Pebble application execution harness, aimed mostly at layout preview and basic functionality prototyping.
The name comes from the fact that claypellet is more malleable than a Pebble and also significatly less solid and robust.
This is very experimental at present. I have implemented just enough to a few watchface apps running. If you need any help getting it working or have suggestions for features, I generally hang out in #pebble on the freenode IRC network.
The code uses the MIT license. See the LICENSE file for details.
Most of the C code is pulled straight out of the Pebble SDK header files and modified slightly to build the requisite CFFI hooks. Some of the resource management code was reverse-engineered from the Pebble SDK resource packaging tools. Also, the core of this is an implementation of a certain amount of the Pebble application environment and API. Since the SDK apparently has no license but is distributed freely to anyone with an account on developer.getpebble.com, I'm assuming that it's okay to use it like this.
You'll need to have the pebble SDK installed. From there it gets a little hairy.
You'll need the current development branch of cffi -- a bug I discovered while getting this project going has been fixed there but not in the v0.6, which is the current release version at the time of writing.
You'll also need pygame which I'm using for the display.
With all the dependencies installed, you should be able to run the following commands in you project directory to get your app running in the simulator:
/path/to/claypebble/compile_pebble_app.sh
/path/to/claypebble/run_pebble_app.py
This assumes that you have a standard project layout with all the source files
in src/*.c
and built files in build/
.
If you're using OSX, you might need to drop a copy of pebble_app.h
in the
current directory and modify it to change ((section (".pbl_header")))
into
((section ("__DATA,.pbl_header")))
. (I had to do this, anyway. I haven't
tested on any other operating systems, so YMMV.)
You can call run_pebble_app.py -h
to get a list of parameters available. The
most exciting param (and the only one without a useful default) is --firmware
which allows you to specify a Pebble firmware file (perhaps acquired using some
fine tools on the internet)
to read system fonts from. Without this, text using system fonts will not be
rendered.
Once you have your app running in claypebble, you can push some buttons on your keyboard to make things happen.
esc
orq
to quit.r
to reload the app. You'll have to runcompile_pebble_app.sh
to recompile if you've made any changes.s
to take a screenshot. (Written to the current directory with a timestamped filename.)
There are a bunch of documented API functions still unimplemented:
- All animation-related APIs.
- A few graphics APIs, including paths, lines and pixels.
- All "hardware" APIs: backlight, buttons, vibration.
- Raw resource access.
- Some miscellaneous functions here and there not mentioned above.
All documented functions have commented C function signatures in harness.py
so they should be fairly easy to find. All functions declared in pebble_os.h
have callback hooks defined and stub implementations in the generated C and
Python code. I've implemented a few undocumented API functions because some
apps I tried to run used them.
System font support requires a separate download of a Pebble firmware bundle. There should probably be a fallback font (or set of fonts) for when this isn't available, but for now we just give up and don't draw text with fonts we can't find.
I haven't really decided how best to do button input, but binding to arrow keys seems like a good plan. This hasn't been a priority since I've only been testing with watchface apps and none of them use input of any kind.
Simulating communication seems like it'll be hard. Maybe someone else wants to give that a go?
This project is very incomplete. I plan to add more functionality in whatever direction strikes my fancy as and when I get around to it. Feel free to contribute if there's something you particularly want it to do.
While I think the basic approach is reasonable for first-pass prototyping stuff like layout tweaks, but claypellet will never be a replacement for actually testing your code on the device. The CPU architecture and libc implementations are just too different.