[w5jwp]

Merge order: #6008 and #6033 must be merged before this PR. This branch is built on #6008 and merges in #6033 (with the conflicts resolved), so it carries their commits until they merge; it will then be rebased onto main. #6030 has merged.

Merge order for the PGXL / 4O3A PRs (the same map is in each of them; #6030 has merged):

Stage 1:
  #6008  authenticate TGXL, PGXL and Antenna Genius
Stage 2 (either order, after what each needs):
  #6027  Peripherals settings UI          needs #6008
  #6033  PGXL drain current (ID meter)    needs #6008
Stage 3:
  #6036  PGXL docked layout               needs #6008 and #6033

How to review: this PR's own work is b4dcd8c9 (the layout change) and 4cc897c2 (the review findings: a neutral OPERATE placeholder with an accessible name, the temperature column measured in the drawn font, and two test fixes). 1b0de46e merges #6030 in, and 3827ffea, 4ef42eae and 2186ef8e merge #6033 in (the last also brings #6008's hostname-bound credential); open them only to check the resolutions. Everything else in the diff is carried from #6008, #6030 and #6033, which are reviewed in their own PRs.

Summary

Fixes #5903.
Fixes #4885.

#6008 puts the connection indicator on its own line and provides the direct connection that shows all of the readings and controls.

In the docked (rail) view, the PGXL panel clipped its controls and readings: the 260 px rail had one row holding a stacked column of readings plus the MEffA, Fan and OPERATE controls. This PR gives the controls and the readings their own lines, and keeps the panel's shape whether the amplifier is offline or connected.

Docked, the panel now reads:

[ MEffA ]  [ Fan: Standard ]  [ OPERATE ]
PA 89.9 F    Vac 122 V
HL 89.2 F    Vdd 51.9 V            ● DIRECT

Changes

• Controls. MEffA, Fan and OPERATE have the row to themselves and share its width, at one height (22 px). MEffA keeps its three state colours but takes OPERATE's 10 px bold lettering.
• Readings in a 2×2 grid, in both presentations: PA over HL, and Vac over Vdd in a second column 10 px to the right. Docked, the connection indicator ends the HL/Vdd row; expanded, it keeps its own line below. Each is its own readout (label value unit) with its own screen-reader name.
• Vac and Vdd never move. A grid column is as wide as its widest cell, so they start at one x; and the temperature column is held at the width of its widest possible reading, measured in the face it is drawn in, so they also stay put as the temperatures change. Fixed-length text alone was not enough: where no true fixed-width face is installed the fallback is proportional, and 9.9 and 106.8 differ in width.
• Temperature units. PA and HL are both clickable, and a click on either switches both between Celsius and Fahrenheit and saves the choice, as the single readout did. HL shows HL — in grey until the direct connection sends a value, rather than disappearing, so the rows never move. The hover outline is gone (the pointing-hand cursor already says it can be clicked); the keyboard-focus outline stays.
• The connection indicator shares Vdd's line exactly. Docked, it takes Vdd's height, so both centre the same way. At one pixel shorter, centring it needed a half-pixel offset that widget geometry rounds away, which lifted its text above Vdd's.
• Padding. Every readout keeps a fixed length, but the spare spaces go at the end instead of in front of the number (HL 89.8, not HL 89.8). The temperature readouts have no left padding, so they line up with the controls above them.
• Spacing. Docked, an 8 px pad separates the controls from the Id gauge, and the readings sit with equal space above and below.
• Fan labels are spelled out: Fan: Standard, Fan: Contest, Fan: Broadcast. The pull-down reserves room for the longest.
• OPERATE's colours come from the theme. It used a hardcoded green (#006030) under the theme's primary text colour, which is near-black under Default Light and unreadable there ([bug] 26.9.4 PGXL Applet cosmetically broken #5903, item 4). It now uses the lit MEffA key's tokens, so the dark theme looks as before and the light theme gets its own green with white text.
• Before the amplifier reports. The docked controls stay in place instead of hiding: MEffA is grey and disabled, OPERATE reads — and is grey and disabled (with the accessible name "Amplifier state not reported"), and Fan shows Fan: —, disabled. None of them claims a state. Clearing the fan selection is done with signals blocked, so it never sends a fan-mode command. PA and HL are grey until a reading arrives, like Vac and Vdd. The expanded panel's keys still wait for a state before appearing.

Maintainer rulings: accepted as built. That covers the expanded (floating) panel's 2×2 grid, the removed hover outline, the right-justified readouts, and showing the controls disabled before the amplifier reports instead of hiding them. On a station without a direct PGXL connection, MEffA and fan stay dimmed, and their descriptions say they need a direct PGXL connection.

Against #5903's triage: items 1 (controls on their own line), 2 (the rail MEffA button at 10 px) and 4 (OPERATE contrast) are done as proposed. Item 3 (the fan pull-down too wide when docked) is solved the other way: the controls share the full row, so the full labels fit, and they are spelled out rather than shortened to Std/Bcast. The docked-width test it asks for is added (below).

#4885 was left open until the docked Fan: Contest label could be checked on a current build: it reads in full here, and the new test checks all three fan labels at the rail's width.

Merging with #6030, #6033 and #5984: #6030 is merged into this branch (commit 1b0de46), because the two conflicted in 10 places across AmpApplet.cpp, AmpApplet.h and tests/amp_applet_test.cpp: #6030 renames the temperature setters and members (setTemp → setPaHeatsinkTemp, m_hasTempA → m_hasPaHeatsinkTemp, …) and changes updateTempLabel and setDirectConnected, all code this PR restructures into two readouts. The resolution keeps this PR's structure with #6030's names and behaviour: HL clears (and turns grey) when the direct connection drops, a late HL write is ignored, and each readout's tooltip names only its own label. #6030's fallback test is rewritten for the two readouts and now also checks the grey tone. #6033 is merged in too (3827ffea, 4ef42eae, 2186ef8e). It deletes the amplifier-status relay handler and AmpApplet::setMeff, and adds setRadioPaHeatsinkTemp and setRadioDrainCurrent beside the PA setter this PR restructures, so it conflicted here in three files. The resolution keeps this PR's two-readout layout and tooltips, takes #6033's deletions and evidence comment, and keeps the grey-to-live style change on the PGXL's own first PA reading. Its single-button tooltip assertions are dropped in favour of this PR's per-readout ones. #5984 (spelling out MEffA on the expanded panel's key) edits panel-key code this PR leaves alone.

Constitution principle honored

Principle XI (Fixes Are Demonstrated). pgxl_panel_test lays the panel out at the rail's real width (260 px) and checks that no control or readout is narrower than its text needs, and testDockedCaptionsFitTheRail repeats that for every caption the controls can show. Both fail against the old layout. The layout was also checked in the running app on the hardware below, offline and connected.

Test plan

• Local build passes
• Behavior verified on a real radio: FLEX-8600 (SmartSDR 4.2.20.41343) with a PowerGeniusXL on firmware 3.9.8. Docked, offline and connected with the direct connection up; Fan: Broadcast, the longest label, fits beside MEffA and OPERATE.
• Tests pass: pgxl_panel_test, amp_applet_test, pgxl_direct_protocol_test, amp_model_test, peripheral_auth_handshake_test, health_applet_test
• Test changes:
  • pgxl_panel_test: in both presentations, PA sits over HL and Vac over Vdd, with Vac and Vdd at one x. Docked at 260 px, the indicator ends the HL/Vdd row at the right edge, at Vdd's height and top edge; nothing is narrower than its text; the controls share one height; the space above and below the readings matches; the controls clear the Id gauge. Expanded keeps the indicator on its own line below, and the layout survives docked → expanded → docked.
  • amp_applet_test: new testDockedCaptionsFitTheRail shows every caption the rail's controls can take (each operate state word, all three fan labels, all three MEffA states, and the before-report placeholders) at 260 and 300 px, and checks each control is at least as wide as its caption needs. It compares against the control's own size hint, so a CI host that falls back to other fonts cannot fail it on a caption that fits. Against the old layout it fails at both widths: the fan pull-down got 70 px of the 130 its label needs. New testVoltagesStayAlignedWithTheTemperatures runs PA and HL through 9.9, 106.8, 100.4, 120.0 and -5.0 and checks that Vac and Vdd stay at one x and never move; without the pinned column width they moved by up to 7 px. testDualSensorToggle checks that a click on either temperature switches both and that the unit chosen from HL is saved. Also the new readout text; fixed length across digit changes; the before-state controls are shown but disabled with no mode; clearing the fan selection sends no command; the PA placeholder tone.
• Reproduction steps documented in the linked issues

Checklist

• Commits are signed (docs/COMMIT-SIGNING.md)
• No new flat-key AppSettings calls
• Code is clean-room
• All meter UI uses MeterSmoother (no meter changed)
• Documentation updated if user-visible behavior changed (no docs/ page describes this panel)
• Security-sensitive changes reference a GHSA if applicable (not applicable)

🤖 Generated with Claude Code

Review follow-up (2026-10-03)

• The column-pin test forces a proportional face on the temperature buttons, drives 9.9 → 106.8, and asserts the column minimum covers the widest reading plus the border. With the setColumnMinimumWidth(0, ...) line removed it fails.
• Justification (see PGXL panel (popped out): the MEffA, fan and STBY keys are squeezed below their size while the port strips are shown, and their captions clip #5983): only the numeric field is right-justified (label + rightJustified(value, 5) + unit), so neither the unit nor the number's right edge moves as the value changes.
• When there is no direct connection, the MEffA and fan descriptions say they need a direct PGXL connection.
• OPERATE gets an accessible description while disabled; Vac/Vdd say "not reported" before the first reading; each readout's name change is announced once (setAccessibleName posts NameChanged itself); with a direct connection and no fan mode yet, the fan pull-down says the amplifier has not reported one; placeReadings() skips re-flow when nothing changed.
• Comments describe current behaviour; the caption check uses the combo's edit-field rect; placeholder-tone checks compare against the resolved disabled colour.
• Known: pgxl_panel_test fails at :553-554 (source->height() == vdd->height()) in my sandbox, where the theme resources fail to load. It fails identically at the previous head f5eafa06, so it predates this change. It passes for you per your review.
• Merged fix(pgxl): read Id from the radio's ID meter. Principle VIII. #6033's branch.

[w5jwp]

Before. Both taken docked in the 260 px rail, on a FLEX-8600 (SmartSDR 4.2.20.41343) with a PGXL on firmware 3.9.8 connected directly (#6008).

1. With the amplifier operating. The temperature is cut off (PA 106.7 /, no HL value visible), and all three controls are squeezed below their captions: MEffA is clipped at both ends, the fan pull-down reads Fan: C… and OPERATE reads PERAT. The readings are stacked in a narrow column beside the controls.

2. The same clipping at another moment, as reported in #5903: PA 35.7 / H, Fan: St… and PERAT, with Vdd and Vac crammed against the left edge.

[w5jwp]

After. Same station, docked in the 260 px rail.

1. Connected, direct connection up. MEffA, Fan: Standard and OPERATE share the row at one height, with nothing clipped. PA over HL and Vac over Vdd form a two-column grid; Vac and Vdd start at one x and do not move as the temperatures change. The connection indicator shares Vdd's line. (Vdd reads 0.0 V because, on the Standard fan profile, the PGXL turns its supply off when idle.)

2. Before a connection. The docked controls hold their place instead of hiding: MEffA, Fan: — and OPERATE are grey and disabled, so none of them claims a state the amplifier has not reported. PA, HL, Vac and Vdd keep the same grid in grey with dashes, and ● OFFLINE sits on the Vdd line. The panel keeps one shape whether the amplifier is offline or connected.

[aethersdr-agent]

1. Issue fit

Yes for both, with one deviation the body already names. #5903's reporter (dark-theme variant, Windows 11, 236 px content width) and @G6PWY-Chris's Linux confirmation are both the docked 260 px rail clipping MEffA / Fan: St… / STANDBY. Giving the three controls their own full-width row at one height removes the ~300-px-of-minimums-in-228-px squeeze the triage measured, and the geometry guards at 260 px are the thing that was missing (the triage explicitly asked for the TGXL parity-test equivalent; tests/pgxl_panel_test.cpp now lays out at 260×380 and asserts width() >= sizeHint().width() for every visible control and readout). #4885 ("Fan: Contes" in the Panel) is met by spelling the labels out instead of shortening them to Std/Bcast — the opposite of triage item 3, disclosed as such, and the better answer to an issue whose complaint is the truncated word. Triage item 4 (light-theme kOperateStyle contrast) is fixed by moving to color.accessory.key.meffa.active.*; both tokens exist in default-dark.json:120-124 and default-light.json:121-125, so this resolves rather than silently dropping through. Note item 4 is not what the reporter saw — they answered "a variation of Dark" — so it is a ride-along fix the triage proposed, not part of the reported symptom.

One requirement neither issue asked for: the docked controls now stay in place, disabled, before the amplifier reports. The body flags it for maintainer review, which is the right call — see Needs-maintainer-decision below.

2. Scope

File(s)	Change	Claimed?	Verdict
src/gui/AmpApplet.cpp, AmpApplet.h	controls row, 2×2 readout grid, pinned temperature column, theme-resolved OPERATE, spelled-out fan labels, before-report placeholders	Yes	In scope
tests/amp_applet_test.cpp	testDockedCaptionsFitTheRail, testVoltagesStayAlignedWithTheTemperatures, dual-readout toggle, before-report assertions	Yes	In scope
tests/pgxl_panel_test.cpp	260 px docked geometry, grid order, indicator on the Vdd line	Yes	In scope
26 other files (src/core/Pgxl*, Tgxl*, PeripheralAuth*, src/gui/PeripheralAuth*, RadioSetupDialog, MainWindow*, docs/, CMakeLists.txt, .gitignore, touchpoint manifest)	The peripheral-auth feature	Yes — first line of the body	Inherited, not this PR's

I verified the stacking claim rather than taking it: git fetch origin pull/6008/head and this checkout share 120ed439 → 80f40fef → 601b4b41 → 00d6f79e exactly, and b4dcd8c9 is the only commit this branch adds (4 files, 563 insertions). So the 26-file tail is byte-identical to #6008's head — no extra change smuggled in under the stack, and nothing here for me to re-review that #6008's thread does not already own. The "Merging" section's conflict inventory against #6030 / #6033 / #5984 matches what I see in the tree.

Expanded-panel readouts also change from one row of three to the same 2×2 grid. Neither issue is about the expanded panel — but splitting one PA x / HL y label into two readouts makes four cells, and four abreast in the panel is worse than 2×2, so I read this as a consequence rather than smuggled restyling. Worth the maintainer's eye since it alters a presentation nobody reported as broken.

No CHANGELOG.md edit. CI on b4dcd8c9 is 5/5 green (build, check-windows, check-macos, Static checks, Sanitizer configures) — note that is the frozen -R gate, so neither amp_applet_test nor pgxl_panel_test actually ran there; both are full-suite.yml.

Socket-test disclosure (§2): this PR modifies tests/pgxl_panel_test.cpp, an existing registered default-graph target that binds a loopback QTcpServer on QHostAddress::LocalHost, 0 and answers as a simulated PGXL peer (tests/pgxl_panel_test.cpp:123-145), skipping with 77 when it cannot bind. The changes here are UI-geometry assertions only — no new bind, listen, connect or peer behaviour, and no new fake peer. It is not in the PR -R gate. tests/amp_applet_test.cpp is socket-free. Recording this per the reporting half of the socket-test rule; it is not a finding.

3. Blockers

None. Everything I could construct that breaks users either did not reproduce in the code or predates this commit.

4. Needs maintainer decision

1. The before-report OPERATE placeholder reuses a real state word (AmpApplet.cpp:1576, inline). With no state, the button reads OPERATE in kBtnStyle — the same caption the operating state shows, and the same style the STANDBY state shows. Disabled-ness is the only thing distinguishing "not reported" from either. The same commit gives the fan combo Fan: — / "not reported yet" and both temperatures "not reported" accessible names, so this is inconsistent with the PR's own pattern; and m_operateBtn has no accessible name at all, so a screen reader announces "OPERATE, dimmed". Suggested in the inline comment.
2. The before-report rule itself. This replaces a deliberate rule whose comment was deleted: "Nothing before the amplifier has reported a state, on the same rule the fan controls follow: a control that cannot say what it is set to is worse than none." The new rule (present, disabled, asserting nothing) is defensible and keeps the row's shape, but it is a behaviour change to working UI that neither #5903 nor #4885 asks for. The body already routes it here, which is the right handling under GOVERNANCE.md §AI Contributors — no split needed, just a ruling. Fixing (1) would make the new rule considerably easier to accept.

5. Nits (non-blocking)

• The unit letter now moves instead of the number (AmpApplet.cpp:234, inline). leftJustified puts the padding after the unit, so Vdd 0.0 V / Vdd 100.4 V place V two columns apart, and PA 9.9 C / PA 106.8 C likewise. The old pad() right-justified the value, which pinned the unit and moved the number's left edge instead — the gap the comment rejects. Both jiggle; you traded one for the other in a change whose headline is stable readings. testVoltagesStayAlignedWithTheTemperatures pins the column, not the inside of the readout.
• Column width is measured from a font the sheet does not ask for (AmpApplet.cpp:981, inline): QFont("monospace") + setStyleHint(QFont::Monospace), while the sheet says only font-family: monospace. Where no real monospace family exists the two can resolve differently, and if the measured face is the narrower one setColumnMinimumWidth under-reserves and Vac/Vdd shift again — on Windows, not on the CI host where the new test runs. m_tempBtn->fontMetrics() after applyStyleSheet would measure what is actually drawn.
• HL keeps a stale, live-toned reading when the direct connection drops (AmpApplet.cpp:1608, inline). setDirectConnected(false) clears Vdd/Vac to — and greys them, and drops fan mode, but leaves m_hasTempB/m_tempB alone — and the radio does not relay HL. Pre-existing (the old combined label froze the same value), so not a regression, but this commit promotes it to its own readout beside a correctly-greyed Vdd, and the body says the #6030 merge resolution is what fixes it. If #6036 lands first, that stale HL ships alone.
• controlsFit dereferences meffa before the null check (tests/amp_applet_test.cpp:810, inline): b->parentWidget() == meffa->parentWidget() runs inside the discovery loop, while bool fits = meffa && fan && operate; is line 814. If ampMeffaButton is ever renamed, the Principle XI test segfaults instead of failing. (The object name does exist today at AmpApplet.cpp:511, so this is latent.)
• The #4885 guard does not actually measure the fan caption (tests/amp_applet_test.cpp:818, inline). With AdjustToMinimumContentsLengthWithIcon, QComboBox::sizeHint() skips the item texts and uses minimumContentsLength * advance('x'), so width() >= sizeHint().width() is a 14-character proxy, not "Fan: Broadcast fits". minimumContentsLength(14) happens to equal that label's length, and the drop-down width is added on top, so it is fine today — but the test would stay green if a future label outgrew the combo, which is exactly the #4885 failure mode. The QPushButton halves (MEffA, OPERATE) are genuinely covered, since their sizeHint derives from the text.

6. What I tried to break (and could not)

Findings below are reasoned from the head checkout at /tmp/aetherclaude/pr-6036; I have no build, no test run and no way to drive the GUI, so nothing here was reproduced at runtime.

• "The grid is placed from applyDensityAtScale, not buildUI" — the obvious break is an empty readout grid on first show. buildUI() ends with applyDensityAtScale(1.0) at AmpApplet.cpp:588, so placeReadings() always runs before the applet is visible. Survived.
• Reparenting m_sourceLabel between the grid and the column on every density pass. placeReadings() calls removeWidget on both containers (one is always a no-op), then re-appends to m_vbox after m_bottomStretch in the floating branch — so docked → floating → docked cannot leave the indicator above the bottom pad, and cannot leave it orphaned (show() at the end). Walked the order by hand in both directions. Survived.
• !known used before its declaration in the rewritten applyStateToControls — const bool known = !m_stateWord.isEmpty(); is at line 1560, ahead of every use. Survived.
• Appending QPushButton { font-size: 10px; } to the panel-key sheets for the rail MEffA button: if those sheets used a class selector (PanelKey {...}) the concatenation would be the only matching rule and MEffA would lose its three state colours. They are all plain QPushButton selectors (AmpApplet.cpp:126-157) and none sets font-size, so there is no conflict and no ordering dependence. Survived.
• Token existence for the re-pointed kOperateStyle, including the :hover variant — all four color.accessory.key.meffa.active.* keys are present in both bundled themes. kOperateStyle has no :disabled rule (unlike kPanelKeyMeffaActiveStyle, which has one precisely because a hard-set background beats the disabled palette), but setEnabled(known) means that sheet is never applied to a disabled button. Survived.
• kValueFieldChars overflow: it is 5, and the widest real temperature (106.8, -88.8) is exactly 5, so leftJustified never has to truncate; and even when a value did overrun, the pinned column min-width absorbs it rather than shifting the voltages. Survived.
• Clearing the fan selection firing a fan-mode command — both setCurrentIndex(-1) sites are inside QSignalBlocker scopes, and the test asserts fanModeChanged count stays 0 across a connect/disconnect cycle. Survived.
• Dangling references to the renamed/removed API: applyTelemetryLayout, m_telemetryInRow and voltsReadout have zero remaining hits anywhere in src/ or tests/; every new objectName is referenced only by this applet and its two tests. Survived.
• Build hazards from the new code — std::max / std::make_pair without <algorithm> / <utility> (both arrive transitively via QtCore containers; pgxl_panel_test.cpp does add <algorithm> explicitly), and no M_PI or windows.h/std::min pairing. check-windows and check-macos are green on this head. Survived.
• CodeGuard's six CG-PATH-001 advisories — refuted. src/core/TgxlConnection.h:19 is a protocol comment (// S0|state key=val ...); MainWindow.cpp:9549 is the NR2 wisdom dialog's label text and :9780-9782 is the macOS bundle-root QDir::cdUp() walk. None is a path operation, none is touched by b4dcd8c9 (MainWindow.cpp is not even in that commit), and #6008's thread has already dispositioned the same six. No action.
• Could not check: anything requiring a run — the actual rendered widths on Windows/Linux with the real font fallbacks (the font-resolution nit above is the specific place that matters), and whether the new 260 px geometry assertions pass on a CI host, since neither modified test is in the PR's -R gate.

7. Recommendation

Needs maintainer decision — on the before-report control states (item 1 is a concrete, cheap fix; item 2 is a ruling), and on the expanded panel picking up the 2×2 grid. The layout work itself is the right fix for both issues and the evidence is unusually good: the geometry guards run at the rail's real 260 px, they are claimed to fail against the old layout, and the scope claim about the inherited #6008 stack checks out commit-for-commit. Next step: settle the OPERATE placeholder caption and accessible name, then land behind #6008 (and fold in the #6030 resolution for the stale-HL nit, whichever order the two take).

Thanks for the merge-order and conflict inventory up front — verifying the stack took one git fetch instead of a guess.

---

_{🤖 aethersdr-agent · cost: $11.7145 · model: claude-opus-5}

---

GitHub rejected the inline anchors (GitHub API 422: Unprocessable Entity); findings below:

src/gui/AmpApplet.cpp:1576
With no state reported the caption is OPERATE in kBtnStyle — the same word the operating state shows, and the same style the STANDBY state shows. Only setEnabled(false) separates "not reported" from either. m_operateBtn also has no setAccessibleName/setAccessibleDescription anywhere in this file, so a screen reader gets "OPERATE, dimmed".

The same commit is careful about exactly this elsewhere — Fan: — plus "Fan speed: not reported yet", and "PA heatsink not reported" / "Harmonic Load heatsink not reported". A neutral caption plus a name would make the new before-report rule much easier to accept:

    m_operateBtn->setText(!known ? QStringLiteral("—")

…paired with m_operateBtn->setAccessibleName(known ? m_operateBtn->text() : tr("Amplifier state not reported")); just below the setToolTip call. Caption wording is the maintainer's call — the point is that the placeholder should not be a state word.

src/gui/AmpApplet.cpp:234
leftJustified puts the spare columns after the unit, so the unit letter is what moves now: Vdd 0.0 V vs Vdd 100.4 V place V two columns apart, and PA 9.9 C vs PA 106.8 C likewise. The old pad() pinned the unit and moved the number's left edge instead — the HL 89.8 gap the comment rejects.

Both jiggle, and you chose consciously, so this is an observation rather than a defect: in a change whose headline is that readings do not move, the readout's own interior still does. testVoltagesStayAlignedWithTheTemperatures pins the column boundary, not the inside of the cell. If it matters, right-justify only the numeric field and keep the fixed total (label + rightJustified(value, kValueFieldChars) + unit) — that pins both edges at the cost of the leading gap.

src/gui/AmpApplet.cpp:981
The column width is measured in a font the style sheet never asks for: here it is QFont("monospace") with setStyleHint(QFont::Monospace), while the sheet applied to these buttons says only font-family: monospace. Where no family literally named monospace is installed, the sheet's resolution and this one can land on different faces — and if the measured face is the narrower one, setColumnMinimumWidth(0, widest + 4) under-reserves and Vac/Vdd shift again, which is the whole bug this hunk exists to prevent.

testVoltagesStayAlignedWithTheTemperatures cannot catch that: it measures in the same process as the sheet, on the CI host. A style sheet's font-size/font-family is pushed onto the widget's own font by QStyleSheetStyle::polish, so measuring m_tempBtn->fontMetrics() after theme.applyStyleSheet(...) would measure what is actually drawn and remove the divergence entirely.

src/gui/AmpApplet.cpp:1608
This clears Vdd/Vac to — and greys them, and drops fan mode just below — but m_hasTempB / m_tempB are le

(truncated)

[aethersdr-agent]

Cisco CodeGuard — static analysis of this PR (6 finding(s))

• [MEDIUM] CG-PATH-001 — Potential path traversal in src/core/TgxlConnection.h /tmp/aetherclaude/pr-6036/src/core/TgxlConnection.h:19
• [MEDIUM] CG-PATH-001 — Potential path traversal in src/gui/MainWindow.cpp /tmp/aetherclaude/pr-6036/src/gui/MainWindow.cpp:9549
• [MEDIUM] CG-PATH-001 — Potential path traversal in src/gui/MainWindow.cpp /tmp/aetherclaude/pr-6036/src/gui/MainWindow.cpp:9661
• [MEDIUM] CG-PATH-001 — Potential path traversal in src/gui/MainWindow.cpp /tmp/aetherclaude/pr-6036/src/gui/MainWindow.cpp:9780
• [MEDIUM] CG-PATH-001 — Potential path traversal in src/gui/MainWindow.cpp /tmp/aetherclaude/pr-6036/src/gui/MainWindow.cpp:9781
• [MEDIUM] CG-PATH-001 — Potential path traversal in src/gui/MainWindow.cpp /tmp/aetherclaude/pr-6036/src/gui/MainWindow.cpp:9782

_{Automated static scan by Cisco DefenseClaw CodeGuard on the changed files. Advisory — some may be false positives; the review above verifies them.}

---

_{🤖 aethersdr-agent · cost: $12.1437 · model: claude-opus-5}

[aethersdr-agent]

Antares Detector — candidate vulnerable file(s)

• src/core/PgxlConnection.cpp

The TcpSocket connection reads all incoming data via m_socket.readAll() without size limiting, allowing untrusted external input to force large memory allocations and potential DoS or resource exhaustion.

_{Localized by Cisco Foundation AI Antares-1B running locally in the AetherClaude sandbox, seeded by the Cartographer security map. Advisory only — please verify before acting.}

---

_{🤖 aethersdr-agent · cost: $12.1437 · model: claude-opus-5}

[w5jwp]

Thanks for the review. Pushed c30e153e, which merges #6033's branch (no force-push). The fixes are all in b12f8f51.

Blocker 1: the telemetry grid has an objectName. The test measures the widest reading in the polished m_tempBtn font, forces a proportional family on the temperature buttons, drives 9.9 → 106.8, and asserts columnMinimumWidth(0) is at least that width plus the 4 px border. With m_telemetryGrid->setColumnMinimumWidth(0, widest + 4) commented out, amp_applet_test fails twice; restored, it passes.

Nits (all in b12f8f51)

• N1: m_operateBtn->setAccessibleDescription(...) while disabled. tools/check_a11y.py src/gui/AmpApplet.cpp reports no issues.
• N2: rewrote the stale comments (AmpApplet.h, the two AmpApplet.cpp spots, pgxl_panel_test.cpp) to say what is true now and dropped the history narration.
• N3: the caption check compares against the SC_ComboBoxEditField rect.
• N4: placeholder-tone assertions check the sheet contains the resolved disabled colour (and a live reading does not).
• N5: Vac/Vdd say "not reported" before the first reading, and NameChanged is sent only when the name changes.
• N6: placeReadings() skips re-flow when the presentation is unchanged.

M2 mitigation: with no direct connection, the MEffA and fan descriptions say they need a direct PGXL connection. M1 mitigation (justification): only the numeric field is right-justified (label + rightJustified(value, 5) + unit), so neither the unit nor the number's right edge moves. Cited #5983 in the body. Left for your ruling: the expanded-panel grid, the hover/justification direction, and the before-report rule; I did not change them.

Heads-up: pgxl_panel_test fails at :553-554 (source->height() == vdd->height()) in my environment, where theme resources fail to load. It fails the same way at the previous head f5eafa06, so it predates this change, and it passed for you. The other focused targets pass; I could not drive the panel on the bridge (no amplifier in demo).

[w5jwp]

@ten9876 the updates from your review are pushed (c30e153e). The itemized reply is above. Ready for another look when you have time.

[ten9876]

Approving at bb8a14c1.

Fixed (each one proved):

• Blocker 1, column-pin test couldn't observe the pin: b12f8f51. With setColumnMinimumWidth(0, widest + 4) removed, amp_applet_test fails "column 0 is pinned to the widest reading (0 < 64)" and "the voltages do not move as the temperatures change". Restored, it passes.
• N1: OPERATE's accessible description, b12f8f51. tools/check_a11y.py src/gui/AmpApplet.cpp reports 0 findings.
• N2: stale comments, b12f8f51 and e2b49b51; history narration is trimmed.
• N3, N4, N6: b12f8f51.
• N5: b12f8f51. In e2b49b51, the temperature readouts no longer post a manual NameChanged, because on Qt 6.11.2 setAccessibleName already posts it once per change (checked with a test harness). The fan pull-down now says the amplifier hasn't reported a mode when a direct connection is up.

Maintainer rulings: accepted as built:

• the expanded-panel 2×2 grid;
• the removed hover outline;
• the right-justified readouts;
• dim-not-hide before the amplifier reports;
• MEffA and fan dimmed, with "needs a direct PGXL connection", on relay-only stations.

Stack: bb8a14c1 merges main with #6033's head as the base. The squash differs from that head only by main's own commits, and this branch's net diff against main is exactly its 4-file increment.

Verified empirically: Linux build. 14 focused targets pass on this head, including pgxl_panel_test on loopback. The :553-554 failure the author saw doesn't reproduce here. The mutation is as listed. CI is green on this head, created after main's tip.

Not verified: the panel on the bridge, because the demo backend has no amplifier. Layout rests on the unit tests and the author's screenshots.