Skip to content

TODO.md

Latest commit

 

History

History
97 lines (80 loc) · 5.99 KB

TODO.md

File metadata and controls

97 lines (80 loc) · 5.99 KB
Raw

TODO

Outstanding work. Items are written to be picked up cold: each says what to change and why it matters.

Background reading, in this order:

Where the project stands

The delivery chain works end to end: a merge to main builds, signs, and publishes a pre-release to GitHub Releases, and one approval promotes that same build to a release. https://whiteboard.sqlbi.com reads its download links from the release manifest deployed beside it and needs no edit per release. The current product version is VersionPrefix in Directory.Build.props (1.2.1). Identity version for the Store package is VersionPrefix.0 (1.2.1.0).

Declaring that number is decision 20 in docs/decisions.md. What 1.0 was waiting on shipped during 0.9.x: Preferences, .wimport, Explorer and VS Code previews, the public documentation site, and Finger drawing (default when no pen is detected).

No numbered work remains. 1.2.0 answered discussion 78 with Mouse drawing — decision 23, with the alternatives kept in docs/mouse-mode.md — and 1.2.1 made it discoverable from the toolbar, decision 24. The video teaser is recorded and served from the landing page itself as site/teaser-av1.mp4 / site/teaser-h264.mp4 — the Vimeo-embed plan was reversed, see decision 19 in docs/decisions.md; the production script and staging assets are in docs/teaser/. The release manifests and the Store submission are done and proven: the manifests are live at https://whiteboard.sqlbi.com/stable.json, and the pipeline's Store stage has carried two releases through certification unattended. The install-side verification list was walked in full for 1.0.0, which was the last part of the chain that had only ever been reasoned about. winget is the one piece still waiting, below. All of it is described in docs/release-management.md.

Pen buttons: what was settled, and what is left

The barrel button is the only assignable one, and it takes Laser or Straight line. Adding an action means an entry in PenButtonAction, a choice in SettingsCatalog, and — if it swaps the tool rather than acting as a modifier — a case in MainWindow.BarrelToolFor. Nothing else needs to know.

Erasing is not assignable. The reverse end of the pen erases, and so does the upper side button, because they cannot be told apart:

  • The upper button and a reversed pen are the same signal. A trace from the development pen (PenTrace, enabled by pointing SQLBI_WHITEBOARD_PENTRACE at a file) settles what several rounds of inference could not. The device exposes exactly two buttons, Tip Switch and Barrel Switch — no eraser button, no secondary tip button. Clicking the upper side button and turning the pen round produce identical events: Inverted goes true, both switches stay up, pressure stays zero, and when either one lands the same tip switch closes. So an inversion is the eraser, full stop. A device that reports a real SecondaryTipButton would be distinguishable, and supporting one would mean re-introducing a second slot — worth doing only if such a device turns up.

  • The barrel switch masks the tip switch, and the ink there is recovered by hand. A barrel press and a barrel release each arrive as a stylus up with InAir true. After a release the pen keeps reporting Tip Switch=Up and InAir until the button is pressed again - while the tip is still on the glass, and while the packets still carry its real pressure (0.54 rising to 0.69 across one such gap in the trace). WPF delivers those as in-air moves, so the InkCanvas collects nothing and the ink drawn in between was lost. AccumulateMaskedTipInk keeps them instead and commits them as a freehand stroke when the gap ends. Two consequences worth knowing: a barrel transition splits the line into separate stroke objects, which shows as a seam where a highlighter overlaps itself and as several undo steps; and the recovered stretch appears when the gap closes rather than under the tip, because there is no wet-ink path for points the InkCanvas never sees. Giving it one means drawing a provisional stroke on the scene surface.

  • The straight-line constraint cannot start mid-stroke from a button on this pen. It can from Shift, and from the barrel button, because both are reported while the tip is down. Anything reported only through Inverted is not, since Invert and Tip are mutually exclusive on this device.

  • Two constants stand in for signals the hardware does not give. AppendPenInk calls four consecutive weightless packets a lift rather than a dropped reading — no digitizer misses four readings in a row. DefaultActivationDistance is 24 px: how far the pen must travel before the axis is settled. It was 8 px, which let a few milliseconds of the previous direction pick the axis; a trace of real strokes is the way to revisit it. Once settled the axis is kept for the whole segment, however far off it the hand drifts — turning a corner instead was tried and produced a staircase out of a diagonal.

Waiting on the first winget submission

Not work, but the reason winget is not finished yet. microsoft/winget-pkgs#421386 submits SQLBI.Whiteboard 0.9.2 and is in review. Nothing in this repository depends on it: the workflow that keeps the package current afterwards is already in place, and the token it needs is set.

Until that pull request merges, .github/workflows/publish-winget.yml fails on every release, because wingetcreate update has no previous version to read. That failure is visible and gates nothing.