hamradiotools

Nexus 1.0.0

Released 4 August 2026. The current release is 1.24.0. Download it · Every version

Nexus has been a beta since its first public build, and this release closes the beta period. What that means here is narrow, and worth saying plainly: the modes, the rig control, the logbook and the awards engine have been run on the air through that period, on more than one station and on rigs the author does not own, and what came back from that is what the last several releases have been made of. It is not a claim that nothing is left — every entry below still says what a change does not do, and that habit does not stop at 1.0. Windows, Linux and Raspberry Pi build from the same tree and ship together, as they have every release. Upgrading asks nothing of you: 1.0.0 installs over 0.27.0 and reads your existing log, settings and layouts as they are.

0.28.0, 0.28.1, 0.28.2 and 0.29.0 were bench builds and were never published. Coming from 0.27.0, which is the last public release, this install brings you the 0.28.0 and 0.28.2 sections below as well as this one.

Fixed: two controls you could not reach — the Logbook's Log button and Program's export row

Both at 1024×768, the supported floor, and both with nothing to scroll:

  • The Logbook's manual entry form.

    Press Log QSO, fill it in, and at a UI zoom of 115 % or more the Log button that commits it was off the bottom of the screen. No scrollbar, nothing to drag. The header above the form and the form itself are both fixed height, so the overflow had nowhere to go.

  • Program's whole delivery row

    — Export for CHIRP…, Export CSV, Save to Memory Bank, Clear. At 120 % and up all four were off the bottom, sitting under two lists that each hold a floor of their own. That is the worse of the two: building a channel list and getting it out to the radio is the section, so the step you could not take was the last one. One cause, and it is three rules deep. The view's wrapper carries a scrollbar, and the panel inside it is told to be exactly the wrapper's height — so the wrapper's contents can never be taller than the wrapper, and that scrollbar can never appear. Not because there was nothing to scroll: by construction. The panel then clipped whatever its own column could not shrink, which is how a button ends up painted nowhere. Fixed at the rule rather than at the two symptoms. The panel keeps its exact height, because that is what bounds the scrolling lists inside it — letting it grow instead would break every view that works today: Settings would become one endless page, and the Logbook would stop virtualising and mount every row in your log. The panel owns its own overflow instead. Five views had each escaped that rule privately, one at a time, which is the tell that the rule was wrong rather than the views; POTA's private escape is dead now and is deleted with the cause. This is a scrollbar, not more room. At 1024×768 with a large zoom the open form is still taller than the pane holding it — what changed is that you reach the button by scrolling instead of not at all. Nothing moves at any size where it already fit.

Fixed: Phone and CW stop cutting controls off at the window Nexus opens at

Reported as Phone and CW looking bad at low resolution, and the diagnosis is upside down from the obvious one. At 1024×768 the panes run in a single column and that column has always scrolled — the floor was the safe case. The clip bit at the default window, where the panes run two columns and the region holding them does not scroll.

A column of control strips is a stack of things that cannot shrink, standing against an edge that cuts, with nothing in between. Phone's leading column is control strips end to end, so the NR slider, the AGC chips, the DSP toggles and the voice keyer's F4–F6 rendered past that edge — not below the fold, gone, with no scrollbar and nothing to drag. The column scrolls now. In the single-column layout the rule is inert, because a column there is exactly as tall as what is in it.

Two more, behind the same door:

  • Every pane that fills had a floor of zero.

    CW's six aux strips cannot shrink, so they took their height out of the one pane that can — the decode transcript, the pane the cockpit exists for — and it could be starved toward nothing. There is a floor under it now, and it is written to yield: stated in text units rather than pixels, and never claiming more than the space it is in.

  • The column count was being read as a claim about width.

    Untick CW's decode, sent echo and aux panes on a 3440-wide window and the log pane went content-height with the whole rest of the window left blank — because the ⊞ Panels menu could drive the region to one column at any width, and one column meant content-height rows. The track count and the width claim are separate things now, and it is the width that decides the vertical behaviour. CW's leading column is no longer drawn when there would be nothing in it. One column used to mean content-height rows, so an empty one was merely cheap; with the fill behaviour above it would have been a strip of dead space beside the log. All of this is layout. No pane changes parents, so a half-typed log entry and a recording in progress survive every flip between one, two and three columns. PTT, Stop TX and Tune are not in the pane region and are untouched.

Changed: the CW decode window is the pane the CW cockpit gives room to

The decode transcript is the pane a CW operator actually works from, and it is declared the primary one — but at the window Nexus opens at it was showing three lines. The reason is arithmetic rather than a bug: at that width the cockpit runs two columns, so the Sent echo, the scope controls, the DSP toggles, the RX DSP levels, Band Activity and the Copilot all share the decode's column, and none of them can shrink. The transcript is the only pane that can, so it paid for all of them and sat on its floor while the column scrolled 191 px past the bottom.

Four boxes that said what the pane frame already says once are gone, and no type was made smaller — no font size, no density setting, no UI zoom:

  • The Sent echo's own header, one label reading "SENT ▲" two lines under a frame head reading SENT.

    The blue stripe down its left edge is what tells you at a glance that these are your own transmissions, and that stays. Deleting the header is what let the strip come down from six lines' worth of reserved height to four, showing the same number of lines of text as before.

  • The Decode pane's own header row.

    Its first word was "DECODE", under a frame head reading DECODE. Everything else on that row — the AI badge, the words-per-minute readout, the AI decoder switch, its status and Clear — moved up into the frame's own title bar, which had been rendering empty on every pane in this cockpit. Same controls, same order, one less row between you and the text.

  • Two of the three rig-control panes.

    Scope controls, DSP toggles and RX DSP levels were three separate framed cards — three title bars, three sets of padding, three borders and three gaps — wrapped around one row of buttons each. They are now one Rig controls card holding one wrapping strip, and at the default width they sit on a single row. The ⊞ Panels menu is unchanged: all three are still separate entries, each still switching its own group on and off, each still explaining itself when your radio cannot feed it. Result at the default window: the unshrinkable stack in the decode's column falls from 513 px to 406 px, what is left to scroll past falls from 191 px to 84 px — one short drag instead of two screens — and the transcript goes from three lines to four. On a wide window, where the aux panes get a column of their own, the decode keeps the whole gain. In the header, the four keyer back-end buttons — CAT, Serial, WinKeyer, Soundcard — become one dropdown, about 110 px wide where they stood nearly 300. That header also carries the band picker, the tuning strip, Tune, Stop TX, speed, pitch, macros, filter width, memories and the rotator, and on a 1024-wide screen its width is what wraps it onto extra rows. Every word of the four explanations survives: each is on its own menu entry, and the one for the back-end you are actually keying with is on the dropdown itself — including the soundcard warning about routing audio to the rig and keeping drive below ALC, which is the one that matters on the air. Stop TX, Tune and Esc are untouched: no control moved into a pane that the ⊞ menu can hide, and none changed what it stops. Not done, deliberately. Band Activity and the Copilot stay in the decode's column. Moving them next to the log form is the only change that would close the remaining 84 px, and the log column has no room to give — its own Log button sits 379 px inside a 392 px pane, so anything added above it puts the button back below the fold, which is the defect the previous change was written to fix.

Changed: the Phone cockpit's control column stops scrolling its own frame

At low resolution the left-hand column of the Phone cockpit — Band Activity, the voice keyer and the rig-scope / DSP / RX-DSP strips — stands taller than the space it has, so the DSP buttons sit below the fold and you scroll a control column in the middle of a contact. Nothing was unreachable; it was simply further away than it needed to be.

Two of those panes were drawing a second card inside the pane's card, and one was printing a paragraph that no longer needed a line of its own. Both are gone: the band strip's own border, padding and margin, and the voice keyer's. That is 60 px, plus 41 px for the keyer's note, out of a column that was over-full by 237 px at the window Nexus opens at — a bit over 40 % of the overflow, recovered without moving a single control and without making any type smaller.

The band strip also stops printing "Band activity" one line under a pane head that already reads BAND ACTIVITY. The live spot count beside it is not the same thing and stays. This lands in the CW cockpit too, which frames the same strip.

What the keyer's note said is not chrome and has not been dropped: ● records from your input device — often the rig's RX audio, not a mic is the sentence that stops a slot going on the air with the wrong audio in it. It now rides both controls that start a recording — the ● button and an empty slot itself — so it reaches you where you are standing when it matters.

In the header, three words that repeated the control beside them are gone: Record QSO next to the ● that records, RX next to a meter already announced as "RX audio level", and Colors next to a palette picker already announced as "Waterfall color palette". Every one of them keeps its name to a screen reader; what went was the printed duplicate, and with it about 120 px of a header row that wraps at small windows.

Changed: the log strip commits above the fold

Reported at low resolution: "the Log button ends up below the fold with a half-typed QSO above it." Measured at the window Nexus opens at (1200×720), the strip stood 422 px inside a 392 px log pane — the Log button 30 px past the bottom edge, so committing a contact meant scrolling to find the button that commits it.

Two boxes above it were saying what the pane already said. The pane's own head reads LOG and is its name to a screen reader; the strip printed "Log this QSO" again two lines below it. And the strip drew its own bordered, padded card inside the pane's card. Both are gone from the Phone and CW cockpits: 43 px off the top of the commit row, 56 px off the strip's height, and the button is inside the pane at every window from 1024×768 up. No type is smaller — the space came out of duplication, not out of the font.

The Satellites section keeps its heading. It hosts the same strip with no title above it, so there the heading is the only thing naming the surface — this is a per-host decision, not a deletion.

Also in the Phone cockpit: the hint line under PUSH TO TALK is gone from the pinned transmit dock, where it held a full line at every window size and repeated the button's own tooltip. The part that was not a repeat — you talk on the rig's mic — is now on the button itself.

Fixed: a cockpit header can no longer be crushed under its own controls

The header every cockpit shares carried a hard 44-pixel floor. A floor written that way forfeits the automatic protection that stops a flex child shrinking below its own contents, so the header was the one child of a cockpit column allowed to be squeezed smaller than what it holds.

That shipped once, and it is worth naming. At 1366×768 the CW header wrapped onto extra rows, was squeezed back to 44 px, and drew those rows outside its own border box — where the scope panel below it, which is opaque, painted straight over them. What went under the paint was the keyer's Speed slider and Tune / Stop TX / CAT. A control that stops a transmission, under opaque paint, on a screen that looked fine.

The repair at the time was a hand-maintained list of the four cockpits it had been seen in — CW, Phone, RTTY and SSTV. That made the protection opt-in per cockpit, could not reach a grid-based host at all, and left the Operating screen out; Operate survived on an accident of what it happens to render today, which was asserted nowhere.

The 44 px is a real requirement — it is what keeps the header the same height across modes, so it does not jog when you switch — and it stays. It moved onto the identity block the header always draws, so it arrives as content: it raises the header's height instead of overriding the header's minimum. The protection is back in every cockpit, including one nobody has written yet, and the allowlist is deleted along with the reason for it. Nothing moves on screen, and Stop TX and Tune stay where they are, outside every removable pane.

Fixed: the APRS header stops cutting off its own controls

APRS was the one full view in Nexus with no way to scroll an overflow at all, and the reason nothing caught it is that nothing ever checked — the census that computes this for the other cockpits did not have APRS in it. Both are fixed.

The APRS header is a control strip that wraps: frequency picker, Re-tune, the TX arm latch, Monitor, both health chips and the internet control. Narrow the window, or raise UI zoom, and it grows onto more rows and cannot shrink below them. It sat directly against a hard edge — the only other thing in the view absorbs none of an overflow — so everything the strip could not fit was cut off with no scrollbar anywhere in the chain. The internet-feed panel is inside that same edge: at 1024×768 under a pinned zoom the header wraps deep enough that the panel's Watched calls row and the note under it fell past the bottom of the view, with no way to reach either.

The view scrolls vertically now. Sideways it still does not, on purpose — the station table scrolls itself — and the view keeps the exact height the map canvas is drawn against, so the map does not start growing again.

APRS has no removable panes, so nothing here was ever hidden by a ⊞ menu. The arm latch that holds the transmit queue was simply out of reach at those sizes, and is not now.

Fixed: the memory strip stops clipping its chips

The favourites strip in the Phone, CW and Operating headers carried a hard 26-pixel ceiling with its vertical axis clipped — no scrollbar, no recovery. What sits under that ceiling is written in your type, not in pixels: the chips and the + and = buttons are sized against the browser's root font, and the strip reserves a horizontal scrollbar under them. At the ordinary 16-pixel root font one row already comes to about 29 px against a 26 px ceiling, so the bottom of every chip and of both buttons was outside the clip before any UI zoom, any font bump, or any theme with a thicker border — on an axis that cuts rather than scrolls. There was no way to reach it.

Re-tuning the number is not the fix. A pixel ceiling over type-sized content is wrong in kind: the two sides are denominated in different units and move on different inputs, so no constant makes the relationship hold.

The ceiling is deleted, and the thing it was written to prevent — the old memory list growing the top bar with every save — is held by the shape that already holds it. The strip is exactly one row however many favourites you keep, bounded sideways, and scrolling horizontally past that; scrolling surplus chips sideways is the right behaviour for a chip strip and is what makes one row enough. The header does not get taller either: its cross-mode height floor is 44 px, well above the strip's row.

Fixed: opening Bonuses no longer flattens the Field Day sections board

The Field Day column is the banner, the header, the operator row, the score tiles, the sections board, the Bonuses list and the log — and every one of them is a fixed height except the board. So the board was the only thing that could give, and it gave for all of them. Open the fifteen-row Bonuses list at 1200×750 and the board a club watches all weekend went to a blank strip, with the growth past that off the bottom of the log and no scrollbar in the chain.

The cap on the Bonuses list was written as the fix for exactly this, and carried a comment saying so. It could never have been: the cap is a fraction of the same window the space is coming out of, so it bounds the list's own height and says nothing whatever about that height displacing its neighbours. The cap stays, and its comment now says what it does — it gives you all fifteen rows inside a list of its own, instead of about 290 px of checkboxes between you and the log.

Two rules do the actual work, and neither is any use alone. The board carries a floor, on the box the outer column actually sizes — the old floor sat one level in, where the column could not see it, and could not shrink, so a squeezed board just painted through the section beneath it. And the column scrolls past that floor, because a floor with nothing above it turns a crush into a clip, which is worse.

Fixed: the scope splitter has travel at both ends, and RTTY's waterfall yields on a short window

Three faces of one defect: pixel limits written onto strips whose height is a share of the window.

  • The scope splitter's drag was dead at both ends.

    Phone and CW handed the drag a minimum of 100 and 90 pixels against a stylesheet floor of 112, and both handed it a maximum of 420 against a ceiling that is 45 % of the effective window — 346 px on a 768-tall screen, 648 px on a 1440-tall one. So at both ends the pointer moved, the number moved, and the panel did not. The drag takes its limits from the same place the panel takes its size now, so the whole travel is live. Nothing renders at a different height on a 768-tall window; only the ends of the drag become real.

  • RTTY's waterfall had a 120-pixel floor and no way out of it.

    RTTY is the one scope in Nexus with no splitter for you to drag, so that floor was unrecoverable: on a 768-tall window at 175 % UI zoom the effective viewport is 439 px and at 200 % it is 384, and a large part of that was a floor you could not move. It yields now, and never claims more than 28 % of the window. Its 220-pixel ceiling went with it — that one bit only above a roughly 1000-pixel window, where it froze a tall display below the share every other size already gets.

  • The Phone scope panel's base rule carried a dead floor and a dead ceiling.

    Both shipped hosts override them, so nothing was wrong on screen. But a dead rule holding a pixel floor and a pixel ceiling is a loaded gun: a third host rendering that strip would have been silently pinned between 120 and 220 px at every window and every zoom. Size belongs to the host now, and the base rule keeps chrome only. RTTY's shorter floor only frees room above its transmit dock. Nothing else on that screen moves.

Every version