Attempt to unbreak the build by building SwiftDriver module with -enable-testing in CMake.

[artemcm]

Pass -enable-testing to SwiftDriver CMake build to allow the tests to be built on Linux in the presence of @testable.
Linux build bots are otherwise failing with:

/home/buildnode/jenkins/workspace/swift-PR-Linux@2/branch-main/swift-driver/Tests/SwiftDriverTests/IncrementalCompilationTests.swift:15:18: error: module 'SwiftDriver' was not compiled for testing

Tests relying on SwiftDriver were disabled in #364. This PR restores these tests.

[cltnschlosser]

Do we build in release configuration? (https://bugs.swift.org/browse/SR-1354)

I ran into this just now too when running the integration tests with spm (This fixed there, diff is broken to fix markdown):

diff --git a/README.md b/README.md
index b17beb4..442acaf 100644
--- a/README.md
+++ b/README.md
@@ -88,7 +88,7 @@ on the command line:

 $ SWIFT_DRIVER_ENABLE_INTEGRATION_TESTS=1 \
   SWIFT_DRIVER_LIT_DIR=/path/to/build/Ninja-ReleaseAssert/swift-.../test-... \
-  swift test -c release --parallel
+  swift test -c release -Xswiftc -enable-testing --parallel

 
 #### Testing against `swift` compiler trunk

[artemcm]

I think in this case the problem is a little bit different, here the CMake build (not part of the test stage, but rather part of the build stage in build-script), on Linux, is building files that belong to the test targets. And that breaks if the imported module has not been built with -enable-testing

[cltnschlosser]

Ah, does it need to build the tests?

So with this change we build with testability when doing things like building a toolchain? If yes, do we have a long term plan for getting rid of @testable?

[artemcm]

Yeah that's a good question, and I definitely don't think this is anything workable long-term. This has been a bit of a perfect storm where our use of SPI caused build breaks with newest development compilers, and it was easier to remove that than debug the compiler, to unblock the build. Then we replaced it with Testable which had the effect this PR is trying to fix.

Long term, we should absolutely move away from Testable and do this properly, probably with library evolution and spi combination. This change is a very temporary workaround to get folks' PR tests working.

[DougGregor]

@swift-ci please test

[DougGregor]

Ah, macOS PR testing seems to have failed with this:

/Users/buildnode/jenkins/workspace/swift-PR-osx-smoke-test/branch-main/swift-driver/Tests/SwiftDriverTests/IncrementalCompilationTests.swift:15:18: error: module 'SwiftDriver' was not compiled for testing
21:39:25 @testable import SwiftDriver
21:39:25 ~~~~~~~~~~       ^
21:39:25 error: fatalError
21:39:25 xcrun --sdk macosx --show-sdk-path

[DougGregor]

Folks, we can't have PR testing broken. I'm going to land #364 to disable the code with @testable import in it and we can sort this out on Monday.

[DougGregor]

Make sure to revert #364 before tackling this.

[artemcm]

@swift-ci please test

[artemcm]

@swift-ci please test