[PATCH] D149996: [test] [ExecutionEngine] Skip the ExecutionEngine tests on mingw targets

[Martin Storsjö via Phabricator via llvm-commits]

mstorsjo added a comment.

In D149996#4331020 <https://reviews.llvm.org/D149996#4331020>, @lhames wrote:

>> The IR doesn't have any triples in it - I'm running into this with e.g. https://github.com/llvm/llvm-project/blob/main/llvm/test/ExecutionEngine/MCJIT/test-loadstore.ll. When running without `-jit-kind=mcjit`, `lli` seems to ignore the triple specified with the `-mtriple` option, and it also doesn't react to changes to `sys::getDefaultTargetTriple()`. So where is the triple set that is used for code generation, when running that testcase with Orc? (I'm curious about where it picks up the msvc vs mingw difference,  if it's not from either of the `-mtriple` option or `sys:: getDefaultTargetTriple()`.)
>
> Oh -- It's using `sys::getProcessTriple()` rather than `sys::getDefaultTargetTriple()`.

Ah, I see - thanks, that explains it.

> I think the failure to honor `-mtriple` should be treated as a bug here -- I'll add a patch to fix this.

Maybe, but do note that there's a subtle existing tweak for MCJIT, it uses triples like `x86_64-pc-windows-msvc-elf`, set up here: https://github.com/llvm/llvm-project/blob/main/llvm/test/lit.cfg.py#L96-L102 If we start honoring that when not using `-jit-kind=mcjit`, the tests that run with Orc end up doing and testing a totally different thing, I'm afraid...


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149996/new/

https://reviews.llvm.org/D149996