535671 - refactor get_locales_from_json

Status: RESOLVED WONTFIX

(Release Engineering :: General, defect, P3)

Description

[Aki Sasaki (not active)]

and l10n-changesets*.json

Pike was thinking of having something like

{
changesets: { locale1: {changeset}, ...},
platforms: {locale1: ["+maemo-multilocale"], "ja": ["-mac"], "ja-JP-mac": ["mac"], ...}
}

with the default list of platforms specified elsewhere.

Also, I was supposed to remove the FIXME bit in the get_locales_from_json and forgot.

Comment 1

[Chris Cooper [:coop] (he/him)]

Mass move of bugs from Release Engineering:Future -> Release Engineering. See
http://coop.deadsquid.com/2010/02/kiss-the-future-goodbye/ for more details.

Comment 2

[Aki Sasaki (not active)]

Axel: I'm looking at this and I'm not sure what the immediate benefit is from separating these out... to me it seems like a 6 of one/half a dozen of the other type bug.

(I can see how having the platforms set this way helps you in effect ignore them, but I can also see how it might be harder to "read" which platforms are in a locale when platforms or locales are added.)

Do you still need this? What benefits does this provide us?

Comment 3

[Axel Hecht [:Pike]]

Right now, we're adding android as platform, right? For those locales that are not special, it'd be nice to just not see any change. The file might be smaller in total size afterwards, too.

Also, the two pieces of information feel perpendicular to me, which revision do we ship, and, on which builds is this locale showing up. And the latter just feels like "we do this, but for ... we do ...,". It just feels right that way.

Comment 4

[Aki Sasaki (not active)]

(In reply to comment #3)
> Right now, we're adding android as platform, right? For those locales that are
> not special, it'd be nice to just not see any change. The file might be smaller
> in total size afterwards, too.

Right, though the proposed format has the side effect of needing to update/check two locations to see what platforms are being built, rather than one.

> Also, the two pieces of information feel perpendicular to me, which revision do
> we ship, and, on which builds is this locale showing up. And the latter just
> feels like "we do this, but for ... we do ...,". It just feels right that way.

I thought that combining platform and changeset information was the whole point of this file?
Can you think of situations we've run into where the proposed format would improve things?

Comment 5

[Aki Sasaki (not active)]

Kind of hoping bug 609265 means we can WONTFIX this.

Comment 6

[Aki Sasaki (not active)]

We now have Django code to generate this file from bug 609265.

We also have scripts/configs that depend on this file in both buildbot and mozharness; specifying default platforms elsewhere adds another layer that both codebases would have to look at.

I showed this bug to a few other RelEng folks, who feel the layout of the file is best as is.

I'm going to close as WONTFIX.  If there are strong reasons that need to be discussed, let's discuss.