zturner added a comment. > Clang called from the builder shows errors on types like char16_t and > char32_t. That's why enums-layout.test and typedefs.test are made to use MSVC > instead;
For me enums-layout.test crashes, but I see the same behavior in typedefs.test. This is because lld-link does not emit `S_UDT` records (note the original test was using %msvc_link, not %lld_link. That's why it passed before). I could probably add a new magic toolchain called clang-cl-with-msvc-linker but I think it's also fine to just keep these using cl.exe. > The builder now doesn't support /Gy and /order options, so the old scheme is > used in function-level-linking.test; Makes sense. Just curious, is the order file strictly necessary for this test? `/Gy` is the same as `-ffunction-sections`, so there could be an argument to be made for exposing `--function-sections` on the command line of `build.py`. On the other hand, there's no harm in falling back to the old mechanism here since this test isn't intended to be portable. > The builder now doesn't support compilation of multiple sources and linking > them together, so the old scheme is used for linking in func-symbols.test; This one is important to fix. > On Windows 32-bit LLDB can debug only 32-bit applications, and 64-bit LLDB > can debug only 64-bit applications. That's why the old scheme is left in > dynamic tests udt-layout.test, variables-locations.test and vbases.test. If > we will always compile the tests as 32-bit (64-bit), then they will fail on > the 64-bit (32-bit) platform. That's why we just assume that the user runs > the tests from the same command prompt where the build was run. Ahh, right. I wonder if we should have something like `--arch=lldb` that means "match the architecture of LLDB". It's not urgent for this patch, but it could be useful as a followup. Repository: rLLDB LLDB CHANGES SINCE LAST ACTION https://reviews.llvm.org/D54942/new/ https://reviews.llvm.org/D54942 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
