prestonbrown/helixscreen v1.0.1

on GitHub

v1.0.1

The first patch release on the 1.0 line. The bulk of it is in five areas: which printer
HelixScreen thinks it is talking to, what the pre-print screen claims the printer is
doing, whether filament system edits survive being written, where a bed mesh actually
gets saved, and the install and restore paths on K2 and ForgeX machines.

Added

• The pre-print screen follows printers that narrate through the display - a
  PRINT_START that reports its phases with SET_DISPLAY_TEXT or M117 was previously
  invisible, because commands inside a macro never echo to the console. Those printers
  now get a tracked pre-print sequence like any other.
• An aux fan role - a printer declaring an auxiliary fan gets it named and mapped,
  and a preset pointing at it resolves instead of silently doing nothing.

Fixed

Printer identification

• A Centauri Carbon is no longer identified as a Qidi - a machine running COSMOS
  matched on a shared command and came up as the wrong printer at 89% confidence, which
  meant the wrong profile for everything downstream. It now identifies exactly.

Starting a print

• Heating estimates are per-printer and hold across prints - the pre-print screen
  learns each machine's real climb rate rather than a shared guess, saves the whole climb
  rather than the part before a hold, and keeps what it learned when a timeout ends the
  pre-print early.
• The pre-print screen waits for heaters to actually reach target - it no longer ends
  on a timeout while a heater is still climbing, and it treats a heater swinging around
  its target as settled rather than still climbing.
• Centauri Carbon pre-print phases are tracked - the COSMOS PRINT_START sequence
  matched nothing, leaving every phase on a single default estimate. Its phases now map.
• A heat soak holds the screen open for the time it announces - a COSMOS heat soak on
  the Centauri Carbon no longer looks finished minutes before it is.
• A heating bed is labelled as one - a heat soak relabels it correctly, and an idle
  mesh step gives way to the nozzle heat rather than sitting in front of it.
• A file that would trigger an emergency stop is refused before printing - a gcode
  file calling a command the printer turns into an emergency stop is blocked up front.
• A file scanned too early is scanned again - a file checked before the printer's
  macros had been read was cleared on incomplete information and stayed cleared for the
  rest of the session. That answer is now reused only while it still applies.
• K2 phase narration is the printer's, not ours - our own pre-start echo stopped
  claiming three phases the printer had not reached.
• A paused print stops advancing - progress no longer creeps forward, and it is no
  longer recorded internally either, whatever the printer reports.
• An active print keeps its status screen - closing the screen during a print no
  longer discards it, and a job that ends while the screen is hidden releases cleanly.
• A filament runout that stops the print is reported - the removal notice also waits
  until a tool change would have refilled, so it is not raised prematurely.
• The 2D preview appears complete - it is built off screen and swapped in, rather than
  drawing itself in front of you.

Filament systems

• AFC and Happy Hare edits survive - a slot's colour and temperatures are no longer
  dropped when the override is written, clearing a slot erases it from disk, and Happy
  Hare's override store is loaded at startup instead of being built empty.
• Black is a real colour - a slot set to pure black is dispatched to AFC and Happy
  Hare rather than read as unset (#1597), and the spool editor treats it the same way
  (#1608).
• A spool weight edit changes only the weight (#1652) - saving a meter reading no
  longer rewrites the rest of the slot with it.
• Filament loading respects temperature limits - a macro prefill stays above the
  extrusion minimum and within the hotend's maximum, is handed the nozzle temperature the
  panel is showing, and on a multi-tool machine uses the extruder the loading slot feeds
  rather than whichever one is active.
• Homing consent is spent once and returned if unused - a confirmed home before a
  macro load homes first, a raw extrude never asks, and a failed dispatch releases the
  consent it did not use.
• An unavailable remap explains itself - the filament card no longer offers a remap it
  would refuse, dims the chips rather than the whole card, and moves its refusal into a
  help modal that offers the fix. A tool changer with no ASSIGN_TOOL refuses outright.
• The file-rewriting warning appears only when the file is rewritten - printers that
  route filament natively were warned that the job file would be modified before
  printing, which was never true for them.
• Lane presence comes from the sensors - silk sensors own whether a lane is loaded,
  rather than inferring it from a Spoolman weight poll, and a single-position system is
  no longer clamped to a one-slot span (#957).
• Snapmaker U1 with a PAXX AFC-Lite - the Snapmaker filament system takes precedence
  instead of the two fighting over the machine.
• CFS error messages match Creality's own table - and the long CFS verdict message no
  longer overflows its box in any language (#1605).

Bed mesh and calibration

• A mesh is saved under the name you chose - calibration reported success under the
  chosen name while storing the mesh in default, and replacing another stored profile
  now asks first.
• Profile names with spaces work everywhere - names are quoted on every path they are
  sent, and a name containing a semicolon is refused rather than silently breaking.
• A mesh is named before it is probed - so it lands under the name printing will look
  for, rather than somewhere printing cannot find it.
• The Centauri Carbon uses its own bed mesh macro - COSMOS machines delegate to the
  firmware's macro and count their mesh points, rather than reporting a mesh that was
  never taken.
• Your own mesh macro is not failed for an optional command - only a shipped
  calibration sequence fails on a command the printer does not know. A wrapper macro of
  your own kept probing while being reported as failed.
• Centauri Carbon mesh calibration no longer waits for a hot bed to cool - and
  heaters the COSMOS sequence turned on are turned back off when it finishes.
• Dismissing the mesh naming dialog leaves Probe working - and the default profile is
  re-probed without asking again.
• The mesh naming dialog speaks every language - its save and replace messages were
  English-only regardless of the language set.

Chamber heaters

• A chamber heater or sensor counts only while the printer reports it - a stale
  assignment no longer presents itself as live, and a sensor named in configuration is
  re-resolved on each discovery.
• Filament and material temperatures read the resolved chamber heater - rather than a
  different one than the rest of the screen is showing.

Installing and restoring

• K2 stock UI restore verifies the boot link (#1641) - it offers a reboot only once
  the link is confirmed, reports an app that failed to start rather than claiming success,
  and kills the carve-out web server properly.
• The K2 web server carve-out survives restarts and reboots (#1641, #1665) - it runs
  as a supervised service, and deploying records it in the ledger with both boot links
  verified (#1667).
• An interrupted K2 or K1 install can be resumed (#1668) - the disabled-services
  ledger is found wherever the interrupted run left it, and the K1 sweep stops the web
  server with the rest of the stock backend.
• ForgeX machines keep the display mode they arrived on - a machine arriving in FEATHER
  mode kept drawing over HelixScreen indefinitely, and the network daemon is now restored
  at startup so first-run WiFi setup is reachable.
• Presets survive the wizard (#837) - post-wizard migration seeds role keys that were
  never set, a preset value is no longer lost during startup, and AD5M variants get the
  second filament sensor they declare.

Networking

• Hidden networks can be joined - both network backends now send what their tool
  requires for a hidden SSID.
• A network daemon that is installed but down is reported as such - rather than being
  read as absent and quietly replaced.

Screen and interface

• The panel goes dark at sleep even where the backlight works (#1594) - previously a
  machine with working backlight control kept showing the last frame.
• A backlight write the kernel rejects is reported (#1595) - instead of being treated
  as applied.
• Home edit mode gestures are classified against their own press point - so a drag
  started in one tile is not attributed to another.
• The first-run tour behaves - it starts once however many things ask for it, waits
  for the wizard to finish, and does not open on top of home edit mode.
• The WiFi password dialog cannot be left dangling (#1579) - anything else destroying
  it no longer leaves a stale reference behind.
• A 2D render is not refused over a saved Thumbnail Only preference - the two settings
  are now one decision instead of disagreeing.
• Chinese reads correctly on the slot reservation message - it used a character the
  bundled font does not carry, so it drew as a missing glyph.
• LED detection reads live configuration - and a synthetic strip is never reported
  missing.
• Print list drops Size and Modified on narrow screens - rather than crowding the
  file name off.

Stability and performance

• Several crashes on screens that rebuild themselves - event guards that had stopped
  applying were restored, and an observer on a value that is never freed is now removed
  correctly.
• History no longer refetches on every timelapse frame - a timelapse component moving
  and rendering frames throughout a print cost a full history reload each time.
• Streaming a large gcode preview is budgeted honestly - a RAM-backed cache was
  charged nothing for the file it streamed, so the memory it used went uncounted.
• A memory warning that never clears escalates - instead of sitting at warning
  indefinitely.
• The Centauri Carbon caches to flash rather than RAM - and the cache reports whether
  it actually landed in RAM instead of assuming it did.
• Debug bundles find logs and crash files on every printer - on Creality, Flashforge
  and Centauri Carbon machines the bundle looked in one place that did not exist there, so
  an uploaded bundle could be missing the crash report entirely.

Changed

• Crash reports name real functions on statically linked printers - on the six
  machines built without position-independent code, the load address was added to symbol
  addresses that were already absolute, so backtraces pointed at the wrong code.
• Moonraker settings the environment file cannot change are no longer shipped - they
  only looked configurable.