ggardet

One last one is with openCL enabled: ``` [ 33s] [ 8%] Building CXX object src/backends/cl/CMakeFiles/armnnClBackend.dir/ClContextDeserializer.cpp.o [ 33s] cd /home/abuild/rpmbuild/BUILD/armnn-21.05/build/src/backends/cl && /usr/bin/c++ -DARMCOMPUTECL_ENABLED -DARMNNREF_ENABLED -DARMNN_SERIALIZER -DARMNN_SERIALIZER_SCHEMA_PATH=\"/home/abuild/rpmbuild/BUILD/armnn-21.05/src/armnnSerializer/ArmnnSchema.fbs\" -DARMNN_TF_LITE_PARSER -DBOOST_ALL_DYN_LINK -DBOOST_ALL_NO_LIB -DDYNAMIC_BACKEND_BUILD_DIR=\"/home/abuild/rpmbuild/BUILD/armnn-21.05/build\"...

@FrancisMurtagh-arm Here is the changelog: ``` - Add '-Wno-error=stringop-overread' to workaround build error with GCC11 with openCL enabled - Add '-Wno-error=uninitialized -Wno-error=array-bounds' to workaround build issues with GCC11 - https://github.com/ARM-software/armnn/issues/548...

Current version is `21.08`, not `21.05` anymore. Would that be ok for you? EDIT: also, I have multiple -Wno-error* flags accumulated. If you want to see all warnings/errors, you should...

I added https://review.mlplatform.org/c/ml/armnn/+/6283 on top of `21.08` and removed all -Wno-error* flags. Now, the errors I get are only when openCL is enabled. * 1st error: ``` [ 158s] In...

Indeed, ` -Wno-error=deprecated-declarations` is not required as it is just a note. But the include path needs to be updated to be future proof.

> I pushed a patch for the pragma: https://review.mlplatform.org/c/ml/armnn/+/6516 You switch from `CL/CL2.hpp` to `CL/opencl.h`. I think it should be `CL/opencl.hpp`. > However would it be possible we are using...

Clearly, ACL should not package those header files and should depend on `opencl-headers` package instead. Latest packages for Arm can be found here: https://build.opensuse.org/package/binaries/openSUSE:Factory:ARM/ComputeLibrary/standard `*.h` vs `*.hpp` are `C` vs...

> you can use #!ArchExclusiveLine: to mark the next line to be architecture specific, eg: > > #!ArchExclusiveLine: x86_64 zypper -n install nbdkit-vddk-plugin > > is this solving your issue?...

@adrianschroeter It looks like it works for the scheduler, but not at build time since the line is executed on build. See: https://build.opensuse.org/package/live_build_log/home:Guillaume_G:branches:Virtualization/cdi-importer-container/containerfile/aarch64

How should we handle the arch properly? Defining `OBSARCH` in PrjConf seems to be hacky. What about having the proper `TARGETARCH` (and others?) defined by OBS?