[windows] Nightly builds need more coverage

[weliveindetail]

The infrastructure for running a stable subset of LLDB tests on Windows is in place on mainline. We enabled it for PR-testing with #78844, but https://ci-external.swift.org/job/swift-main-windows-toolchain/ uses a custom (hardcoded) invocation that I cannot access.

Bringing LLDB testing to swift-main-windows-toolchain seems important to monitor flakes. We filtered loads of flaky test in the bringup phase, but future changes may introduce new ones and it's critical to find them early. Otherwise they might clog the PR-testing bots.

Right now, it invokes swift\utils\build.ps1 directly and passes -Test swift,dispatch,xctest. We could either extend it to -Test swift,dispatch,xctest,lldb or drop the parameter entirely so that build.ps1 selects the default.

[weliveindetail]

@shahmishal What do you think?

[weliveindetail]

While we are here, we should also add -Summary to monitor build and tests times.

[weliveindetail]

New test flake that popped up randomly in one of my own PRs: #79154 We should get better monitoring here.

[weliveindetail]

@shahmishal Gentle ping. Is there anything I can do to help with this task?

[weliveindetail]

@shahmishal Any chance to move on here this week?

It's hard to track flakes in PR-testing. And it's hard to tell how much it slows down the review process. Updating the mainline bots would help a lot.

[weliveindetail]

@shahmishal ping

[weliveindetail]

@shahmishal ping

[weliveindetail]

@shahmishal ping

We also see more and more issues when building the Android SDKs. Might be useful to enable them (or at least one) in the nightly. @compnerd Can you confirm?

Summary: we need -Test swift,dispatch,xctest,lldb, -Summary and -Android (or at least -AndroidSDKs aarch64)

[shahmishal]

lldb added to the testing, -Summary and -Android flag now configured for Windows toolchain bots.

[shahmishal]

https://ci-external.swift.org/job/swift-main-windows-toolchain/1141

[weliveindetail]

Thank you very much!