[Lldb-commits] [lldb] [lldb] Remove default_byte_order from CoreDefinition (PR #190494)

Sergei Barannikov via lldb-commits lldb-commits at lists.llvm.org
Sun Apr 5 00:49:07 PDT 2026


s-barannikov wrote:

> The tests fail because LLDB and LLVM disagree on whether `armv6m` is `arm` or `thumb`. See [fadc210](https://github.com/llvm/llvm-project/commit/fadc210817bb197c249758880814abf57dda9937) that changed LLVM to classify it as `thumb`, and [a3a0452](https://github.com/llvm/llvm-project/commit/a3a045283133bb2733b90ad0826ed13780659334) that added `armv6m` support to LLDB (more than a year and a half earlier). I guess this should be fixed separately.

I investigated this further, and it appears it is not the only problem. Here is a complete list of Cores, which when round-tripped through llvm::Triple return ArchType different from what Core specifies:
```
armv8  aarch64 -> arm
armv8a aarch64 -> arm
mipsr2 The rest are UnknownArch
mipsr3
mipsr5
mipsr2el
mipsr3el
mipsr5el
mips64r2
mips64r3
mips64r5
mips64r2el
mips64r3el
mips64r5el
ppc601
ppc602
ppc603
ppc603e
ppc603ev
ppc604
ppc604e
ppc620
ppc750
ppc7400
ppc7450
ppc970
ppc970-64
i486sx
hexagonv4
hexagonv5
```
This means that `arch_spec.GetMachine()` and `arch_spec.GetTriple.getArch()` return different values for them and thus these methods are not interchangeable. Is that expected? I'm confused.


https://github.com/llvm/llvm-project/pull/190494


More information about the lldb-commits mailing list