Currently profiling on device requires roughly these steps: 1) Connect device via USB to computer 2) Run ./profile.sh start with a list of sometimes esoteric options 3) Perform actions to be profiled 4) Run ./profile.sh capture 5) Upload resulting file to cleopatra website While this is certainly doable and works well, the number of steps is a barrier to getting more developers to use the profiler. This bug is to investigate/discuss ideas for making profiling easier on device. One idea: 1) Developer setting that places a begin/end profile buttons on the screen. 2) User presses "begin" profile button. 3) User performs actions to be profiled. 4) User presses "end" profile button. 5) Device automatically uploads resulting files to cleopatra (or another pre-processing service). 6) Device displays, logs, or emails resulting cleopatra URL to user. This would require picking a set of profiler options that make sense most of the time. Any other ideas?

Its a good idea but keep in mind: - Requiring UI interaction means that we're going to sample the user interacting with the UI at the stop/end of the profile. We can ignore these easily in the analysis but it's not ideal. - We will need to translate addresses into symbol on the device or store the symbols in our symbol server. Otherwise our profile will have to omit any native addresses collected that aren't resolved.

A better option could be to hook up that with the remote debugging tools in fx desktop instead of having on-phone UI.

Firefox OS is not being worked on