feat(backend): upgrade to Anki 26.09.3

[david-allison]

Note

Assisted-by: GPT-6

Purpose / Description

Bump the backend

Fixes

• Part of Integrate Anki 26.09 #21926
• Fixes [BUG] FSRS - Help Me Decide - Simulate causes BackendFatalError: PoisonError  #21923

Approach

Pretty simple bump:

• handle the new endpoints
• Functionality was pre-prepared, see links in Integrate Anki 26.09 #21926

How Has This Been Tested?

Unit tests and integration tests for Card Info and Deck Options have been added.

MathJax is now tested, thanks to a code review.

Learning (optional, can help others)

In future, specify that I want GPT to fix bugs which are introduced, not 'bugs', as it debugged the entire implementation

• https://github.com/ankitects/anki/releases/tag/26.08
• https://github.com/ankitects/anki/releases/tag/26.09
• https://github.com/ankitects/anki/releases/tag/26.09.1
• https://github.com/ankitects/anki/releases/tag/26.09.2
• https://github.com/ankitects/anki/releases/tag/26.09.3

Checklist

• You have a descriptive commit message with a short title (first line, max 50 chars).
• You have commented your code, particularly in hard-to-understand areas
• You have performed a self-review of your own code
• UI changes: include screenshots of all affected screens (in particular showing any new or changed strings)
• UI Changes: You have tested your change using the Google Accessibility Scanner

[github-actions]

Snapshot diff report vs main. Open screenshot-diff for diffs.

• PreferencesScreenshotTest: 1 change

All 1 changed screenshots

PreferencesScreenshotTest

• AboutFragment_compare.png

[david-allison]

[mikehardy]

In future, specify that I want GPT to fix bugs which are introduced, not 'bugs', as it debugged the entire implementation

😆

("oops! sorry I fixed that")

[mikehardy]

Seems reasonable enough on a human scan. It's in the queue for auto review as well

[mikehardy]

Requesting changes on the rebase, not on the idea of landing 26.09.3.

Local (post-rebase e7817c7): lint / unit / package / emulator (API36_GAPI_PS16K) all green here. GitHub CI on the new tip may still be catching up — please confirm it goes green before merge.

Do not Fixes #21926. That issue is the 26.09 integration umbrella. #22156 is still open, explicitly “Part of #21926”, and already has requested changes. Use Part of #21926 here (keep Fixes #21923 only if you are willing to close that crash without an AnkiDroid-level regression).

MathJax (PreviewerHelpers.kt): Removing the eager <script> tags is the 26.09 lazy-load path. There is no test that a \(...\) card in previewer / new study screen still typesets, and bodyClassForCardOrd still always stamps mathjax-rendered. Please add a failing-then-fixed check (even stdHtml no longer contains those two asset URLs, plus @NeedsTest for actual typesetting) or say why ViewerResourceHandlerTest is enough.

RPC allow-list (PostRequestHandlerTest.kt): The new skip list will hide missing handlers for any name you classified as editor/desktop-prefs. Confirm none of those names are posted by shipped pages (deck-options, card-info, graphs, import-*, image-occlusion, congrats). Prefer listing skip names next to the Anki route that owns them.

Coverage vs the 26.09 checklist: CSV close-path is the only new behavioral test. Card Info / FSRS panic / CSP javascript: / deck-options save+optimize were called out on #21926 and are not re-validated on this bump. If that work is entirely in already-merged prep PRs, say so on the PR so reviewers are not guessing.

Thanks for splitting the bump from the prep PRs — the surface area here is small and the import close emulator test is the right kind of check. I still would not merge until #21926 is not auto-closed and the MathJax/FSRS gaps are either tested or explicitly deferred with @NeedsTest.

[david-allison]

In addition on the testing front:

• fix(anki-26.09): defer reloading the page until 'Save & Optimize' has been handled #22165 - we already caught a reload race condition [fixed in this prep commit]
• fix(reviewer): handle Anki 26.09 '/_anki/js/mathjax.js'  #22108 - bundled MathJax assets are tested, but this was not an integration test

done

[mikehardy]

hahahaha my review bot is harsh!

[mikehardy]

Hey David 👋

Thanks for the follow-up — Part of #21926, the MathJax previewer typesetting test, Card Info + Help Me Decide, importDialogRequireClose, and the RPC skip-list docs all close out the previous RC.

GitHub CI is green on b947618; local lint / package / emulator were green on a clean rebase onto current main (same four commits; not force-pushed). Local full unit still hits the known ReminderLogTree suite-order pollution (solo green; same on main) — not blocking. LGTM.

[mikehardy]

My review bot is slightly less harsh now ;-)

[david-allison]

Thank you all

And now the fun begins 😁