Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

Windowing foundation: one model, decided once a frame, tested without a screen

Open
  • Is something wrong?
#243 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
25/100
Issue type
Refactor
Clarity
Mostly clear
Activity status
Active
Tech stack
zig
Domain
backend, desktop

Research direction

Start with docs/WINDOWING_FOUNDATION_PLAN.md, which contains the review findings and breaks the work into separate PRs; note the stated dependencies on #237, #238, and #242. Choose one step and follow its named scope, such as running the app module tests in step 1 or measuring Objective-C messages in step 4. Done means completing that step while preserving today's behaviour; this issue describes a broad, multi-step plan rather than a single newcomer-sized change.

Written by the indexing model from the issue text.

Description

The plan is docs/WINDOWING_FOUNDATION_PLAN.md (#246). Every step keeps today's behaviour. They make the windowing (floats, menus, dialogs, view drags, drop zones) testable without a screen and cheap in native calls, share one model across platforms, and leave the backend usable by any dvui app. It builds on NATIVE_WINDOWS_PLAN.md and WINDOWS_LINUX_GLASS_PLAN.md, and doesn't repeat them.

Findings from the 2026-10-08 review are in the doc. In short:

  • one Viewport struct serves five kinds of window;
  • Popout re-decides every window by hand each frame, through five near-copies of fill-and-present;
  • there are seven coordinate spaces and two different band thresholds;
  • drag hit tests repeat once per region per frame;
  • macOS makes about 20 Objective-C messages per float per frame, changed or not;
  • glass looks are mirrored four deep and the shader is hand-synced three ways;
  • 32 layout tests never ran;
  • nothing popped out is tested.

Steps, each its own PR:

  • 1. Run the app module's own tests (32 that never ran): #244
  • 2. Menus and dialogs stay auxiliary windows on macOS: #245
  • 3. Typed coordinate spaces and one bandOf in viewport_map, after #237 and #238
  • 4. OsWindow: kinds as a union, native handles cached, dressing applied only on change, per-kind budgets (measure Objective-C messages per frame before and after)
  • 5. The frame's windows as a pure plan the backend reconciles, plus a fake backend and the first integration tests that pop a float out
  • 6. One fill for every OS window, with a target pool
  • 7. A drag model computed once a frame, with tests, and the release rules in one table
  • 8. Float, window and drag lifecycles as explicit states
  • 9. Glass: published capabilities, one look record end to end, shader parity (after #242)
  • 10. Full-screen transitions driven by Core Animation, if release builds don't hold 8.3 ms
  • 11. dvui changes on the fork (subwindow render redirection, a screens hook, subwindow kind and parent), the backend as a package, and dvui's example app as the acceptance test
  • 12. Remove the dead viewport API, along with steps 3–5
Dominant language
Zig
Stars
1.5k
Forks
40
Avg merge
5h 35m
Merged PRs (30d)
151

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from fizzyedit/fizzy

All issues in fizzyedit/fizzy

Similar issues

More Backend & API Design issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.