Inject everything when assembling a MG hierarchy

[pbrubeck]

---

name: "MG: Inject coefficients"
description: ''
title: ''
labels: ''
assignees: ''

---

Description

Things were breaking when a solver is reused for a bilinear form with a coefficient that is not a Constant , e.g. functions in R or DG0, after updating that coefficient.
Turns out that we also need to manually inject everything in problem.F.coefficients().
But wait, there's also another issue. The manual injection has to ensure that the level above the current one is up-to-date with respect to the finest level, an assumption that only holds in the two-level case. PCMG will traverse the levels in the order [1, 2, 3, ..., finest, 0 = coarsest]. So we should inject all coefficients across the hierarchy the first time we are reassembling the problem.

Fixes Issues:

• TODO report the issue

Checklist for author:

• I have linted the codebase (make lint in the firedrake source directory).
• My changes generate no new warnings.
• All of my functions and classes have appropriate docstrings.
• I have commented my code where its purpose may be unclear.
• I have included and updated any relevant documentation.
• Documentation builds locally (make linkcheck; make html; make latexpdf in firedrake/docs directory)
• I have added tests specific to the issues fixed in this PR.
• I have added tests that exercise the new functionality I have introduced
• Tests pass locally (pytest tests in the firedrake source directory) (useful, but not essential if you don't have suitable hardware).
• I have performed a self-review of my own code using the below guidelines.

Checklist for reviewer:

• Docstrings present
• New tests present
• Code correctly commented
• No bad "code smells"
• No issues in parallel
• No CI issues (excessive parallelism/memory usage/time/warnings generated)
• Upstream/dependent branches and PRs are ready

Feel free to add reviewers if you know there is someone who is already aware of this work.

Please open this PR initially as a draft and mark as ready for review once CI tests are passing.

[wence-]

Is it sufficient to only do this in kspcomputeoperators case, or should it also be done in form_jacobian and form_function? When we did this before (cc @pefarrell) we did it by setting up hooks master...pefarrell/transfer-manager-inject-everything

That was never merged because it does too much work (reinjects too frequently). I think this suffers from the same problem. There's no need to reinject a coefficient of F if it hasn't changed. These days you can ask the dat for its state, keep track of that and if it changed since the last time you saw it, you know its changed and you must inject the function.