fix(timerangeselector): support uifigure parents (replace uicontrol labels with uilabel when needed)

[HanSur94]

Root cause

TimeRangeSelector is used in two contexts:

1. Classical figures — DashboardEngine.render() creates a figure(...); uicontrol is fine.
2. uifigures — CompanionEventViewer.buildFigure_() creates a uifigure(...); uicontrol errors out with "Functionality not supported with figures created with the uifigure function."

The latter is currently broken — the MATLAB Tests (A-D) CI failure on main (TestCompanionEventViewer/testConstructorOpensFigure and ~14 sibling tests) all blow up at TimeRangeSelector/buildGraphics_:726 where the first of three uicontrol('Style', 'text', ...) calls runs.

Fix

• Detect the parent figure type once in the constructor via isprop(hAncFig, 'AutoResizeChildren'). That property exists on uifigure only; classical figure() handles lack it even though both report class matlab.ui.Figure on R2020b+. Stored in a new private IsUIFigureParent_ flag.
• Branch in buildGraphics_:
  • Classical → existing uicontrol('Style','text', ...) with normalized Position (extracted into buildLabelsClassical_, bit-for-bit identical to the previous code).
  • uifigure → new buildLabelsUIFigure_ that creates uilabel widgets with pixel Position and installs a SizeChangedFcn on obj.hPanel to recompute positions on resize. The pixel rectangles mirror the classical normalized layout ([0.045 0.05 0.30 0.32], [0.36 ...], [0.66 ...]). The same pattern (panel-SizeChangedFcn → recompute pixel layout) is used by MultiStatusWidget, IconCardWidget, and TextWidget.
• Chain any prior SizeChangedFcn on the panel so we don't strand sibling resize handlers; the saved handle is restored in restoreCallbacks_ (called from delete).
• Refactor setRangeLabels to dispatch through a new private setLabelText_(hLabel, str) helper that picks Text (uilabel) or String (uicontrol) based on IsUIFigureParent_, with a defensive fallback to the other property if the primary set throws. setRangeLabels itself is now branch-free.

Files changed

• libs/Dashboard/TimeRangeSelector.m — 178 insertions, 21 deletions (1 file)

New private members:

• properties: IsUIFigureParent_, OldPanelSizeChangedFcn_
• methods: buildLabelsClassical_, buildLabelsUIFigure_, layoutUIFigureLabels_, onPanelResized_, setLabelText_

restoreCallbacks_ extended to restore the saved SizeChangedFcn when we hijacked it.

Backward compatibility

• The dashboard path (DashboardEngine.render → classical figure(...)) hits buildLabelsClassical_, which contains the original three uicontrol calls verbatim — same options, same positions, same String writes via setLabelText_.
• Public API of TimeRangeSelector is unchanged. setRangeLabels(leftText, rightText, middleText) signature is unchanged. Public properties (hRangeLabelLeft/Middle/Right, RangeLeftText/MiddleText/RightText) are unchanged.
• The hRangeLabel* handles change underlying class (uicontrol→uilabel) on the uifigure path. They're declared SetAccess = private and a grep across libs/, tests/, examples/, benchmarks/, demo/ shows no external code touches .String, .Text, .ForegroundColor, .BackgroundColor, or any other widget-type-dependent property — the only readers are inside TimeRangeSelector.m itself, which already dispatches through setLabelText_.
• No new toolboxes; pure MATLAB.
• Namespaced error IDs preserved (TimeRangeSelector:*).

Test files exercised by this change

Existing tests still hold (no API surface changed):

• tests/test_time_range_selector.m — 6 invariants over setSelection/setDataRange/getSelection/OnRangeChanged (uses classical figure('Visible','off') → classical path)
• tests/test_time_range_selector_event_markers.m — setEventMarkers polyline behavior (classical path)
• tests/test_time_range_selector_reinstall_after_rerender.m — reinstallCallbacks after rerender (classical path)
• tests/suite/TestTimeRangeSelectorEventMarkers.m — class-based variant (classical path)
• tests/test_dashboard_preview_overlay.m, test_dashboard_preview_envelope.m, test_dashboard_range_selector_integration.m, test_dashboard_engine_event_markers.m — DashboardEngine integration (classical path)
• tests/suite/TestCompanionEventViewer.m — the failing suite (uifigure path). testConstructorOpensFigure + ~14 siblings should now pass since buildGraphics_ no longer throws when the parent is a uifigure.

Test plan

• CI: MATLAB Tests (A-D) — TestCompanionEventViewer/testConstructorOpensFigure and siblings flip green
• CI: MATLAB Tests for TestRangeSelector / TestTimeRangeSelectorEventMarkers / classical TimeRangeSelector tests stay green
• CI: Dashboard integration tests (test_dashboard_range_selector_integration.m etc.) stay green
• Manual: open a dashboard via DashboardEngine — slider labels still update on selection drag (classical path)
• Manual: open CompanionEventViewer — slider mounts inside the uifigure without error and label text updates on drag (uifigure path)

Caveats / known limitations

• The SizeChangedFcn chaining is best-effort. If a future parent (e.g. some uigridlayout cell) refuses SizeChangedFcn, the labels stay at their initial pixel position. This is acceptable for a fixed-height slider strip and isolated by try/catch so it never breaks construction.
• The fallback paths inside setLabelText_ (try Text → fall back to String, and vice versa) exist purely for defensive parity if a future refactor crosses widget types under this class. Today only one branch ever fires.

Pre-existing mh_style warnings (bg unused at L440, try, ... catch, end extra-comma at L702/703) were left untouched — they are not on lines this PR modifies.

Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 67.10526% with 25 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
libs/Dashboard/TimeRangeSelector.m	67.10%	25 Missing ⚠️

📢 Thoughts on this report? Let us know!

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'FastSense Performance'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.10.

Benchmark suite	Current: bbf3e4c	Previous: c271749	Ratio
Downsample mean std(5M)	0.099 ms	0.063 ms	1.57
Render mean std(5M)	1.701 ms	0.121 ms	14.06
Zoom cycle mean std(5M)	0.726 ms	0.377 ms	1.93
Downsample mean std10M)	0.213 ms	0.075 ms	2.84
Instantiation mean std10M)	0.901 ms	0.79 ms	1.14
Render mean std10M)	3.855 ms	2.544 ms	1.52
Downsample mean (50M)	86.153 ms	77.586 ms	1.11
Instantiation mean std50M)	14.535 ms	2.592 ms	5.61
Zoom cycle mean std50M)	0.439 ms	0.352 ms	1.25
Downsample mean (100M)	173.133 ms	152.908 ms	1.13
Downsample mean ( std00M)	1.684 ms	0.548 ms	3.07
Instantiation mean ( std00M)	148.409 ms	46.52 ms	3.19
Downsample mean ( std00M)	3.504 ms	0.548 ms	6.39
Instantiation mean ( std00M)	718.82 ms	46.52 ms	15.45
Render mean ( std00M)	94.167 ms	2.858 ms	32.95
Zoom cycle mean ( std00M)	1.069 ms	0.617 ms	1.73

This comment was automatically generated by workflow using github-action-benchmark.

CC: @HanSur94