[decay71]

Fix for the temperature detour reported in #172 (nozzle wipe at 250 after a background swap), ported from dev.

• Background swap: the engine no longer switches a head's heater to 0 when its op ends. It hands the heater back to the newest request from the slicer since it took the heater (for example the M109 for the next tool), else to the target it found, else 0. Once the print has ended it is always 0.
• Swap: a head that arrives at a target below min extrude no longer gets that target back while printing. It gets the last print temperature a swap saw on that head, else the swap temperature. The post-load wipe uses the same value, so it stays at the print temperature instead of jumping to the database load temperature (250 for Generic PLA).

Tests: tests/unit/test_bg_heat_intent.py (observer and claim/release on stubs).

[docx-tech]

@decay71 I can confirm that the temperature is no longer spiking to 250C after the IDLE action when a head is picked up, and goes to the last print temperature.

Full BG Swap Blob Issue

I a full bg swap is completed (and its purging in the back) we get the blob in front of the prime tower again.

Blob
klippy (1).log

Heat Spike has been resolved

[decay71]

Thanks for the log, it shows the blob exactly.

Both cases (13:34 and 13:38 UTC) are the same sequence: the bg swap was still in LD_PRIME when the T arrived. Stock parks the old head, lifts the bg head out of the dock (pick extruder 13:34:26.2) and activates its extruder 2.4 s later (13:34:28.6). The engine's interlock only sees that activation, and only between two prime chunks, so it noticed the pick at 13:34:29.8, about 3.5 s after the nozzle had left the dock. In that window the prime kept running (roughly 10-15 mm at 4 mm/s) and the anti-ooze retract never happened, because the abort fired first. The head was carried to the tower with the purge still coming out, which is the blob in your photo. The pick-check wipe and the cushion only come 8 s later at the discard position. It has nothing to do with the heater hand-back; those lines are all correct in your log (standby 70, print 220).

Fix on branch fix/bg-pick-notice (PR follows): ace.py wraps T0-T3 and announces the toolchange to the bg engine BEFORE the stock macro runs, then waits (up to 8 s) for the handover. The engine stops at its next move boundary while the head is still docked: the prime ends cleanly, sets its anti-ooze cushion and records the missing purge, which the arrival pick-check tops up as before. The other polling phases take their existing pick-abort path, just earlier; the bulk retract is untouched. Log tells: toolchange announced during LD_PRIME, prime stopped at N/153 mm for the announced toolchange - cushion set, background op handed over before the toolchange.

For #186 this is the same spot: ACE_BG_WAIT already sits before the T. In LD_GRIP/LD_PRIME it should request the abort and wait for the handover instead of passing through, otherwise the blob stays with the wait feature. The new request_pick_abort(head) in the engine is exactly that call, you can use it from ACE_BG_WAIT once the branches are merged.