Pull request overview

Fixes macOS panadapter popouts rendering stale/static spectrum after cross-window reparenting, and addresses layout/state issues when floating/docking pans.

Changes:

• Add macOS-specific reparent refresh logic for SpectrumWidget plus a float/dock signal hook for dimension re-push.
• Rebuild the docked pan splitter after float/dock to avoid blank placeholder space.
• Prevent late-arriving pans from re-triggering the initial “restore floating state” debounce.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Thanks for tracking down a particularly thorny native-window/QRhi interaction, @rfoust — and for already absorbing the Copilot feedback (AETHER_GPU_SPECTRUM guard, no setParent(nullptr) transient, layout-aware rebuild). This looks good overall; just a couple of small cleanup observations, none blocking.

Redundant size handling in dockPanadapter() (PanadapterStack.cpp:564-583)

After this change, the existing splitter-sizing block is now dead/dueling code:

m_splitter->addWidget(applet);   // adds to OLD splitter
applet->show();
rebuildDockedSplitter();         // reparents applet again into NEW splitter, schedules equalizeSizes()

const int count = m_splitter->count();   // operates on NEW splitter (top-level slots, not pan count)
if (count > 1) {
    int total = ...;
    int each = total / count;
    ...
    m_splitter->setSizes(sizes);          // immediately overwritten by queued equalizeSizes()
}
m_splitter->updateGeometry();

Two small things worth tidying:

• m_splitter->addWidget(applet) happens against the soon-to-be-deleted old splitter; rebuildDockedSplitter() then reparents it again. You can drop the first addWidget (rebuildDockedSplitter already collects the docked applet via m_pans once it's been removed from m_floatingWindows).
• The manual setSizes(...) block runs synchronously, but rebuildDockedSplitter() queues equalizeSizes() via QTimer::singleShot(0, ...) which immediately overrides it. Safe to delete.

Saved PanadapterLayout may not match what's on screen after float

rebuildDockedSplitter() falls back to a count-based default (2v/2h1/2x2/vertical) when the saved layout doesn't match docked.size(), which is the right call. But AppSettings::PanadapterLayout is left at the previous value — so if the user had 4v saved, floats one, then re-docks, you re-enter rebuildDockedSplitter() with 4 docked applets and 4v matches, but 4v has no branch in the if/else chain and falls through to addVertical() (which is the right behavior, just by accident). Worth either updating PanadapterLayout on float/dock or being explicit about 4v/3v mapping to addVertical(). Not urgent.

Minor: setProperty("fired", ...) (MainWindow.cpp:2079, 2134, 2899)

Using QObject dynamic property as a one-shot latch works, but a plain bool m_layoutRestoreTimerFired{false}; member next to the existing m_layoutRestoreTimer is a hair clearer and avoids the QVariant round-trip. Style-only, take it or leave it.

Things that look right

• refreshAfterReparent() macOS path is appropriately scoped behind both Q_OS_MAC and AETHER_GPU_SPECTRUM, so software-spectrum builds keep the lighter resetGpuResources() path.
• The panFloated/panDocked signal + 200 ms display pan set xpixels=... repush is a reasonable workaround for the radio not knowing the new surface dimensions until after the reparent settles.
• AppSettings::instance().save() after saveFloatingState() matches the convention used elsewhere for state that needs to survive a hard exit.
• Layout-aware rebuildDockedSplitter() correctly handles the 2x2/2h1/12h/etc. nested-splitter cases, and reparents directly via addWidget() without a transient top-level state.

Nothing here is blocking from my read.