Release v2.4.4 by tig · Pull Request #5466 · gui-cs/Terminal.Gui

tig · 2026-06-02T01:36:52Z

Release v2.4.4

Version: 2.4.4
NuGet Package: Terminal.Gui 2.4.4

What happens when this PR is merged

✅ The Finalize Release workflow will automatically create tag v2.4.4
✅ The Publish workflow will build and push to NuGet.org
✅ A GitHub Release will be created with auto-generated notes
✅ A back-merge PR from main → develop will be opened

Checklist

CI passes on this PR
Version looks correct: 2.4.4
Release notes reviewed (will be auto-generated on merge)

NeedsDrawRect and the viewPortRelativeRegion parameter of SetNeedsDraw are now in viewport-LOCAL coords: (0, 0) is the top-left visible cell of the View's Viewport, independent of Viewport.Location (scroll offset, possibly negative under AllowNegativeX/Y). Previously the no-arg SetNeedsDraw() passed Viewport (with the scroll offset baked in), DoDrawAdornments escalation did the same, and the subview cascade compared parent's viewport-relative region to subview.Frame (in parent's content coords) without accounting for the parent's scroll. PR #5431 worked around the resulting inconsistency by gating its region-aware ClearViewport narrowing on "Viewport.Location == Point.Empty"; this commit normalizes the convention so that workaround can be removed in a follow-up commit. Changes: - SetNeedsDraw() no-arg now passes new (Point.Empty, Viewport.Size). - DoDrawAdornments NeedsDrawRect escalation does the same. - SetNeedsDraw(Rectangle) subview cascade translates the viewport-local region through: parent scroll -> content coords -> intersect with subview.Frame -> subview frame-local -> subtract subview adornment offset and subview scroll -> clip to subview viewport bounds. If the result is empty (dirty area is only in subview's adornments/scrolled- off content), fall back to a no-arg subview SetNeedsDraw so the subview redraws safely. - View.Layout.SetFrame translates the union(old, new) frame from SuperView content coords to SuperView viewport-local before passing to SuperView.SetNeedsDraw. - Update NeedsDrawRect_Is_Viewport_Relative test scrolled-frame assertions to reflect the new convention (NeedsDrawRect no longer acquires scroll-position offsets). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…e normalized PR #5431 added a "Viewport.Location == Point.Empty" early-return to CanNarrowClearToNeedsDrawRect because the cascade in SetNeedsDraw produced rects in a different coordinate space than ViewportToScreen expects. Issue #5359 normalized NeedsDrawRect on viewport-local coordinates, so the conversion via ViewportToScreen is now safe regardless of the view's scroll state. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Nine tests pin the viewport-local coordinate convention: - SetNeedsDraw() no-arg on a scrolled view stores (0, 0, W, H). - SetNeedsDraw() no-arg on a view with negative Viewport.Location stores (0, 0, W, H). - Subview cascade from a scrolled parent produces correct subview- viewport-local rect (parent scroll translated out). - Subview-with-padding cascade produces subview-viewport-local coords (adornment offset subtracted), not subview-frame-local. - Scrolled-subview cascade subtracts the subview's own scroll on top of the frame and adornment offsets. - Dirty region overlapping only the subview's adornment area falls back to a full subview SetNeedsDraw (the safe fallback). - Framework ClearViewport narrowing now works on scrolled views (the scrolled-view guard removal from this PR). - Zero-size viewport is a no-op. - Repeated small invalidations on a scrolled view accumulate as a viewport-local union. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…ests CoPilot review of PR #5435 caught three issues: 1. (Blocking) The subview cascade in View.NeedsDraw.cs subtracted subview.Viewport.Location on top of the adornment offset when converting to subview-viewport-local. That's wrong — viewport-local coords are scroll-INDEPENDENT ((0, 0) is always the top-left visible cell regardless of scroll; scroll changes which content cell appears there, not the viewport-local coords themselves). Subtracting the scroll shifted the dirty rect to a different on-screen position than the cells that were actually dirtied. ViewportToScreen treats input as scroll-independent, so the bug would also corrupt the framework's narrowing in CanNarrowClearToNeedsDrawRect for the subview's later self-redraw. Fix: only subtract the adornment offset. 2. (Medium) The RegionAwareClearViewportTests.FrameworkDraw_ScrolledView_FallsBackToFullClear test documented the workaround guard removed by this PR. With the guard gone the test was passing accidentally because its dirty rect (1, 6, 3, 2) was outside the 5-row viewport visible area (Y=6 > 4), so the narrowing intersected to empty and fell through for an unrelated reason. Delete the test — NeedsDrawCoordTests's FrameworkNarrowing_NowWorks_OnScrolledView covers the correct behavior. Replace with a comment block explaining the removal. 3. (Low) SetFrame's scroll translation in View.Layout.cs was only exercised by tests with SuperView.Viewport.Location == Point.Empty (where the translation is a no-op). Add SetFrame_ScrolledSuperView_TranslatesInvalidationToViewportLocal to cover the scrolled case. Also update SetNeedsDraw_ScrolledSubview_... test: the previous assertion locked in the buggy shifted rect. With the cascade corrected, parent dirty viewport-local (2, 2, 4, 2) → subview viewport-local (2, 2, 4, 2) (same on-screen cells, just rendered by the subview now). Full suites still green: 17285 / 74 / 433. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Correct the step-4 cascade comment to say GetViewportOffsetFromFrame() is the combined Margin/Border/Padding inset (not just Padding), and update the plan to match the implemented contract — the cascade does not subtract the subview's Viewport.Location because viewport-local coordinates are scroll-independent. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

The Copilot autofix in 491f190 opened <para> twice with a single closing tag. Remove the stray opening tag so the doc comment is well-formed. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…zation Fixes #5359. Normalize NeedsDrawRect to viewport-local coordinates

Back-merge v2.4.3 from main into develop

Extract the self-content draw (Text + Content + local-context merge) into a private DrawSelfContent helper that owns the full _localDrawContext lifecycle for a self-redraw pass: create, populate, merge. Move the context creation out of the viewport-clear block — DoClearViewport only writes the shared context, so the local context isn't needed until DrawSelfContent, keeping create/populate /merge in one place. This replaces three separate `if (needsDrawSelf)` blocks (whose alignment had to be maintained by hand) with a clear/self split documented as intentional around DoDrawSubViews, plus an authoritative lifecycle comment on the _localDrawContext field describing why it is preserved across child-only passes (TransparentMouse CachedDrawnRegion contract from #5431). Behavior-preserving; draw order unchanged. Side effect: collapsing creation and use into one method lets nullable flow analysis prove _localDrawContext non-null, removing a CS8602 warning. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>