1283355 - is there potential for signature analysis of existing add-ons based on factors making them likely to break with e10s

Status: RESOLVED WORKSFORME

(Firefox :: Extension Compatibility, defect)

Description

[:shell escalante]

This is exploratory.  Starting conversation with billm to see if there are any markers or ways to predict when e10s non-friendly CPOWs come into play.  Common good or bad signatures that can be seen on AMO without waiting for Gecko to run to flag them.

Including Andreas (theone), Philipp (fallen), and Jorgev for their expertise in AMO.

aside: someone mentioned that there had been talk of running validation in a VM (idea from a different project) that might be an applicable solution, if dynamic analysis is the only option.

The end goal is to be able to make an intelligent predication on how the mass of add-ons that have not been marked either way for MPC flag will behave.  Is it mostly going to be OK with a few failures?  Is there a way to be certain of what will fail in advance, flag them as not-compliant, and just deploy to all add-ons except not-compliant?  Or does it look really uncertain - so we should only deploy to MCP OK flagged add-ons or webextension based ones.

Comment 1

[:shell escalante]

so not sure if andy or luca is going to own this - but looking through SDK for high level APIs issues with e10s and also known lower level ones that will break (ex: requiring chrome).

Idea is to be able to articulate that to expect or compare against AMO to identify issues we will have.

after talking to billm, andy, dzeber- there weren't any more markers beyond Shim and CPOW that we can get from telemetry.  making necessary adjustments to telemetry pull to get more actionable information from that.

Comment 2

[Andy McKay]

First step: Luca, Kris and I will run through the SDK APIs to find the areas that are most likely going to have problems.

Comment 3

[Andy McKay]

I feel we did comment 2. But I'm not sure where we are with e10s these days. Shell do we still think we need to do anything here, it kinda feels like the wider audience of add-ons for 51 is coming either way.