[clang-tidy] encounteres "error: unknown target CPU 'cortex-a76.cortex-a55' [clang-diagnostic-error]"

[nick-potenski]

While trying to scan files in a cross-compiled project with clang-tidy version 20, all of my source files encounter the following error:

Error while processing <Source file name redacted>.
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]
note: valid target CPU values are: a64fx, ampere1, ampere1a, ampere1b, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-a15, apple-a16, apple-a17, apple-a7, apple-a8, apple-a9, apple-m1, apple-m2, apple-m3, apple-m4, apple-s4, apple-s5, carmel, cobalt-100, cortex-a34, cortex-a35, cortex-a510, cortex-a520, cortex-a520ae, cortex-a53, cortex-a55, cortex-a57, cortex-a65, cortex-a65ae, cortex-a710, cortex-a715, cortex-a72, cortex-a720, cortex-a720ae, cortex-a725, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78ae, cortex-a78c, cortex-r82, cortex-r82ae, cortex-x1, cortex-x1c, cortex-x2, cortex-x3, cortex-x4, cortex-x925, cyclone, exynos-m3, exynos-m4, exynos-m5, falkor, fujitsu-monaka, generic, grace, kryo, neoverse-512tvb, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-n3, neoverse-v1, neoverse-v2, neoverse-v3, neoverse-v3ae, oryon-1, saphira, thunderx, thunderx2t99, thunderx3t110, thunderxt81, thunderxt83, thunderxt88, tsv110
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]

The project is compiled with gcc version 11.5.0 and the cortex-a76.cortex-a55 seems to be a valid cpu option for it.

[llvmbot]

@llvm/issue-subscribers-clang-tidy

Author: Nick Potenski (nick-potenski)

While trying to scan files in a cross-compiled project with clang-tidy version 20, all of my source files encounter the following error:

Error while processing <Source file name redacted>.
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]
note: valid target CPU values are: a64fx, ampere1, ampere1a, ampere1b, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-a15, apple-a16, apple-a17, apple-a7, apple-a8, apple-a9, apple-m1, apple-m2, apple-m3, apple-m4, apple-s4, apple-s5, carmel, cobalt-100, cortex-a34, cortex-a35, cortex-a510, cortex-a520, cortex-a520ae, cortex-a53, cortex-a55, cortex-a57, cortex-a65, cortex-a65ae, cortex-a710, cortex-a715, cortex-a72, cortex-a720, cortex-a720ae, cortex-a725, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78ae, cortex-a78c, cortex-r82, cortex-r82ae, cortex-x1, cortex-x1c, cortex-x2, cortex-x3, cortex-x4, cortex-x925, cyclone, exynos-m3, exynos-m4, exynos-m5, falkor, fujitsu-monaka, generic, grace, kryo, neoverse-512tvb, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-n3, neoverse-v1, neoverse-v2, neoverse-v3, neoverse-v3ae, oryon-1, saphira, thunderx, thunderx2t99, thunderx3t110, thunderxt81, thunderxt83, thunderxt88, tsv110
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]

The project is compiled with gcc version 11.5.0 and the cortex-a76.cortex-a55 seems to be a valid cpu option for it.

[llvmbot]

@llvm/issue-subscribers-backend-aarch64

Author: Nick Potenski (nick-potenski)

While trying to scan files in a cross-compiled project with clang-tidy version 20, all of my source files encounter the following error:

Error while processing <Source file name redacted>.
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]
note: valid target CPU values are: a64fx, ampere1, ampere1a, ampere1b, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-a15, apple-a16, apple-a17, apple-a7, apple-a8, apple-a9, apple-m1, apple-m2, apple-m3, apple-m4, apple-s4, apple-s5, carmel, cobalt-100, cortex-a34, cortex-a35, cortex-a510, cortex-a520, cortex-a520ae, cortex-a53, cortex-a55, cortex-a57, cortex-a65, cortex-a65ae, cortex-a710, cortex-a715, cortex-a72, cortex-a720, cortex-a720ae, cortex-a725, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78ae, cortex-a78c, cortex-r82, cortex-r82ae, cortex-x1, cortex-x1c, cortex-x2, cortex-x3, cortex-x4, cortex-x925, cyclone, exynos-m3, exynos-m4, exynos-m5, falkor, fujitsu-monaka, generic, grace, kryo, neoverse-512tvb, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-n3, neoverse-v1, neoverse-v2, neoverse-v3, neoverse-v3ae, oryon-1, saphira, thunderx, thunderx2t99, thunderx3t110, thunderxt81, thunderxt83, thunderxt88, tsv110
error: unsupported argument 'cortex-a76.cortex-a55' to option '-mcpu=' [clang-diagnostic-error]

The project is compiled with gcc version 11.5.0 and the cortex-a76.cortex-a55 seems to be a valid cpu option for it.

[vbvictor]

Hi, the error you get (clang-diagnostic-error) is from clang compiler itself. clang-tidy uses clang fronted to parse code with appropriate flags and build AST for analysis. So to use clang-tidy you need to be able to compile your code with clang, at least it should be able to parse C++ and produce AST.

[vbvictor]

It seems that clang doesn't know this target and to fix this you could configure your project without specified target at all because it doesn't matter for analysis AFAIK. Or you can manually delete it from compile_commands.json with e.g. sed .

[davemgreen]

We have never added those fake "cpus" that model a big+little, as very few people have use for them (and later generations became much more complex). If you need to compile with clang you can pick one of -mcpu=cortex-a55 or -mcpu=cortex-a76 and it should be fine.

[vbvictor]

you could configure your project without specified target at all because it doesn't matter for analysis AFAIK

That was a bit of rush, there is 1% of rules that has different behavior with -target, like bugprone-narrowing-conversions, but apart from them clang-tidy checks must be platform independent.

[nick-potenski]

Thanks, yes, I can certainly work around this by using a sed-like approach to modify the compile command database. That is, however, a bit clunky and annoying to carry around. It would be nice if we could use clang tooling without that sort of hack in place.

[carlosgalvezp]

you could configure your project without specified target at all

It's also possible that some ARM-specific code is not understood by clang-tidy/clang without the -mcpu flag, leading to compiler errors or false positives. So choosing the "closest" CPU would be best.

[vbvictor]

It would be nice if we could use clang tooling without that sort of hack in place.

IMO, we should not provide a way of hiding compiler errors because they lead to broken AST. And when clang-tidy runs on broken AST, it may produce random FPs and users would create issues "clang-tidy produces random warnings on valid code".
To avoid such problems we suggest users to make their code compilable in a way they want (not how clang tooling decided)

[carlosgalvezp]

Agreed. Should we document this somehow? "A prerequisite for running clang-tidy is that the code must compile without errors on clang"

[vbvictor]

Agreed. Should we document this somehow? "A prerequisite for running clang-tidy is that the code must compile without errors on clang"

Yeah, I think we should state that in docs. Maybe also mention that for best compatibility, clang-tidy version should be the same as clang-compiler (but newer versions of clang-tidy may also work otherwise user gets compiler errors)

[nick-potenski]

IMO, we should not provide a way of hiding compiler errors because they lead to broken AST. And when clang-tidy runs on broken AST, it may produce random FPs and users would create issues "clang-tidy produces random warnings on valid code". To avoid such problems we suggest users to make their code compilable in a way they want (not how clang tooling decided)

To be clear, I'm not suggesting the compiler error should be hidden. I'm only suggesting supporting feature parity between compilers.