ponder-lab / Hybridize-Functions-Refactoring / #2530

82%

main: 83%

push

github

Separate category (b) into its two dispositions, in the message and the reason (#785)

* Split the category (b) status message by its evidence source

The `INPUT_SIGNATURE_INFERENCE` status message told users the recovery was
"tracked at #509", but #509 closed as completed on 2026-06-18. The pointer was
user-visible: it is surfaced on the PyDev wizard, so the tool directed
developers to a closed issue. Two tests pinned the string, and the Javadoc on
`inferInputSignature` and `AbsenceReason.NO_SHAPE_OR_DTYPE_EVIDENCE` repeated it.

The references cannot all be re-pointed at one tracker, because the two ways to
reach category (b) have opposite evidence situations. A container parameter has
concrete per-element shapes and dtypes that Ariadne holds and
`getTensorContainers` discards; that is recoverable and is now tracked at #781. A
type-hint-only parameter carries no dtype at all, and since an input signature
admits no dtype-⊤ (#494), there is nothing for any tool to synthesize and no
follow-up to cite.

Split the message on `Parameter.isTensorContainer()` so each case names its own
disposition, and point only the container half at a tracker. The absence reason
stays `NO_SHAPE_OR_DTYPE_EVIDENCE` for both; distinguishing them at the reason
level is #781's call, not this cleanup's. Update the two pinning assertions to
match, one now asserting the type-hint message cites no tracker.

`./mvnw verify` green: 617 tests, 0 failures, 0 errors, 0 skipped.

Closes #782.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K8tXuyy3YWhM7BzvmCxKAn

* Drop the issue reference from the wizard-facing text

The point of #782 is not that the status message cited the wrong issue but
that it cites one at all: a tracker is not actionable to a user, and a closed
one actively misleads. Replacing the `#509` pointer with `#781` reproduced the
defect rather than fixing it.

State the container disposition in plain ... (continued)

13 of 13 new or added lines in 2 files covered. (100.0%)

2154 of 2631 relevant lines covered (81.87%)

0.82 hits per line