I'm just hopeful that this new push on ML for the EOS-M will result in a reliable hack that I can use. As you know, I've been using the BMPCC and it is very reliable. The last time I was using ML it wouldn't turn audio back on for H.264. If I switched to photo mode without re-setting FPS the LV would get scrambled.
This is unlikely without proper
testing feedback and
bug reports. This section feels kinda empty:
https://bitbucket.org/hudson/magic-lantern/issues?version=EOS-MWelcome to
Linux Airlines. Unlike BMPCC, ML is a spare-time community project and nobody of us is developing full-time. Identifying the bugs is a very time-consuming process for us, and doesn't require any coding skills, so you can help a lot simply by documenting these issues properly.
Just so you get an idea about the complexity of software testing: imagine you have a menu with only 16 boolean items (just on/off). To cover all the combinations (e.g. check the interactions between various features and such), you would need only 65536 test cases. Increase the number of menu items to 32, and you need 4294967296 test cases to cover everything. Sure, in practice you can figure out what features are likely to interact, and only test those combinations.
These days, ML has hundreds of options in menu (if not one thousand). I'm pretty sure if somebody sits down and tries every single menu item (just one at a time, not combinations), he/she will find a bunch of bugs. It happens to me every time I get bored in the train and I start playing with the camera.