Mangler enabled in hyenie/CI part of pipeline?

[connor4312]

I noticed recently that PR's takle longer to merge because the mangler is enabled in the compile/hygenie step, e.g. https://github.com/microsoft/vscode/actions/runs/19175726948/job/54819910297?pr=276137

I'm guessing the intention is to catch failures that may happen due to private access in mangling. However, this change won't actually catch those -- those failures happen when running tests in code that has already been mangled. That doesn't happen in this CI check, as-is the CI is only checking that the mangler itself doesn't error out, which I don't think has ever been the issue.

We'd need to actually wait and run tests on mangled code, but would make PR merging even slower...

[lszomoru]

Great callout. The intention is to close the gap between what is being run as part of the PR build vs. what is being run in the CI/product build. Anything that we can do to narrow the gap saves people precious time when the CI/product build turns red.