fix(gui): reopen the Aetherial strip after minimize — Principle XI.

[crypticpy]

Summary

Fixes #5365.

Minimizing the Aetherial Audio Channel Strip made it unrecoverable from its own
button: no number of presses on the AetherVoice button (or the chain applet's
nub) would bring the window back, leaving the Dock/taskbar as the only way to
reach it.

Two Qt behaviours combine to cause it, and toggleAetherialStrip() was written
as if neither existed:

1. QWidget::isVisible() stays true while a window is minimized, so
   if (m_aetherialStrip->isVisible()) sent a minimized strip down the hide()
   branch.
2. QWidget::show() on a minimized window restores its saved state — still
   minimized — so the follow-up press did not recover it either.

The decision is extracted into gui/WindowShowState.{h,cpp}:

• windowIsShowing(w) — visible and not minimized.
• showAndRaiseWindow(w) — showNormal() when minimized, show() otherwise,
  then raise + activate.

showNormal() stays guarded on isMinimized() because calling it
unconditionally would clear a Maximized or FullScreen window — the same reason
the net-reminder and tray raise paths in MainWindow_Nets.cpp guard it (#3918).
That helper is where those two sites should eventually converge; they are
deliberately not touched here, to keep this change scoped to the reported
bug.

The docked CWX and DVK panels use the same bare isVisible() shape but are
child widgets that cannot be minimized independently, so they are unaffected and
are left alone.

Constitution principle honored

Principle XI — Fixes Are Demonstrated. New regression test
window_show_state_test drives a real QWidget through
hidden → shown → minimized → restored and runs the toggle exactly as
MainWindow does. Mutation-checked three ways, each failing a different
assertion:

• windowIsShowing() without the !isMinimized() term → 3 failures
• showAndRaiseWindow() with an unconditional show() → 2 failures
• showAndRaiseWindow() with an unconditional showNormal() → 1 failure
  (the Scheduled Event Doesn't Show Correct Slice Matching Tuned Frequency #3918 maximized guard)

Principle VIII — Evidence Over Assertion. Both Qt behaviours above are
pinned against a real QWidget on the offscreen platform, not asserted from
memory.

Test plan

• Local build passes (cmake --build build) — clean, exit 0
• Behavior verified on a real radio if applicable — N/A: pure window
  management in MainWindow, independent of the connected backend
• Existing tests pass (CI) — window_show_state_test passes; full local
  suite run on this build tree shows only the two failures that reproduce
  unchanged on clean main (bridge_docs_check, hl2_state_restore_test),
  neither of which touches GUI window state
• Reproduction steps documented if user-reported bug — in Aetherial Audio Channel Strip cannot be reopened after minimizing — its own button no longer restores it #5365

Checklist

• Commits are signed (docs/COMMIT-SIGNING.md) — GPG, GitHub reports
  verified: true
• No new flat-key AppSettings calls — this change adds no settings
• Code is clean-room — AetherSDR's own widget code against public Qt API
  (Principle IV)
• All meter UI uses MeterSmoother — N/A, no meter UI is touched
• Documentation updated if user-visible behavior changed — no doc change
  needed; no document describes the old toggle behaviour. CHANGELOG.md
  deliberately untouched
• Security-sensitive changes reference a GHSA if applicable — N/A

---

Note on the claim protocol (AGENTS.md §Issue / PR Claim Protocol): assignee
changes are rejected for an account without write access to this repo, so the
Fixes #5365 link is the visible claim on the issue timeline instead.

[ten9876]

Issue fit

#5365: the Aetherial strip button treated a minimized strip window as "already showing", so pressing it did nothing — the operator had to un-minimize by hand. The fix makes windowIsShowing() require !isMinimized() and adds showAndRaiseWindow() that un-minimizes before raising. Correct and directly on-target, extracted into a testable free function with a mutation-checked regression test (window_show_state_test) that pins both Qt behaviors and the #3918 maximize guard. Principle XI honored.

Scope

Clean — the helper, its one call site, the test. CHANGELOG.md correctly untouched. Preflight: no sockets, pure widget-state test.

Blockers

None.

Nits (non-blocking)

1. Minimize-from-maximized reopens un-maximized (inline). showNormal() clears both the Minimized and Maximized bits, so maximize the strip → minimize → reopen brings it back at normal size, not maximized — a corner of the exact reopen path this PR fixes, and one the test misses (case 7 only checks maximize is preserved when the window was never minimized). w->setWindowState(w->windowState() & ~Qt::WindowMinimized) clears only the minimized bit and preserves maximize.
2. The isMinimized()/showNormal()/raise pattern now lives in three places (MainWindow_Nets.cpp:226-247 has two copies for #3918). The body acknowledges convergence "eventually"; this helper takes a QWidget* and could absorb those net-reminder/tray paths now, so a future fix (like nit 1) lands once.

What was verified vs read

• Verified by me: showNormal() clears the maximized bit (nit 1's mechanism, uncovered by the test); the helper resolves unqualified in MainWindow.cpp's namespace.
• From the automated pass, verified: first-press-creates-then-shows still works; the mutation test genuinely pins the two Qt behaviors.
• Not run: no bridge session (the widget-state helper is unit-covered), tests read not executed — no CI yet, which is a merge gate.

[ten9876]

Nit (non-blocking) — showNormal() drops a prior maximize. It clears both the Minimized and Maximized bits, so minimize a maximized strip and reopen → it returns un-maximized. Clear only the minimized bit to preserve the pre-minimize size:

Suggested change

	w->showNormal();
	if (w->isMinimized())
	w->setWindowState(w->windowState() & ~Qt::WindowMinimized);
	else
	w->show();

Add a test row (maximize → minimize → reopen → still maximized); case 7 doesn't cover the minimized-from-maximized path.

[crypticpy]

@ten9876 Both nits taken, pushed as 5a48797.

Nit 1 — minimize-from-maximized. showAndRaiseWindow() now clears only Qt::WindowMinimized through setWindowState(), exactly as you suggested. That also removes the isMinimized() branch: on a hidden widget the cleared state is pending and show() applies it, on a visible one it takes effect at once and show() is a no-op. Case 8 in window_show_state_test pins it (maximize → minimize → toggle → still maximized). Mutation-checked against the old showNormal() body: only the final "restored window is still maximized" assertion fails, so the row is measuring what it claims.

Nit 2 — the two copies in MainWindow_Nets.cpp. Both (net-reminder tune, tray-message click) now call showAndRaiseWindow(this). They had the identical latent un-maximize, so the nit-1 fix reaches them in the same commit. Disclosing here since it widens the diff by one file beyond #5365's strip: no behaviour change other than preserving maximize on those two raise paths.

One style-only change of my own: the strip toggle in MainWindow.cpp gets its braces back to match the surrounding code.

[Ozy311]

Issue fit

The fix resolves #5365. Independently reproduced the lockout in the actual pre-fix application and verified one-click recovery in the PR application on macOS 26.5.2 (25F84), arm64, Qt 6.11.1. The latest commit also addresses both points from the earlier maintainer review: restoring a minimized window preserves maximized/fullscreen state, and the two net-reminder raise paths share the helper.

Scope

Files	Change and provenance	Verdict
MainWindow.cpp, WindowShowState.{h,cpp}	Correct the strip toggle and restore only the minimized state bit; described by #5365 and the review follow-up	In scope
MainWindow_Nets.cpp	Reuse the helper at both raise sites, explicitly requested in the earlier maintainer review and disclosed in the author's follow-up	In scope
CMakeLists.txt, tests/tests.cmake, window_show_state_test.cpp	Compile the helper and register socket-free Qt widget regression coverage	In scope

No unrelated changes, new settings, protocol changes, dependencies, or changelog churn. The PR summary still describes the earlier showNormal() implementation and says the Nets sites are untouched; the follow-up comment accurately describes the final code.

Blockers

None found.

Nits — non-blocking

1. Refresh the PR summary and test comments to describe clearing WindowMinimized and the two Nets callers; references to an isMinimized() guard are stale.
2. Add braces to the new helper's null guard and the test's toggle branches to match AGENTS.md's control-flow convention (inline suggestion).

Independent verification

Built both the unchanged merge base 7f3eb98cd2df426dc4381e2a2203a1d59f600d1a and exact PR head 5a48797deeeef0057e761a68c652111215cd5ac7 as Mac applications. Both used RelWithDebInfo, Qt 6.11.1, and ENABLE_ASR=OFF; optional ASR is unrelated to this window path. The sandbox's icon-generation step failed identically on both; rerunning the builds outside the sandbox succeeded without source changes.

The full application checks used separate settings stores, explicit per-instance automation sockets, offscreen Qt, and only DEMO-0001 (AetherSDR Demo), with transmission pinned off. The actual AetherVoice launcher was invoked through VfoWidget/Aetherial Audio Channel Strip; restoration was not forced through the bridge's restore verb.

Recorded strip state:

Sequence	Pre-fix application	PR application
Open, minimize, press AetherVoice once	visible:false, windowState:"minimized"	visible:true, windowState:"normal"
Press AetherVoice again on pre-fix build	visible:true, windowState:"minimized" — still inaccessible	Normal hide/reopen behavior verified separately
Maximize, minimize, press AetherVoice	—	visible:true, windowState:"maximized"
Fullscreen, minimize, press AetherVoice	—	visible:true, windowState:"fullscreen"

The PR application also passed ordinary hide/show toggles, hide/show while maximized, and close/reopen while maximized. Both owned demo instances exited normally after validation; final demo model state reported transmitting:false and txPower:0.

The registered window_show_state_test passed through CTest. Its 15 assertions also passed independently on native Cocoa with no skips. Three deliberate mutations were caught: removing the minimized predicate (4 failures), omitting the minimized-state clear (3 failures), and using unconditional showNormal() (3 failures). An additional socket-free driver passed 11 assertions covering hidden/minimized restoration, fullscreen preservation, repeated geometry-preserving toggles, and a parent-owned top-level window matching the strip's ownership.

Inspected callers, ownership, surrounding code, and intervening main changes; no stale-base overwrite or overlapping fix found. Local test-registration and engine-boundary checks passed (existing tracked boundary warnings only). No separately invocable automated code-review skill was available; a second code audit was performed and its conclusions checked against the patch and runtime evidence.

Review snapshot and limits

Trusted governance snapshot: 50d70ebe9fdc4d6f01171b5cb17479bc89537be2, manifest PASS; canonical constitution and root mirror match. Both PR commits have valid GitHub-verified signatures. Approval remains subject to the repository's human CODEOWNERS requirements: infrastructure for CMakeLists.txt, reviewers for source/tests.

This review demonstrates the actual strip path on Mac plus native Cocoa helper behavior. Native Dock stacking, actual OS tray notifications, and physical-radio operation were not exercised. The full local suite was not run; the author's historical full-suite failure claims were not independently repeated.

Verdict

Approve with the non-blocking notes above. CI is green for Linux, macOS, and Windows, and Static Checks is green. Inspected the completed job logs; the platform test steps ran and passed. The head and existing discussion were rechecked immediately before submission. No merge action is included in this review.

[Ozy311]

Non-blocking convention nit: AGENTS.md requires braces on all control flow. Please brace this guard (and the toggle branches in the new test).

Suggested change

	if (!w)
	return;
	if (!w) {
	return;
	}