Additionally, the explicit linux target has been removed from the test.
Details
Diff Detail
Event Timeline
clang/lib/Driver/Driver.cpp | ||
---|---|---|
1639 | The correct thing for the mingw linker interface would be to do the same thing as for GNU style linkers, i.e. using the unix parameter form. (With the lld mingw interface, the canonical way to pass options through to the underlying lld-link interface is e.g. -Xlink=/reproduce:. But as an unintended loophole, as long as you pass lld-link options with a leading slash instead of a dash, e.g. if you pass e.g. /reproduce: directly to the mingw interface, the mingw interface considers it a filename and passes it through as-is to lld-link, and lld-link the interprets it as an option. But that's not the kosher way to do it IMO.) Thus, here, I think the correct thing would be to check C.getDefaultToolChain().getTriple().isWindowsMSVCEnvironment(). |
clang/lib/Driver/Driver.cpp | ||
---|---|---|
1639 | Thanks |
Thanks! LG with comment.
clang/lib/Driver/Driver.cpp | ||
---|---|---|
1638–1642 | Twines shouldn't be on the stack. There's no reason for this to be a Twine either -- make it a StringRef (and then cast below if needed) |
The removal of the explicit target broke the test on Mac: http://45.33.8.238/macm1/42754/step_7.txt
(Just need to be more permissive about leading underscores)
Please take a look and revert for now if it takes a while to fix.
Thanks should be fixed by https://github.com/llvm/llvm-project/commit/c175d80be2ed496debb98d47f315aa5e60116768
@mstorsjo, will this do the right thing for mingw?