The QR decoder has a single item that is 8908 bytes big and placed
on the stack. (struct datastream)
Without enlarging the stack, this will reliably crash.
* Removes VectorShape from user python interactions
* Re-integrates vectorio with displayio behind draw protocol implementations
* Implements draw protocol with VectorShape
* Composes VectorShape behaviors into Rectangle, Circle and Polygon
* Fixes terrible pixel garbage being left behind
* Improves redraw performance (heuristically) by tracking dirty area separately from current area.
Known Issues:
It does not work with transposed views.
Call `supervisor.disable_ble_workflow()` and the BLE workflow will
be disabled until the chip is reset.
This also includes a couple fixes:
1. Terminals can now be deinit by setting the tilegrid to NULL. This
prevents using the tilegrid before display is init.
2. Fix BLE serial send amount when sending more than a single packet.
Fixes#5049
A pointer to the FIL structure was kept after it went out of scope.
Apparently this happened to work until cf97793 added the `result`
variable that clobbered it.
Fixes#5062
This is a breaking change with previous palette semantic with respect to python code that uses vectorio.
Displayio has breaking changes in cpy 7 for Group's removal of max_size parameter so this is as good a
time as any to break everything.
Currently:
To color vectorio shapes correctly you have to pass in a palette with length 2. Palette[0] must be set transparent and palette[1] must be the color you want.
New:
To color vectorio shapes correctly you pass in a palette with length >= 1. Palette[0] will be the color of the shape.
Also improves pixels per second when skipping areas that aren't covered by the shape.
On ESP ctrl-c during fake sleep will now stop the sleep. A crash
on real deep sleep is now fixed as well. (Exception string saving
was crashing on reading the deep sleep exception.) Fixes#4010
This also fixes nRF fake sleep after the first time. The internal
variable wasn't being reset early enough. Fixes#4869
@dhalbert noticed that if a fork was called `micropython`, the
existing check could fail during CI because it will be cloned into a
directory also called `micropython`.
Instead of hardcoding a range of strings that are OK as top directories,
find the location of the top directory relative to the script.
This will be conflicty after #5069 is merged, but it should be easy to
fix.
