Jake Bailey
Jake Bailey
I'm happy to say that [CL 600296](https://go.dev/cl/600296) does fix the performance issue I was observing. (It appears to be _slightly_ slower than a straight revert of [CL 584597](https://go.dev/cl/584597), but there...
This is an intentional divergence. This sort of global analysis is problematic because it means that the emitted code for one file requires analyzing all other files (and therefore, no...
It's definitely supposed to report an error.
We were actually considering adding a prepack script to TypeScript so that people could install via git for bisect purposes, but found out that it doesn't actually work in npm...
If you hadn't seen it, there's also: https://github.com/cspotcode/workaround-broken-npm-prepack-behavior Which is what `ts-node` uses to work around this problem.
FWIW I do not think the above linked PR is correct. The PR here seems fine-ish, though the layout of the upstream package is truly, truly strange.
Can you add a test for your case?
> (it lives in the formatter 😬) This scared me enough to go look, but it seemingly doesn't actually do anything formatting related; I'm not sure why it lives there...
I would be curious if TS 5.7's project loading changes would make this more well-behaved; might be worth a try. There have been improvements to project loading which should result...
I wouldn't think so, given you have checkJs on and my PR would only fix errors when it's off.