ponder-lab / Hybridize-Functions-Refactoring / #3182

86%

main: 88%

push

github

Stop the forced static-shape-read check from failing a project (#986)

* Stop the forced static-shape-read check from failing a project

With `alwaysCheckStaticShapeReads` on, every function reaches signature
inference, including one the program never calls. When such a
function's parameter is classified as a tensor from its type hint
alone, it has no call-graph node and so no tensor types, and inference
dereferenced them, aborting the whole project with a
`NullPointerException` (#984).

A parameter with no tensor types has no evidence to exclude, so the
expected-failure check now treats it as having none, and it falls
through to the existing no-evidence disposition. Only the forced check
reaches this state: the default configuration runs the check only for
a function with a tensor parameter and no primitive one, which an
unreachable function never is.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

* Pin the state the static-read check used to crash on

The test now witnesses that the parameter is classified as a tensor from
its hint with no tensor types inferred, and that it reaches the
no-evidence disposition, rather than only that the check passes. The
null return is documented on `getTensorTypes()`.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NTnpjPjJj2aZy1ZkUD1eAn

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

1 of 1 new or added line in 1 file covered. (100.0%)

3680 of 4287 relevant lines covered (85.84%)

0.86 hits per line