[Windows] Stop using multiroot data file

[ahoppen]

I have seen testing of swift-foundation fail without an error message (eg. https://ci-external.swift.org/job/sourcekit-lsp-PR-windows/3152/, build log attached to this PR), which likely started occurring after #80122. Revert that change to unbreak CI.

sourcekit-lsp-PR-windows-3152.txt

[ahoppen]

@swift-ci Please smoke test

[compnerd]

I am unsure that this is the true cause of that silent crash. This is something that @bnbarham and I have been trying to isolate (it is CI only). I wonder if we should do a more judicious revert of just the parameter. This is unfortunate, but if this is causing it to reproduce more often, I am curious if this might help us narrow down what is going on.

[bnbarham]

Looks like it somehow happens for swift-foundation now, rather than just the swift-corelibs-foundation that we've been seeing 🤔 I very much doubt this is the actual cause, but I have no explanation of how it suddenly causes it in swift-foundation. Interestingly not all of them have icu_packaged_data.cpp at the end like we've seen previously. Eg. https://ci-external.swift.org/view/Pull%20Request/job/swift-PR-windows/44259/console doesn't have it.

And merging to unblock CI, since I do see a bunch of failed builds since that PR was merged :(.

[ahoppen]

That’s interesting. Maybe the difference is whether we test in debug or release configuration (swift-foundation currently tests in release while swift-corelibs-foundation tests in debug due to an incorrect merge conflict resolution that I described in #80122 (comment).

Trying to line the configurations up in #83720, maybe that can give some insights into the root cause.

[bnbarham]

Oh I only saw this after I commented on the other PR. Yeah, that is the current assumption (which your PR looks to have confirmed) 👍