The issue is very apparent with a cold extruder. For example when starting a single color print with the MMU. I can hear grinding in the extruder gears at the very end of the sequence. This means the sequence is too long and we are
likely pushing the filament too far into the extruder.
The purpose of the test is to check if the filament is actually
present in the PTFE tube. My proposal is we reduce the length
of the test to only cover the *length* of the PTFE tube.
This will also reduce printing time a little bit.
Fixes issue where First Layer Cal. triggers an unload before any
filament is loaded
The purpose of this if() statement is to handle the case where
a user or developer is sending T-codes to the printer directly
via Serial. Such as when one first sends T0 and then T4.
And unload must be triggered in-between, we can do this automatically
if FINDA detect filament.
Change in memory:
Flash: -10 bytes
SRAM: 0 bytes
Use an unversioned root to store dependencies, so that they can be cached more easily.
Do not re-download archives if the build environment has already been set up.
Attempt to cache dependencies with the travis cache. This avoids
redownloading avr-gcc for each stage of the pipeline.
This /might/ improve the download failures we're seeing recently,
although the travis cache is also networked.
The printer shouldn't continue endlessly asking for a version if an old MMU is detected.
Also if the protocol is supposed to be turned off, it
shouldn't be sending S0 queries.
In M600 and M702, the Z-lift is taken care of outside this function
so it makes no sense to have a Z-lift within unload_filament()
Only exception to this is in the wizard, there I moved the Z-lift to
happen before we start heating the nozzle.
Change in memory:
Flash: -12 bytes
SRAM: 0 bytes
Ensure that fan checks are reset and use the new speed at each step of
the calibration.
This also gives extra time to the fan to ramp-up from a cold start,
when a fancheck could previously start right *after* the speed change.
Should fix#3791
The cmake build system is effectively on-par (and a lot better in most
areas) compared to the old build system.
There's no need to warn developers anymore.
The option SECONDARY_LANGUAGES is no longer being used.
cmake always allows en/multilang for all selected variants
by using the appropriate target now (ALL_ENGLISH).