[BPF][Test] Disable codegen test on AIX

[jsji]

Summary

https://reviews.llvm.org/D101194 changed the default getMultiarchTriple in toolchain.
So -march=bpf on AIX will get triple of bpf-ibm-aix now,
this is unexpected and causing test failures.

BPF on AIX is not supported (yet), disable the codegen test on AIX in lit cfg.

[yonghong-song]

We have a lot of tests using -march=bpf[eleb], why only these two tests? Is this due to the fact llvm-objdump won't work any more? Also, we have quite some objdump-* tests,

$ ls objdump*.ll
objdump_atomics.ll    objdump_cond_op.ll  objdump_imm_hex.ll     objdump_nop.ll         objdump_trivial.ll
objdump_cond_op_2.ll  objdump_dis_all.ll  objdump_intrinsics.ll  objdump_static_var.ll  objdump_two_funcs.ll

They all have similar llc | llvm-objdump | FileCheck test commands? Do other objdump_* tests also need to change to triple?

[jsji]

In D101866#2737982, @yonghong-song wrote:

We have a lot of tests using -march=bpf[eleb], why only these two tests? Is this due to the fact llvm-objdump won't work any more? Also, we have quite some objdump-* tests,

Good question. Not due to llvm-objdump, it is due to the fact AIX does not support bpf (yet).
You should be able to see what is wrong on Linux with following triple as well.

$ llc -mtriple=bpfel-ibm-aix -filetype=obj -o - llvm-project/llvm/test/CodeGen/BPF/BTF/binary-format.ll
<unknown>:0: error: Undefined temporary symbol 

The failure are:

<unknown>:0: error: Undefined temporary symbol 
..bin/llvm-readelf: error: '<stdin>': The file was not recognized as a valid object file
FileCheck error: '<stdin>' is empty.
FileCheck command line:  .../bin/FileCheck -check-prefixes=CHECK,CHECK-EL .../lvm/test/CodeGen/BPF/BTF/binary-format.ll

[jsji]

But yes, I believe it is better to update all testcases to use -mtriple instead of -march, to avoid trying to test unsupported triple combination. Let me know if you want to to update the patch to include all of them.

[yonghong-song]

If I understand correctly, on AIX system, BPF is not supported, so if -march=bpf is specified, it will internally expand to -mtriple=bpfel-ibm-aix and the test will fail since it is not supported, if llc uses -mtriple=bpfel and then llvm understands it is not really related to AIX and it is fine, is that right?

I don't like to change all tests (-march ==> -mtriple) for unsupported platform. Other platforms e.g., macos/windows seems working fine, could you check how they did?
In the worst case, if what windows/macos did won't work for you, you could change BPF test directory lit.local.cfg file to mark config.unsupported = True in your platform, but not individual tests.

[jsji]

In D101866#2740729, @yonghong-song wrote:

If I understand correctly, on AIX system, BPF is not supported, so if -march=bpf is specified, it will internally expand to -mtriple=bpfel-ibm-aix and the test will fail since it is not supported, if llc uses -mtriple=bpfel and then llvm understands it is not really related to AIX and it is fine, is that right?

Yes, it is correct.

I don't like to change all tests (-march ==> -mtriple) for unsupported platform. Other platforms e.g., macos/windows seems working fine, could you check how they did?

Toolchain can choose whether they want to override the default handling of triple. AIX toolchain choose not to override and use the default one.

In the worst case, if what windows/macos did won't work for you, you could change BPF test directory lit.local.cfg file to mark config.unsupported = True in your platform, but not individual tests.

That is fine for us if you prefer.

[yonghong-song]

Toolchain can choose whether they want to override the default handling of triple. AIX toolchain choose not to override and use the default one.

Maybe it is better to do what other platform is doing? What is the advantage of not overriding the default triple for not supported platform?

[jsji]

In D101866#2740835, @yonghong-song wrote:

Toolchain can choose whether they want to override the default handling of triple. AIX toolchain choose not to override and use the default one.

Maybe it is better to do what other platform is doing? What is the advantage of not overriding the default triple for not supported platform?

Yes, we will evaluate and consider overriding.
But the new change is breaking our internal buildbot, we would like to fix the failures first to unblock build.

[jsji]

Disable test on AIX.

[yonghong-song]

This looks better. Do you have a plan to eventually remove this fix as it really looks incompatible with other platforms?

[jsji]

In D101866#2741033, @yonghong-song wrote:

This looks better. Do you have a plan to eventually remove this fix as it really looks incompatible with other platforms?

Yes, we will.

[yonghong-song]

Ok, sounds good then.

[jsji]

Thanks @yonghong-song

[qiucf]

I believe targets don't redirect OS in triple. Because most of the CodeGen check is asm check, they just 'happen to be right'. For example, on aarch64-darwin, llc -march=aarch64 -filetype=obj BPF/BTF/binary-format.ll is ok, llc -mtriple=riscv64-unknown-linux-gnu -filetype=obj BPF/BTF/binary-format.ll is also ok. However llc -march=riscv64 -filetype=obj BPF/BTF/binary-format.ll fails, and OS is inferred as 'darwin'.

llvm/test/CodeGen/BPF/lit.local.cfg

 if not 'BPF' in config.root.targets:
     config.unsupported = True
+if 'system-aix' in config.available_features:
+    config.unsupported = True