Wire up support for module wrapping on non-Darwin platforms #145
Conversation
owenv
requested review from
jakepetroules,
mirza-garibovic,
mhrawdon and
neonichu
as code owners
|
@swift-ci test |
|
fyi @cmcgee1024 since I think we discussed this at some point |
| switch self { | ||
| case .macho: | ||
| return false | ||
| default: |
There was a problem hiding this comment.
nit: can we prefer to list the cases explicitly? I like to stay away from default since it forces re-evaluation of use sites when new cases are added. Not that adding new formats is particularly likely, but I could at least see a new one for wasm being added in the future.
There was a problem hiding this comment.
This is really an is-not-macho check, I don't expect we will ever move away from module wrap for any future object formats
|
@swift-ci test |
| @@ -2127,6 +2127,11 @@ public final class SwiftCompilerSpec : CompilerSpec, SpecIdentifierType, SwiftDi | |||
| return (inputs, outputs) | |||
| }() | |||
|
|
|||
| if cbc.scope.evaluate(BuiltinMacros.PLATFORM_REQUIRES_SWIFT_MODULEWRAP) && cbc.scope.evaluate(BuiltinMacros.GCC_GENERATE_DEBUGGING_SYMBOLS) { | |||
| let moduleWrapOutput = Path(moduleFilePath.withoutSuffix + ".o") | |||
There was a problem hiding this comment.
Windows really should use .obj instead of .o.
There was a problem hiding this comment.
I filed #149 to track this because it needs to be done in a few other places at the same time to avoid a regression
There was a problem hiding this comment.
@compnerd I think it's more like "MSVC ABI should use .obj and gcc/mingw ABI should use .o", right?
No description provided.