Revert "Use file URL for source map paths (#1771)"

[cspotcode]

Revert #1771

Fixes #1797
Fixes #1790

@PaperStrike if we choose to implement #1769 in the future, it'll need a bit more work to:

a) prove that we are fixing a real-world use-case, not merely an academic desire to comply with a spec
b) prove that debugging scenarios continue to work. Will likely require automated testing of:

• VSCode debugging
• code coverage via nyc
• code coverage via c8 which, in turn, uses node's native coverage mechanism

[codecov]

Codecov Report

Merging #1821 (861f3e9) into main (bf13086) will decrease coverage by 0.01%.
The diff coverage is 100.00%.

Impacted Files	Coverage Δ
src/index.ts	79.95% <100.00%> (-0.09%)	⬇️

[PaperStrike]

I have no time for issuing an academic desire that has no real world use case. I forgot to mention that it's found when using c8 with --exclude-after-remap on Windows. But recently I am really really busy, I am not (not until about July 9) going to prepare a repro or a test case or things other than commenting, sorry about that.

[cspotcode]

Time constraints are understandable; I've been very busy myself, so I understand where you're coming from.

I think there may be some confusion about my comments.

This change broke nyc coverage reports and debugging scenarios. Given the evidence available to me, it looks like this change did more harm than good. So reverting is the most appropriate action. We can always discuss further in the future if/when you or anyone else has the time to investigate.

[PaperStrike]

I totally understand that. Maybe we can try the basename solution in the future.

I remember the problem was like, c8 totally relies on test-exclude when --exclude-after-remap is enabled, so it met bcoe/c8#183 and istanbuljs/test-exclude#43 when combining with ts-node and Windows. The two issues aren't resolved for some reason, so I filled #1769 here.

[cspotcode]

Yeah I had no luck getting in touch with anyone from nyc. I don't think nyc has any meaningful maintainership. Which is unfortunate, nyc clearly needs a patch to support file URIs.

The debugger breakpoints issue, I haven't had time to investigate. I created a ticket #1822 for setting up some sort of automated testing for these things. I have no idea how I'll do it, but I think it will be important. And then perhaps those tests will help us validate the basename solution.

I want to thank you again for your contribution, even though we had to revert it.