Revert "Clean up the code and some of the API of compiling plugins, and improve persistence of cached compiler outputs"

[compnerd]

Reverts #4290

[compnerd]

@swift-ci please smoke test

[compnerd]

CC: @abertelrud @tomerd - the change regressed on Windows, which has been broken multiple times in the past week. I'd like to revert rather than fix forward, and then retry with testing.

[tomerd]

@abertelrud @compnerd can we fix this forward easily? if not, we should revert

[compnerd]

I can't test it locally atm, can you test and merge this quickly? I think it is better to revert and then fix.

[abertelrud]

@compnerd What's the failure? I understand wanting to revert, but will need to fix it.

[compnerd]

@abertelrud https://ci-external.swift.org/job/oss-swift-windows-toolchain-x86_64-vs2019/509/console

[tomerd]

I C:\Users\swift-ci\jenkins\workspace\oss-swift-windows-toolchain-x86_64-vs2019\swift-corelibs-libdispatch\src\BlocksRuntime -I T:\13\swift @CMakeFiles\Workspace.rsp    -Xlinker -implib:lib\Workspace.lib  && cd ."
C:\Users\swift-ci\jenkins\workspace\oss-swift-windows-toolchain-x86_64-vs2019\swiftpm\Sources\Workspace\DefaultPluginScriptRunner.swift:250:27: error: type 'ProcessResult.ExitStatus' has no member 'signalled'

                    case .signalled(let signal):

                         ~^~~~~~~~~

[abertelrud]

@abertelrud https://ci-external.swift.org/job/oss-swift-windows-toolchain-x86_64-vs2019/509/console

Thanks for the reference. Let's revert for now and I'll put up a fixed PR. Is there any way to run the Windows build against CI so I can tell if I'm breaking it again?

[abertelrud]

Or probably better I can just #if out that part. There are other cases of this in the code, so I don't think that will hold things back.

[abertelrud]

Put up #4299. Can we test that on Windows somehow? Would hate to go back and forth on this.

[compnerd]

It is possible to do cross-repository testing - that is the only option atm

[compnerd]

More specifically you need to build a full toolchain for Windows ("@swift-ci please build toolchain Windows platform") with the associated PR

[abertelrud]

Great, I'll give that a try. Thanks @compnerd

[abertelrud]

Great, I'll give that a try. Thanks @compnerd

Oh but I'd have to do that from within a PR in the Swift compiler I guess.

[abertelrud]

In this case I'd prefer to merge the fix in order to make progress. If needed we can revert this and then I'll put up a PR to put it back with the fix, but there's still the question of testing that. I never was able to get SwiftPM on a local Windows VM to work.

[compnerd]

Right, until we get the CI setup on this repo, it would need to be done on that repo (and why it requires a full toolchain build)

[abertelrud]

Meanwhile, I'm preparing a cross-repo test run with a DNM branch in the Swift repo to test #4299. If we must revert quickly we can do that, then I would convert #4299 to putting the code back with the fix.

[abertelrud]

swiftlang/swift#42405