search

mozilla / participation-metrics-org

Participation metrics planning repository
4 stars 4 forks source link

Come back with an estimation on how long it would take to improve the tools so we can filter Firefox-only code from gecko-dev repo #100

Closed nukeador closed 7 years ago

nukeador commented 7 years ago

Come back with an estimation on how long it would take to improve the tools so we can filter Firefox-only code from gecko-dev repo

jgbarah commented 7 years ago

Can we assume everything in the gecko-dev repo is project Gecko, except for directory browser, which is project Firefox? In particular, I was reading this comment, by Myk Melez, which suggests that maybe for project Firefox we should have some more directories into account:

As noted in the description for the Firefox module, the primary source directory for "just Firefox" code in gecko-dev is browser/. Almost everything else is a part of Gecko, although the toolkit/ and chrome/ directories are a part of the Toolkit module, which is not entirely independent of Firefox (and shares a module owner and peers).

nukeador commented 7 years ago

For the toolkit/ and chrome/ forlders we don't a good way to know which is Firefox-only and which is Gecko-only so we are assuming it's part of Gecko contributions.

@mykmelez we didn't get any other tips to separate the code there right?

mykmelez commented 7 years ago

For the toolkit/ and chrome/ forlders we don't a good way to know which is Firefox-only and which is Gecko-only so we are assuming it's part of Gecko contributions.

In the absence of more precise distinctions between the various directories in toolkit/, I would actually count toolkit/ as Firefox, because its module's peers are identical to the Firefox module's peers, and its code is more "frontendy" than "backendy".

However, I still think @Mossop is the better person to answer this question, as he is the module owner for both modules.

@mykmelez we didn't get any other tips to separate the code there right?

Correct, although here are a couple of possible heuristics:

jgbarah commented 7 years ago

@mykmelez we didn't get any other tips to separate the code there right?

Correct, although here are a couple of possible heuristics:

  • count directories which contain predominantly C++ code as Gecko, the rest as Firefox.
  • count directories whose primary contributors' primary contributions outside toolkit/ are to Gecko code as Gecko, the rest as Firefox.

Both heuristics are tricky in the timeframe we hace, so I will stick to using directories. We can revisit this later if you're interested.

So far, I understand then that browser, toolkit and chrome directories are project Firefox, the rest is Gecko.

Any hint on what to do with commits touching files in both parts? (Firefox and Gecko). For now, I'm planning to split in like two commits, one for each "part".

jgbarah commented 7 years ago

Finally, decided on:

Results:

Screenshot with commits per month during the last 10 years. Ignore since Jan 2017, please. Does the rest of the evolution make sense?

selection_207

alpgarcia commented 7 years ago

I just updated Community notebook to use new data: Community metrics updated applying this: https://github.com/Bitergia/mozilla-contribution-analysis/blob/master/data-analysis/rc1/html/Community.html

nukeador commented 7 years ago

I'll check about the numbers, let's assume for now these are correct.

nukeador commented 7 years ago

From yesterday's meeting, this is ready and implemented.

alpgarcia commented 7 years ago

After checking data we realized some commits associated to Firefox are also present (and associated to Gecko) in gecko-projects and universal-search-gecko-dev repos.

We decided after talking with @nukeador to remove both repos from our indexes as probably they are only introducing noise to the analysis.