1503864 - Determine strategy for converting bytecode references to the bytecode interface

Status: RESOLVED WONTFIX

(Core :: JavaScript Engine, enhancement, P3)

Description

[Matthew Gaudet (he/him) [:mgaudet]]

Now that Bug 1499544 is landed, we have a skeleton of a bytecode interface upon which to build. 

The next important step is to figure out how to make forward progress on this. 

Kannan has previously suggested we build out something like 'Are We XBL still?' [1]. I'd be willing to take a stab at a basic form of this, though not nearly as advanced as the one in [1]. 

My preference I think would be to follow the strategy I used in Bug 1459038. In that bg I opened a series of dependent bugs, each one with a template that described in detail the background of the problem, the scope, and suggestions on how to make forward progress. These bugs were all mentored, and ultimately were all closed by contributors, a success in my mind. 

The nice thing about this strategy is that these bugs show up in Codetribute, which is the homepage for new contributors we have [2]. 

In this case, we would open all these bugs as dependents of Bug 1478034.

The difference here is one of scope: grepping for |jsbytecode*| through the code-base today we see 984 hits; as we're changing types, it's possible for a lot of these changes to be quite cross-cutting, so it's not clear what the best strategy is for opening the bugs. Perhaps we should spend some time triaging our highest-value targets? 

Any other ways we can help encourage this work to get done, and keep it tracked? 

[1]: https://bgrins.github.io/xbl-analysis/ 
[2]: https://codetribute.mozilla.org/

Comment 1

[Matthew Gaudet (he/him) [:mgaudet]]

Here's a first draft dashboard... https://mgaudet.github.io/AreWeBytecodeStill/ 

This is... not excellent code, but could easily be run as a cron job somewhere, and produce nicely updating charts over time.

I'm a little concerned the data growth could be an issue eventually... but I also think it could be a while. ¯\_(ツ)_/¯