Upcoming breaking change to Trio private internals

[Zac-HD]

Things to check first

• I have searched the existing issues and didn't find my bug already reported there

• I have checked that my bug is still present in the latest release

AnyIO version

latest

Python version

any

What happened?

python-trio/trio#3256 adds some debug information to trio.Cancelled. I think it would be nice for AnyIO to have something similar, but the more important part is that this changes the private Cancelled._create constructor, which will break AnyIO-on-Trio at the upcoming Trio release.

How can we reproduce the bug?

Run any program involving cancellation, using Trio's main branch.

[agronholm]

OK, how should I go about instantiating trio.Cancelled going forward?

[Zac-HD]

CancelReasonLiteral: TypeAlias = Literal[
    "KeyboardInterrupt",
    "deadline",
    "explicit",
    "nursery",
    "shutdown",
    "unknown",
]
#####
        @classmethod
        def _create(
            cls,
            *,
            source: CancelReasonLiteral,
            source_task: str | None = None,
            reason: str | None = None,
        ) -> Self: ...

I think the minimal action is just to pass source="unknown"; longer term we've been pretty happy about the source-task tracking and CancelScope.cancel(reason=) argument too, as ways to distinguish what's happening when there's an ExceptionGroup of way too many cancellations.

[agronholm]

Would it be okay to add unknown as the default value to Trio? That would preserve backwards compatibility at no cost.

[Zac-HD]

Yep, seems very obviously a good idea, and renders this a non-breaking change. Want to make a Trio PR and tag me to approve + merge it?

[agronholm]

Linking it here too: python-trio/trio#3271

[Zac-HD]

Fixed in Trio so this is no longer a breaking change 😍