fix: Don't report undefined symbols when writing executables with -z undefs#1741
Conversation
| @@ -1,18 +1,29 @@ | |||
| //#AbstractConfig:default | |||
| //#EnableLinker:lld | |||
There was a problem hiding this comment.
What was the reason for LLD here?
There was a problem hiding this comment.
Mostly to verify that LLD has the same behaviour as GNU ld. i.e that it errors and doesn't error in the same circumstances. I wonder if we should change linker-diff's default behaviour to match-all rather than match-any, then we could enable match-any just for tests that need it. It doesn't affect this test much, since diffing isn't so important for this test - all that really matters is whether we're failing or succeeding the link.
There was a problem hiding this comment.
Ah, right. Expect error makes it so all the linkers must fail similar way.
I don't have strong opinion on whether running all tests with additional linker is worthwhile. It might prove time-consuming, but on high core count machines it likely won't be very noticeable since c++/libc/rust integration tests dwarf all the other tests.
No description provided.