[Lldb-commits] [PATCH] D69619: [lldb/lit] Introduce %clang_host substitutions

Adrian Prantl via Phabricator via lldb-commits lldb-commits at lists.llvm.org
Wed Oct 30 09:30:42 PDT 2019

aprantl added a comment.

this looks like a reasonable incremental improvement. That said, I'm wondering how useful `clang_host` is as a concept in the LIT tests. In my own mental model the tests are grouped into

1. end-to-end tests that test lldb against a variety of user-configurable compilers and with different debug info formats
2. unit-tests/internal consistency tests where we may use a compiler for the sake of not having to check in a binary, but where we expect a specific output from the compiler and thus should always use the just-built clang.

The `packages` subdirectory covers category (1) and the `lit` subdirectory covers category (2).

Based on that I would expect that we either always hard-code a triple in the lit tests (may result in less coverage if the target isn't available), or we run them for each available target. Giving the host target preference seems arbitrary, given how extensively LLDB is used on an x86_64 host, debugging AArch64 targets, for example.

  rG LLVM Github Monorepo



More information about the lldb-commits mailing list