[lldb] Mark parsing Swift expressions with generics as not cacheable

[augusto2112]

Because a Swift expression may have different generic instantiations in different invocations, we cannot cache the parsing of them.

[bnbarham]

There were only a couple LLVM PRs when macOS PR testing started failing in https://ci.swift.org/view/Swift%20rebranch/job/oss-swift-rebranch-pr-test-macoss/2324/ with:

  lldb-api :: lang/swift/playgrounds-repl/old_playground/TestNonREPLPlayground.py
  lldb-api :: lang/swift/playgrounds/TestPlaygrounds.py

#7461 ran PR testing and those two tests succeeded, so I suspect this PR is probably the cause? From a very quick look, it seems like code_manipulator may not be set for playgrounds (though I'm not sure if the TestPlaygrounds.py case has repl set or not)?

  std::unique_ptr<SwiftASTManipulator> code_manipulator;
  if (repl || !playground) {
    code_manipulator = std::make_unique<SwiftASTManipulator>(
        *source_file, repl, options.GetBindGenericTypes());

    if (!playground) {
      code_manipulator->RewriteResult();
    }
  }

[augusto2112]

There were only a couple LLVM PRs when macOS PR testing started failing in https://ci.swift.org/view/Swift%20rebranch/job/oss-swift-rebranch-pr-test-macoss/2324/ with:

  lldb-api :: lang/swift/playgrounds-repl/old_playground/TestNonREPLPlayground.py
  lldb-api :: lang/swift/playgrounds/TestPlaygrounds.py

#7461 ran PR testing and those two tests succeeded, so I suspect this PR is probably the cause? From a very quick look, it seems like code_manipulator may not be set for playgrounds (though I'm not sure if the TestPlaygrounds.py case has repl set or not)?

  std::unique_ptr<SwiftASTManipulator> code_manipulator;
  if (repl || !playground) {
    code_manipulator = std::make_unique<SwiftASTManipulator>(
        *source_file, repl, options.GetBindGenericTypes());

    if (!playground) {
      code_manipulator->RewriteResult();
    }
  }

Yeah that was the problem. I merged #7494 around half an hour ago, which should solve the issue.

[bnbarham]

Ah I missed that, thanks! FWIW PR testing should be in a fairly reasonable state for macOS and Linux at least (minus a debug info crash on Linux, but easy to check the current crashes). So would be good to run PR testing 🙏

[augusto2112]

Ah I missed that, thanks! FWIW PR testing should be in a fairly reasonable state for macOS and Linux at least (minus a debug info crash on Linux, but easy to check the current crashes). So would be good to run PR testing 🙏

Will do! Sorry for the mess!