 31 Mar 2025 15:21:33 +0300 Message-Id: <20250331-mesa-25-v4-0-e62fc54a6efc@oss.qualcomm.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" To: openembedded-core@lists.openembedded.org Cc: Khem Raj <raj.khem@gmail.com>, =?utf-8?b?Wm9sdMOhbiBCw7ZzesO2cm3DqW55?= =?utf-8?b?aQ==?= <zboszor@gmail.com> Content-Transfer-Encoding: quoted-printable
Series	mesa: upgrade to 25.0.2 \| expand [v4,0/9] mesa: upgrade to 25.0.2 [v4,1/9] llvm: move libLLVM.so.N.M to llvm-libllvm package [v4,2/9] llvm: use OECMAKE_SOURCEPATH to specify cmake dir [v4,3/9] llvm: handle libdir in llvm-config [v4,4/9] llvm: also use llvm-config wrapper for nativesdk recipes [v4,5/9] llvm: support building libclc [v4,6/9] llvm: add SPIRV-LLVM-Translator support [v4,7/9] mesa: upgrade 24.0.7 -> 25.0.2 [v4,8/9] bindgen-cli: a tool to generate Rust bindings [v4,9/9] mesa: add support for RustiCL under PACKAGECONFIG "opencl"

Dmitry Baryshkov March 31, 2025, 12:21 p.m. UTC

libclc is required in order to build Intel drivers in Mesa 24.1 and
later. Thus OE-Core is currently locked to the 24.0.x branch, which is
no longer maintained, having last release in June 2024.

A clean solution is to package Clang, then use it to build libclc and
SPIRV-LLVM-Translator (required for Mesa OpenCL support). However
creating proper recipes for Clang is a long process (ongoing by Khem
Raj).

This patchset attempts to solve the issue in a simpler way: by building
libclc and SPIRV-LLVM-Translator together with the LLVM. This way we
don't have to create target Clang recipes, there is no extra toolchain
to support, etc. Clang binaries are creted as a part of the build
process (in order to support building target packages), however it only
hits the LLVM sysroot-native and is only used by the LLVM target build.

Future work by Khem (once Clang is properly supported in OE-Core) should
allow us to create proper recipes for libclc and spirv-llvm-translator
and drop them from the llvm / clang recipe.

I think this approach provides a good balance between OE-Core being
stuck with the outdated Mesa release and being able to upgrade it,
bringing support for new hardware.

Last two patches of the series bring in RustiCL support, a new OpenCL
implementation within Mesa. They are optional for the sake of the
upgrade and can be dropped if there is any issue with those. RustiCL
support is mostly compile-tested at this point.

Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
---
Changes in v4:
- Fixed QA issues reported against lib32-mesa-megadriver (Richard)
- Removed release notes link from the mesa's commit message (Alexander)
- Added link to the rust.bindgen() issue to RustiCL-related
  cross-compilation patch (Alexander)
- Renamed 'spirv' PACKAGECONFIG option to 'spirv-llvm-translator' in
  order to be more precise
- Link to v3: https://lore.kernel.org/r/20250329-mesa-25-v3-0-f4d0328eb867@oss.qualcomm.com

Changes in v3:
- Dropped llvm/fix-native-compilation.patch and
  llvm/0001-cmake-modules-CrossCompile-pass-SPIR-V-headers-path-.patch
  in favour of installing prepare_builtins into the sysroot, dropped
  corresponding EXTRA_OECMAKE params
- Dropepd -DCLANG_TABLEGEN from llvm's EXTRA_OECMAKE
- Enabled llvm-config wrapper for nativesdk too, fixing QA errors for
  nativesdk-mesa
- Added --libdir handling to llvm-config, dropping the need for
  mesa/0001-meson-use-target-path-for-Clang-libdir.patch
- Dropped extra LIBCEXTENSION from BINDGEN_TARGET, it is already a part
  of TARGET_OS
- Expanded commit messages to explain the patches
- Link to v2: https://lore.kernel.org/r/20250327221807.2551544-1-dmitry.baryshkov@oss.qualcomm.com

Changes in v2:
- Dropped YOCTO_ALTERNATE_LIBDIR patch
- Reworked mesa / LLVM integration to always use get_option('libdir')
- Reenabled AMD Vulkan driver (got disabled because of c&p error)
- Moved mesa-clc options to the libclc PACKAGECONFIG to fix several
  build errors
- Link to v1: https://lore.kernel.org/r/20250327135520.3920189-1-dmitry.baryshkov@oss.qualcomm.com/

---
Dmitry Baryshkov (8):
      llvm: move libLLVM.so.N.M to llvm-libllvm package
      llvm: use OECMAKE_SOURCEPATH to specify cmake dir
      llvm: handle libdir in llvm-config
      llvm: also use llvm-config wrapper for nativesdk recipes
      llvm: support building libclc
      llvm: add SPIRV-LLVM-Translator support
      mesa: upgrade 24.0.7 -> 25.0.2
      bindgen-cli: a tool to generate Rust bindings

Zoltán Böszörményi (1):
      mesa: add support for RustiCL under PACKAGECONFIG "opencl"

 meta/conf/distro/include/maintainers.inc           |   1 +
 .../bindgen-cli/bindgen-cli-crates.inc             | 186 +++++++++++++++++++++
 .../bindgen-cli/bindgen-cli_0.71.1.bb              |  19 +++
 meta/recipes-devtools/llvm/llvm/llvm-config        |   3 +
 .../llvm/llvm/spirv-internal-build.patch           |  30 ++++
 meta/recipes-devtools/llvm/llvm_20.1.0.bb          |  77 ++++++++-
 ...evert-meson-do-not-pull-in-clc-for-clover.patch |  53 ------
 ...d-Include-missing-llvm-IR-header-Module.h.patch |  41 -----
 .../files/0001-drisw-fix-build-without-dri3.patch  |  58 -------
 ...eno-don-t-encode-build-path-into-binaries.patch |  46 ++++-
 ...vide-cross-compilation-arguments-to-LLVM-.patch |  67 ++++++++
 ...ext-don-t-try-zink-if-not-enabled-in-mesa.patch |  42 -----
 meta/recipes-graphics/mesa/mesa.inc                | 101 ++++++-----
 13 files changed, 470 insertions(+), 254 deletions(-)
---
base-commit: dc2ef2cc1183a14cd3a05d388a1d0485a1bc8d20
change-id: 20250328-mesa-25-6b78a0ce2c90

Best regards,

Mathieu Dubois-Briand April 1, 2025, 7:40 a.m. UTC | #1

On Mon Mar 31, 2025 at 2:21 PM CEST, Dmitry Baryshkov via lists.openembedded.org wrote:
> libclc is required in order to build Intel drivers in Mesa 24.1 and
> later. Thus OE-Core is currently locked to the 24.0.x branch, which is
> no longer maintained, having last release in June 2024.
>
> A clean solution is to package Clang, then use it to build libclc and
> SPIRV-LLVM-Translator (required for Mesa OpenCL support). However
> creating proper recipes for Clang is a long process (ongoing by Khem
> Raj).
>
> This patchset attempts to solve the issue in a simpler way: by building
> libclc and SPIRV-LLVM-Translator together with the LLVM. This way we
> don't have to create target Clang recipes, there is no extra toolchain
> to support, etc. Clang binaries are creted as a part of the build
> process (in order to support building target packages), however it only
> hits the LLVM sysroot-native and is only used by the LLVM target build.
>
> Future work by Khem (once Clang is properly supported in OE-Core) should
> allow us to create proper recipes for libclc and spirv-llvm-translator
> and drop them from the llvm / clang recipe.
>
> I think this approach provides a good balance between OE-Core being
> stuck with the outdated Mesa release and being able to upgrade it,
> bringing support for new hardware.
>
> Last two patches of the series bring in RustiCL support, a new OpenCL
> implementation within Mesa. They are optional for the sake of the
> upgrade and can be dropped if there is any issue with those. RustiCL
> support is mostly compile-tested at this point.
>
> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
> ---

Hi Dmitry,

Thanks for the new version.

It looks like we still have issues on the autobuilder, now during
oe-selftests. I truncated a bit the logs below.

ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
...
| FAILED: src/intel/shaders/intel_gfx80_shaders.spv
| /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
| /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found

But also a lot of issues with sstate tests:

2025-03-31 19:31:36,349 - oe-selftest - INFO - 6: 46/49 214/622 (88.62s) (0 failed) (sstatetests.SStateHashSameSigs.test_sstate_32_64_same_hash)
2025-03-31 19:31:36,349 - oe-selftest - INFO - testtools.testresult.real._StringException: Traceback (most recent call last):
  File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 418, in test_sstate_32_64_same_hash
    self.sstate_hashtest("i686")
  File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 409, in sstate_hashtest
    self.assertCountEqual(files1, files2)
  File "/usr/lib/python3.11/unittest/case.py", line 1233, in assertCountEqual
    self.fail(msg)
  File "/usr/lib/python3.11/unittest/case.py", line 703, in fail
    raise self.failureException(msg)
AssertionError: Element counts were not equal:
First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_install.sigdata.f1121eb56f1f6b8335722e32d0e21eb777e95ab4af83b777971f3666d1265bed'
First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_spdx.sigdata.4c53505a0e3a5bba214d854ae8eea7c551f794d6b85392564d7c3923180e5d7f'
First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_package_spdx.sigdata.cefe11ad38c3c40968e4757314ad08b91f9584c9845cbcf08cc8635e1216912e'
First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_populate_sysroot.sigdata.2e3e3b9c17855a8a865dd9e99c002016e697024fc93cdfed690b7d54bb05e34e'
First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_prepare_recipe_sysroot.sigdata.73d0df5170ce762727a517856505f3505b2c47f479affdddb242ddcad34d5e29'
...

https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264
https://autobuilder.yoctoproject.org/valkyrie/#/builders/23/builds/1350
https://autobuilder.yoctoproject.org/valkyrie/#/builders/48/builds/1214

Can you have a look a these please?

Alexander Kanavin April 1, 2025, 8:06 a.m. UTC | #2

On Tue, 1 Apr 2025 at 09:40, Mathieu Dubois-Briand via
lists.openembedded.org
<mathieu.dubois-briand=bootlin.com@lists.openembedded.org> wrote:
> It looks like we still have issues on the autobuilder, now during
> oe-selftests. I truncated a bit the logs below.
>
> ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
> ...
> | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
> | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
> | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found

The particular failing selftests are:

2025-03-31 21:57:27,426 - oe-selftest - INFO - RESULTS -
runtime_test.TestImage.test_testimage_virgl_headless: FAILED (348.30s)
2025-03-31 21:57:27,426 - oe-selftest - INFO - RESULTS -
sstatetests.SStateHashSameSigs.test_sstate_32_64_same_hash: FAILED
(88.63s)
2025-03-31 21:57:27,426 - oe-selftest - INFO - RESULTS -
sstatetests.SStateHashSameSigs.test_sstate_sdk_arch_same_hash: FAILED
(102.40s)

On Fedora mesa-native seems to build ok in the virgl_headless test,
but then there's a different error:

runqemu - ERROR - Failed to run qemu: MESA-LOADER: failed to open dri:
/srv/pokybuild/yocto-worker/genericx86/build/build/tmp/work/x86_64-linux/mesa-native/25.0.2/recipe-sysroot-native/usr/lib/gbm/dri_gbm.so:
cannot open shared object file: No such file or directory (search
paths /srv/pokybuild/yocto-worker/genericx86/build/build/tmp/work/x86_64-linux/mesa-native/25.0.2/recipe-sysroot-native/usr/lib/gbm,
suffix _gbm)

Alex

Mathieu Dubois-Briand April 1, 2025, 8:27 a.m. UTC | #3

On Tue Apr 1, 2025 at 9:40 AM CEST, Mathieu Dubois-Briand wrote:
> On Mon Mar 31, 2025 at 2:21 PM CEST, Dmitry Baryshkov via lists.openembedded.org wrote:
> > libclc is required in order to build Intel drivers in Mesa 24.1 and
> > later. Thus OE-Core is currently locked to the 24.0.x branch, which is
> > no longer maintained, having last release in June 2024.
> >
> > A clean solution is to package Clang, then use it to build libclc and
> > SPIRV-LLVM-Translator (required for Mesa OpenCL support). However
> > creating proper recipes for Clang is a long process (ongoing by Khem
> > Raj).
> >
> > This patchset attempts to solve the issue in a simpler way: by building
> > libclc and SPIRV-LLVM-Translator together with the LLVM. This way we
> > don't have to create target Clang recipes, there is no extra toolchain
> > to support, etc. Clang binaries are creted as a part of the build
> > process (in order to support building target packages), however it only
> > hits the LLVM sysroot-native and is only used by the LLVM target build.
> >
> > Future work by Khem (once Clang is properly supported in OE-Core) should
> > allow us to create proper recipes for libclc and spirv-llvm-translator
> > and drop them from the llvm / clang recipe.
> >
> > I think this approach provides a good balance between OE-Core being
> > stuck with the outdated Mesa release and being able to upgrade it,
> > bringing support for new hardware.
> >
> > Last two patches of the series bring in RustiCL support, a new OpenCL
> > implementation within Mesa. They are optional for the sake of the
> > upgrade and can be dropped if there is any issue with those. RustiCL
> > support is mostly compile-tested at this point.
> >
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
> > ---
>
> Hi Dmitry,
>
> Thanks for the new version.
>
> It looks like we still have issues on the autobuilder, now during
> oe-selftests. I truncated a bit the logs below.
>
> ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
> ...
> | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
> | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
> | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found
>
> But also a lot of issues with sstate tests:
>
> 2025-03-31 19:31:36,349 - oe-selftest - INFO - 6: 46/49 214/622 (88.62s) (0 failed) (sstatetests.SStateHashSameSigs.test_sstate_32_64_same_hash)
> 2025-03-31 19:31:36,349 - oe-selftest - INFO - testtools.testresult.real._StringException: Traceback (most recent call last):
>   File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 418, in test_sstate_32_64_same_hash
>     self.sstate_hashtest("i686")
>   File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 409, in sstate_hashtest
>     self.assertCountEqual(files1, files2)
>   File "/usr/lib/python3.11/unittest/case.py", line 1233, in assertCountEqual
>     self.fail(msg)
>   File "/usr/lib/python3.11/unittest/case.py", line 703, in fail
>     raise self.failureException(msg)
> AssertionError: Element counts were not equal:
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_install.sigdata.f1121eb56f1f6b8335722e32d0e21eb777e95ab4af83b777971f3666d1265bed'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_spdx.sigdata.4c53505a0e3a5bba214d854ae8eea7c551f794d6b85392564d7c3923180e5d7f'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_package_spdx.sigdata.cefe11ad38c3c40968e4757314ad08b91f9584c9845cbcf08cc8635e1216912e'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_populate_sysroot.sigdata.2e3e3b9c17855a8a865dd9e99c002016e697024fc93cdfed690b7d54bb05e34e'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_prepare_recipe_sysroot.sigdata.73d0df5170ce762727a517856505f3505b2c47f479affdddb242ddcad34d5e29'
> ...
>
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/23/builds/1350
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/48/builds/1214
>
> Can you have a look a these please?

Also, I suspect this series is linked with this Xorg segfault in
parselogs.ParseLogsTest.test_parselogs:

Central error: Fatal server error:
***********************
[     1.468] 	compiled for 1.21.1.16, module version = 1.0.1
[     1.468] 	ABI class: X.Org ANSI C Emulation, version 0.4
[     1.512] (EE)
[     1.512] (EE) Backtrace:
[     1.512] (EE) 0: /usr/bin/Xorg (xorg_backtrace+0x52) [0x566b2522]
[     1.512] (EE) 1: /usr/bin/Xorg (0x56583000+0x1332b5) [0x566b62b5]
[     1.512] (EE) 2: /lib/libc.so.6 (0xf75e8000+0x34060) [0xf761c060]
[     1.512] (EE) 3: /usr/lib/libgallium-25.0.2.so (0xf5140000+0x512634) [0xf5652634]
[     1.512] (EE)
[     1.512] (EE) Segmentation fault at address 0x0
[     1.512] (EE)
Fatal server error:
[     1.512] (EE) Caught signal 11 (Segmentation fault). Server aborting

https://autobuilder.yoctoproject.org/valkyrie/#/builders/11/builds/1274

Alexander Kanavin April 1, 2025, 8:41 a.m. UTC | #4

On Tue, 1 Apr 2025 at 10:27, Mathieu Dubois-Briand via
lists.openembedded.org
<mathieu.dubois-briand=bootlin.com@lists.openembedded.org> wrote:
> Also, I suspect this series is linked with this Xorg segfault in
> parselogs.ParseLogsTest.test_parselogs:
>
> Central error: Fatal server error:
> ***********************
> [     1.468]    compiled for 1.21.1.16, module version = 1.0.1
> [     1.468]    ABI class: X.Org ANSI C Emulation, version 0.4
> [     1.512] (EE)
> [     1.512] (EE) Backtrace:
> [     1.512] (EE) 0: /usr/bin/Xorg (xorg_backtrace+0x52) [0x566b2522]
> [     1.512] (EE) 1: /usr/bin/Xorg (0x56583000+0x1332b5) [0x566b62b5]
> [     1.512] (EE) 2: /lib/libc.so.6 (0xf75e8000+0x34060) [0xf761c060]
> [     1.512] (EE) 3: /usr/lib/libgallium-25.0.2.so (0xf5140000+0x512634) [0xf5652634]
> [     1.512] (EE)
> [     1.512] (EE) Segmentation fault at address 0x0
> [     1.512] (EE)
> Fatal server error:
> [     1.512] (EE) Caught signal 11 (Segmentation fault). Server aborting
>
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/11/builds/1274

This one is specific to x86-x32. We've had to do this previously,
maybe it's not enough, or there's some other issue:
https://web.git.yoctoproject.org/poky/commit/?id=f32128bda2a035b01d2dbf7bc73738a310eb955c

I'd say we drop testing anything graphical from that platform, as the
only way to get it started is with fbdev. No one's using it.

Alex

Dmitry Baryshkov April 1, 2025, 4:19 p.m. UTC | #5

Hello Mathieu,

On 01/04/2025 10:40, Mathieu Dubois-Briand wrote:
> On Mon Mar 31, 2025 at 2:21 PM CEST, Dmitry Baryshkov via lists.openembedded.org wrote:
>> libclc is required in order to build Intel drivers in Mesa 24.1 and
>> later. Thus OE-Core is currently locked to the 24.0.x branch, which is
>> no longer maintained, having last release in June 2024.
>>
>> A clean solution is to package Clang, then use it to build libclc and
>> SPIRV-LLVM-Translator (required for Mesa OpenCL support). However
>> creating proper recipes for Clang is a long process (ongoing by Khem
>> Raj).
>>
>> This patchset attempts to solve the issue in a simpler way: by building
>> libclc and SPIRV-LLVM-Translator together with the LLVM. This way we
>> don't have to create target Clang recipes, there is no extra toolchain
>> to support, etc. Clang binaries are creted as a part of the build
>> process (in order to support building target packages), however it only
>> hits the LLVM sysroot-native and is only used by the LLVM target build.
>>
>> Future work by Khem (once Clang is properly supported in OE-Core) should
>> allow us to create proper recipes for libclc and spirv-llvm-translator
>> and drop them from the llvm / clang recipe.
>>
>> I think this approach provides a good balance between OE-Core being
>> stuck with the outdated Mesa release and being able to upgrade it,
>> bringing support for new hardware.
>>
>> Last two patches of the series bring in RustiCL support, a new OpenCL
>> implementation within Mesa. They are optional for the sake of the
>> upgrade and can be dropped if there is any issue with those. RustiCL
>> support is mostly compile-tested at this point.
>>
>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
>> ---
> 
> Hi Dmitry,
> 
> Thanks for the new version.
> 
> It looks like we still have issues on the autobuilder, now during
> oe-selftests. I truncated a bit the logs below.
> 
> ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
> ...
> | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
> | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
> | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found


I've fixed all other issues, except the x32 (working on reproducing it 
now) and this one. Would you have a pointer to the build job?


> 
> But also a lot of issues with sstate tests:
> 
> 2025-03-31 19:31:36,349 - oe-selftest - INFO - 6: 46/49 214/622 (88.62s) (0 failed) (sstatetests.SStateHashSameSigs.test_sstate_32_64_same_hash)
> 2025-03-31 19:31:36,349 - oe-selftest - INFO - testtools.testresult.real._StringException: Traceback (most recent call last):
>    File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 418, in test_sstate_32_64_same_hash
>      self.sstate_hashtest("i686")
>    File "/srv/pokybuild/yocto-worker/oe-selftest-debian/build/meta/lib/oeqa/selftest/cases/sstatetests.py", line 409, in sstate_hashtest
>      self.assertCountEqual(files1, files2)
>    File "/usr/lib/python3.11/unittest/case.py", line 1233, in assertCountEqual
>      self.fail(msg)
>    File "/usr/lib/python3.11/unittest/case.py", line 703, in fail
>      raise self.failureException(msg)
> AssertionError: Element counts were not equal:
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_install.sigdata.f1121eb56f1f6b8335722e32d0e21eb777e95ab4af83b777971f3666d1265bed'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_spdx.sigdata.4c53505a0e3a5bba214d854ae8eea7c551f794d6b85392564d7c3923180e5d7f'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_create_package_spdx.sigdata.cefe11ad38c3c40968e4757314ad08b91f9584c9845cbcf08cc8635e1216912e'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_populate_sysroot.sigdata.2e3e3b9c17855a8a865dd9e99c002016e697024fc93cdfed690b7d54bb05e34e'
> First has 1, Second has 0:  '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53795/tmp-sstatesamehash/stamps/core2-32-poky-linux/mesa/25.0.2.do_prepare_recipe_sysroot.sigdata.73d0df5170ce762727a517856505f3505b2c47f479affdddb242ddcad34d5e29'
> ...
> 
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/23/builds/1350
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/48/builds/1214
> 
> Can you have a look a these please?
> 


-- 
With best wishes
Dmitry

Alexander Kanavin April 1, 2025, 4:28 p.m. UTC | #6

On Tue, 1 Apr 2025 at 18:19, Dmitry Baryshkov via
lists.openembedded.org
<dmitry.baryshkov=oss.qualcomm.com@lists.openembedded.org> wrote:
> > ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
> > ...
> > | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
> > | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
> > | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found
>
>
> I've fixed all other issues, except the x32 (working on reproducing it
> now) and this one. Would you have a pointer to the build job?

https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264/steps/14/logs/stdio

This is in context of oe-selftest, but it is simply building
qemu-system-native on debian 12 - I suspect the mesa-native fail is
specific to that host distro. I can try to reproduce if you need help
with it.

Alex

Dmitry Baryshkov April 2, 2025, 1:45 a.m. UTC | #7

On 01/04/2025 19:28, Alexander Kanavin wrote:
> On Tue, 1 Apr 2025 at 18:19, Dmitry Baryshkov via
> lists.openembedded.org
> <dmitry.baryshkov=oss.qualcomm.com@lists.openembedded.org> wrote:
>>> ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
>>> ...
>>> | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
>>> | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
>>> | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found
>>
>>
>> I've fixed all other issues, except the x32 (working on reproducing it
>> now) and this one. Would you have a pointer to the build job?
> 
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264/steps/14/logs/stdio
> 
> This is in context of oe-selftest, but it is simply building
> qemu-system-native on debian 12 - I suspect the mesa-native fail is
> specific to that host distro. I can try to reproduce if you need help
> with it.

It's really strange as my build host is Debian 12. And it's even more 
strange as mesa-clc was built few lines above (16406).

-- 
With best wishes
Dmitry

Mathieu Dubois-Briand April 2, 2025, 6:55 a.m. UTC | #8

On Wed Apr 2, 2025 at 3:45 AM CEST, Dmitry Baryshkov wrote:
> On 01/04/2025 19:28, Alexander Kanavin wrote:
>> On Tue, 1 Apr 2025 at 18:19, Dmitry Baryshkov via
>> lists.openembedded.org
>> <dmitry.baryshkov=oss.qualcomm.com@lists.openembedded.org> wrote:
>>>> ERROR: mesa-native-2_25.0.2-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/temp/run.do_compile.506477' failed with exit code 1
>>>> ...
>>>> | FAILED: src/intel/shaders/intel_gfx80_shaders.spv
>>>> | /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc ../mesa-25.0.2/src/intel/shaders/libintel_shaders.h ../mesa-25.0.2/src/intel/shaders/generate.cl ../mesa-25.0.2/src/intel/shaders/generate_draws.cl ../mesa-25.0.2/src/intel/shaders/generate_draws_iris.cl ../mesa-25.0.2/src/intel/shaders/query_copy.cl ../mesa-25.0.2/src/intel/shaders/util.cl -o src/intel/shaders/intel_gfx80_shaders.spv -- -DNDEBUG=1 -DGFX_VERx10=80 -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/shaders/. -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/compiler/libcl -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/intel -I/srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/mesa-25.0.2/src/intel/genxml
>>>> | /bin/sh: 1: /srv/pokybuild/yocto-worker/oe-selftest-debian/build/build-st-53842/tmp/work/x86_64-linux/mesa-native/25.0.2/build/src/compiler/clc/mesa_clc: not found
>>>
>>>
>>> I've fixed all other issues, except the x32 (working on reproducing it
>>> now) and this one. Would you have a pointer to the build job?
>> 
>> https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1264/steps/14/logs/stdio
>> 
>> This is in context of oe-selftest, but it is simply building
>> qemu-system-native on debian 12 - I suspect the mesa-native fail is
>> specific to that host distro. I can try to reproduce if you need help
>> with it.
>
> It's really strange as my build host is Debian 12. And it's even more 
> strange as mesa-clc was built few lines above (16406).

Ok, I will try to play a bit with it today, let's see if I can reproduce
easily.

I launched a rebuild of the previous job there:
https://autobuilder.yoctoproject.org/valkyrie/#/builders/35/builds/1274
I will try to reproduce it locally a bit later today.

Also, maybe this is linked with other fails. So if you have a new
version of your series fixing them, feel free to send it, I will be
happy to test it.

Alexander Kanavin April 2, 2025, 7:56 a.m. UTC | #9

On Wed, 2 Apr 2025 at 08:55, Mathieu Dubois-Briand
<mathieu.dubois-briand@bootlin.com> wrote:
 >
> > It's really strange as my build host is Debian 12. And it's even more
> > strange as mesa-clc was built few lines above (16406).
>
> Ok, I will try to play a bit with it today, let's see if I can reproduce
> easily.

I took v4, and ran 'bitbake mesa-native' on the exact build host
(debian-12-vk-1) where it happened, and it built without issues.
mesa_clc is built and used in build logs.

If this happens again, we should take care to preserve the failing
build directory, so it can be inspected post-mortem. 'Not found' error
can be misleading: it may mean there's a problem with one of the
libraries, or the executable is somehow in an incorrect format, etc.

Alex

Dmitry Baryshkov April 2, 2025, 12:50 p.m. UTC | #10

On 02/04/2025 10:56, Alexander Kanavin wrote:
> On Wed, 2 Apr 2025 at 08:55, Mathieu Dubois-Briand
> <mathieu.dubois-briand@bootlin.com> wrote:
>   >
>>> It's really strange as my build host is Debian 12. And it's even more
>>> strange as mesa-clc was built few lines above (16406).
>>
>> Ok, I will try to play a bit with it today, let's see if I can reproduce
>> easily.
> 
> I took v4, and ran 'bitbake mesa-native' on the exact build host
> (debian-12-vk-1) where it happened, and it built without issues.
> mesa_clc is built and used in build logs.
> 
> If this happens again, we should take care to preserve the failing
> build directory, so it can be inspected post-mortem. 'Not found' error
> can be misleading: it may mean there's a problem with one of the
> libraries, or the executable is somehow in an incorrect format, etc.

Okay, I was able to reproduce the issue. It happens when the llvm-native 
was built with a different build-dir as the current build, because then 
the other build-dir (which might not exist ATM) crawls into llvm-config 
--ldflags.

For example, I caught this while doing builds in build-test, while the 
sstate-cache contained llvm-native build from the build-test-st (from 
selftests builds).

[v4,0/9] mesa: upgrade to 25.0.2

Message

Comments