Version 3.7.0-rc.1

Compatible versions:

• RepRapFirmware 3.7.0-rc.1
• DuetWebControl 3.7.0-rc.1

Upgrade notes:

• DuetControlServer reserves 256 MiB of memory via MemoryMin now, which requires the memory cgroup controller. New DuetPi images enable it, on an existing installation add cgroup_enable=memory to cmdline.txt, else the setting stays without effect

New features:

• Job and macro file errors report the filename instead of the channel, matching RRF, and the filenames are derived from the physical path
• Object height is detected from max_z_height (OrcaSlicer), MAXZ (Cura) and Height (Fusion 360) comments via the new ObjectHeightFilters setting, like RepRapFirmware does
• Added LAYER_COUNT (Cura) and total layer number (OrcaSlicer) to NumLayersFilters

Object model changes:

• Added move.minSpeed
• Added boards[].accelerometer.samplingRate and boards[].accelerometer.resolution

Changed behaviour:

• Meta command gating uses a per-channel executing-state shadow instead of the lagging inputs[].active checks, so no full model sync is needed after M596
• M598 from a file channel makes the file readers rendezvous like the firmware streams do
• Local variables of closed code blocks are deleted only when they are redeclared or the file is closed, so a pause can rewind into their scope
• DuetHttpClient URL-encodes every query parameter and sends virtual paths the way DWC does
• Increased FileInfoReadLimitFooter from 256 KiB to 400 KiB to match RepRapFirmware
• Codes that exceed the transfer buffer report a proper error instead of a generic failure
• Aborted macro files and file abort requests are logged at Debug level on the Daemon channel

Bug fixes:

• Matched file sync requests were never removed, so every following sync at the same file position deadlocked
• A resume arriving while a file task was still draining towards its pause was silently discarded
• A job was treated as complete while the firmware was still executing its moves, so a late pause left an unrecoverable state
• Result variables of M260.1/M261.1 did not mark their code block as having local variables, so the block was closed without waiting for codes that still referenced them
• Variables declared before any code block was opened were only logged as a warning and never tracked
• A failed M260.1/M261.1 still registered its result variable, so deleting it later reported an unknown variable
• Uploaded file names lost plus signs and percent escapes on the rr_ endpoints because query values were decoded twice (#159)
• InputShapingType.EI2 and EI3 serialized as eI2 and eI3
• Jobs pages whose path contains a dot could not be reloaded
• Resuming a job in Fanuc or LaserWeb style G-code lost the modal G0/G1/G2/G3 command, so the first line without a command letter was not recognised
• Object height of files whose last layer is bigger than the footer read limit was not detected (forum 39435)
• M99 in SBC mode could cause a spurious out-of-order reply warning (forum 39434)