Skip to content

[Bug] Re-using previous linearisation matrices (dIydI, dIydtheta, and dvdId) in evolutive solves #92

Description

@kpentland

Codex flagged (in PR #64) that when callers supply the supported dIydI argument while leaving the default mode_removal=True, build_linearization takes the cached-Jacobian branch and never creates self.dvdId or self.initial_currents_plasma_descriptor. The constructor then unconditionally indexes those attributes here, so initialization fails with AttributeError; initialize compatible descriptor state before applying the retained-mode mask, even when the current Jacobian is supplied.

Given we never really used this feature of re-using linearisation matrices, we should create a few examples and stress test this to find the bugs. Then explicitly state how the user needs to provide the matrices in order to do solves rapidly.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions