Anybody have any ideas how to get pyright to emit an error on the unused type: ignore here? None of my ideas seem to work (see each commit in isolation for what I've tried).

Weird, locally on main:
python .\tests\pyright_test.py .\stubs\invoke\@tests -p .\pyrightconfig.testcases.json works as expected, but
python .\tests\pyright_test.py -p .\pyrightconfig.testcases.json does not

Maybe it's something specific to do with how pyright-action invokes pyright

Maybe it's something specific to do with how pyright-action invokes pyright

It's the same if I call pyright directly: pyright .\stubs\invoke\@tests -p .\pyrightconfig.testcases.json

Ok, the ** path just isn't working
It works if I use:

    "include": [
        "stubs/invoke/@tests/test_cases",
    ],

Edit: bissected to be a regression in 1.1.284
Likely because of

Bug Fix (from pylance): Fixed exclude/include to work with '.' prefixed directories.

npx pyright@1.1.283 -p .\pyrightconfig.testcases.json works fine with wildcard paths

Edit 2: What I saw from 1.1.283 is unrelated.

Amazing, thanks for the investigation! I'll be AFK for a few hours, but I'll see about filing a bug report to pyright later if you don't beat me to it ;)

Nvm the bisect part, I celebrated too early and didn't realize the errors I was getting were from a different file. :/

But what I said about using the direct path working is still true. So if it's broken, it's been broken since forever. I've open an issue with a minimal reproduction: microsoft/pyright#4553

A fix has been released! https://github.com/microsoft/pyright/releases/tag/1.1.292
Not sure if you'd like to use this PR or a new one to start clean. But adding test_cases subfolders to the base and stricter configs' exclude is a good call.

A fix has been released! https://github.com/microsoft/pyright/releases/tag/1.1.292 Not sure if you'd like to use this PR or a new one to start clean. But adding test_cases subfolders to the base and stricter configs' exclude is a good call.

Brilliant. I opened #9646 for the pyright bump.

Closing in favour of #9650