Fix llorg/sycl-web pulldown issue

[zahiraam]

Signed-off-by: Zahira Ammarguellat zahira.ammarguellat@intel.com

[smanna12]

Could you please add a test? Thanks

[smanna12]

Could you please add a comment here?
Example: https://github.com/intel/llvm/blob/4d42a5c2c5eb3d77f720cb298691e165e768079e/clang/include/clang/AST/DeclCXX.h#L110-#L111

[Fznamznon]

Shouldn't the fix be included into #5427 . @vmaksimo WDYT?

[zahiraam]

Shouldn't the fix be included into #5427 . @vmaksimo WDYT?

I think it should. But I am wondering if better yet, the changes to SemaDecl.cpp shouldn't be up-streamed? And only make the change here for accessor.hpp?

Shouldn't the fix be included into #5427 . @vmaksimo WDYT?

[Fznamznon]

I am wondering if better yet, the changes to SemaDecl.cpp shouldn't be up-streamed?

I'm sure that the check on NewFD->getIdentifier() should be upstreamed. About the check on SYCLIsDevice I'm not that sure, since the attribute doc doesn't say that it shouldn't be used only for device compilation. It is our headers for some reason put some __init methods under SYCL_DEVICE_ONLY macro.
Probably it makes sense to commit the fix together with pulldown as well to unblock pulldown.

[zahiraam]

I am wondering if better yet, the changes to SemaDecl.cpp shouldn't be up-streamed?

I'm sure that the check on NewFD->getIdentifier() should be upstreamed. About the check on SYCLIsDevice I'm not that sure, since the attribute doc doesn't say that it shouldn't be used only for device compilation. It is our headers for some reason put some __init methods under SYCL_DEVICE_ONLY macro. Probably it makes sense to commit the fix together with pulldown as well to unblock pulldown.

I will make the change for NewFD->getIdentifier() in llorg and ask you to review it. To minimize the next pulldown "headaches" :)

[zahiraam]

I am wondering if better yet, the changes to SemaDecl.cpp shouldn't be up-streamed?

I'm sure that the check on NewFD->getIdentifier() should be upstreamed. About the check on SYCLIsDevice I'm not that sure, since the attribute doc doesn't say that it shouldn't be used only for device compilation. It is our headers for some reason put some __init methods under SYCL_DEVICE_ONLY macro. Probably it makes sense to commit the fix together with pulldown as well to unblock pulldown.

I will make the change for NewFD->getIdentifier() in llorg and ask you to review it. To minimize the next pulldown "headaches" :)

And the rest of the changes can just be committed here. Thanks.

[vmaksimo]

I see it's the best solution - to commit the fix as a part of #5427 and to upstream the needed fix to llorg to minimize next week's pulldown problems. Thanks!