Pyrefly ignores suppressions after multi-line nested strings in f-strings

[QEDady]

Describe the Bug

Pyrefly fails to recognize ignore comments if they appear after an f-string that contains a nested triple-quoted string spanning multiple lines e.g.

# Corrupt the state with nested triple quotes in f-string
_ = f'start{"""message
"""}end'
y: int = "hello"  # pyrefly: ignore[bad-assignment]

Pyrefly still errors in the last line.

Sandbox Link

https://pyrefly.org/sandbox/?project=N4IgZglgNgpgziAXKOBDAdgEwEYHsAeAdAA4CeSIAxAAQDCuATgwK7EAu1bAFjNXG6ja8A7hG7V08IZk4MIxWNQCOzXELjUI6amAC0-OegDmAHXQB9agF4dAcn6oGbYCZCvXAW3hojMM%2B7cQAF8YLFt-dDNSRE10DhtXHigoXFdqahoyBhgwKGjNI3RGXmxUTF1UODgIQq84sxAAGhAsnLzCNlwPKAoaAAVSbNzSPgwcAmoAY1x0SCNmBkEIGcIzGgBlGF4uNjZiOEQAekPW4cJGI0PQw8xcSbhD6dmahaWZw51GalQAN1RoVDYRRPOavNjLbS4dgQuCrdBkbgzXQ-GAMaozazUVwAZkIAEYAEzudAAbVRDEYcAAumZmOgIB5iIxpLpMBBspNwSjMbYakVsuF0EU2Lpsip2TBygBrGCkCqTSbeHnCRzoQXBZqoTkQFEAMWgMAoaCweCIZGCQA

(Only applicable for extension issues) IDE Information

No response

[loulanyue]

I reproduced this on current main (00ff2e4) and have a regression test that fails with no suppressions found (left: [], expected [(Pyrefly, 3)]). #claim

The lightweight comment scanner carries triple-quote state across physical lines but does not retain the surrounding single-quoted f-string context. It therefore interprets the nested string's closing """ on the second line as a new top-level triple-quoted string and hides the real suppression on the following line.

I am testing a scoped fix that keeps the current fast path and falls back to the existing Ruff parser's real comment tokens only when the lightweight scanner masks text that looks like a suppression. This avoids parsing every module twice. The original reproduction now reports 0 errors (1 suppressed), and a negative test confirms the same text inside a real multiline string is not treated as a suppression.

Work and analysis are assisted by Codex; I will disclose that in any PR and include the full test/lint results.

[github-actions]

@loulanyue you've claimed this issue — it's now assigned to you. Thanks for picking it up! 🎉

[github-actions]

This issue has someone assigned, but has not had recent activity for more than 2 weeks.

If you are still working on this issue, please add a comment so everyone knows. Otherwise, please unassign yourself and allow someone else to take over.

Thank you for your contributions!

[WilliamK112]

I reproduced this on current main with a failing suppression-parser regression test. I am working on a fix for the f-string/string-state handling and will link the PR here once it is validated.

[WilliamK112]

I’m taking this issue on, and the fix is now submitted in #4953. I’ll follow through on CI and review feedback there. I wasn’t able to self-assign the issue, but I’m happy for a maintainer to assign it to me (@WilliamK112).

[kinto0]

Thank you!! Assigning you now. In the future, you can hashtag claim (with a real hashtag) and it'll assign this to you