[clang] A more precise matching for the driver test (PR #129611)

Fangrui Song via cfe-commits cfe-commits at lists.llvm.org
Mon Mar 17 19:59:10 PDT 2025


MaskRay wrote:

> #128894 seems to have caused a problem on the ppc64le bot. I _think_ this fixes it, but this test seems a bit janky to me. I don't really see why the driver's sub-commands would ever be out of order, and the imprecise matches seem flaky.
> 
> Not sure if @MaskRay or @yxsamliu have any insights.

The issue was likely that the build directory name
`/home/buildbots/llvm-external-buildbots/workers/ppc64le-lld-multistage-test/ppc64le-lld-multistage-test` contains `lld`.

We usually don't test specific strings like `clang-[[#]]`. I recall that this broke Google's internal lit tester, where clang becomes a hashed string in a content-addressable storage, unless  `-no-canonical-prefixes` is specified.

 Instead, we just check `-cc1` (which is a good anchor for the clang -cc1 line)

https://github.com/llvm/llvm-project/pull/129611


More information about the cfe-commits mailing list