1255407 - Create test to ensure that l10n bits are "proper" inside en-US

Status: RESOLVED INCOMPLETE

(Testing :: General, defect)

Description

[Axel Hecht [:Pike]]

I'm writing a test that checks if the en-US code is actually sane.

So far I found the two bugs 1255382, 1255404 where chrome.manifest gets told about one thing being an overlay of another, but not being (remotely) similar.

I also want to write something that'll make it easy to find un-used strings as part of this. Bug 1203547 is the starting point there.

Actually the latter came first, but I need to understand the chrome registry a bit to actually find all the files I want, so I looked there, and started finding the chrome overlay errors in the process.

I started to hack on this as a mach command in /testing/l10n (new dir), with docs in /testing/l10n/docs. Sadly, they're hard to tie in to the docs in /python/compare-locales/docs, as far as I tried to. Comments on the where and what welcome.

Comment 1

[Axel Hecht [:Pike]]

Attachment: svg of overlay problems

Progress report, I find a few issues, and this is the svg output of a graphviz asset the current code creates.

Those are chrome overrides that don't point back to any sources that get packaged.

The loops in the graph are the sparse-toolkit overrides, so that's working as intentioned (unless broken by jar.mn, as in the case of passwordmgr.properties.)

Good that much of those jar.mn mistakes are mine originally. I guess.

Not yet much value in sharing the code I use, I'm mostly stumbling right now, trying to find out what's right, and what's wrong in my code, and what's wrong in the tree ;-)

Comment 2

[Axel Hecht [:Pike]]

I stopped this effort, and with the conversion to Fluent, the chrome overlay problems go away by itself. Thus we can drop this for good.