While at it, extend test for FreeBSD and check for -lrt iff on Linux.
Details
Details
Diff Detail
Diff Detail
Event Timeline
lib/Driver/Tools.cpp | ||
---|---|---|
8683 ↗ | (On Diff #88871) | addOpenMPRuntime would add -lomptarget even for this call if JA satisfies the newly added check. That's a deviation from exisiting behavior. Am I missing some invariant here where JobActions along this path don't have an offloading action? |
lib/Driver/Tools.cpp | ||
---|---|---|
8683 ↗ | (On Diff #88871) | Right, as I noted in the summary: This enables libomptarget for Darwin, FreeBSD and NetBSD I don't see a problem with that, it's rather an inconsistency in the current code. |
lib/Driver/Tools.cpp | ||
---|---|---|
10334 ↗ | (On Diff #89110) | Now that you've moved the comment, it is not clear what "this" means here. Please reword this to say that you should only pass true for GompNeedsRT on platforms that really need it. |
lib/Driver/ToolChains/CommonArgs.cpp | ||
---|---|---|
434 | Good idea, this even allows this change to become fully NFC |
Do you really need to pass a reference to JobAction here or it is enough to pass a bool value for JA.isHostOffloading()?