This is really to set up a discussion probably between me and rudi when the sun comes up for him, but everyone should at least be aware of this problem.

This morning I noticed that zh-CN had reached 100%, but that translation has been thick with errors for the past few weeks, so I decided to send a note to trans-list identifying the types of errors. I was later surprised to see that rudi had pushed the zh-CN to docs.fp.o, surprising since the document wouldn't build, at least, as of a job that started about an hour before rudi pushed it!

Anyway, when I looked through the document, I noticed a number of entire sections that were untranslated. Odd, since Transifex shows it at 100%. Now I have to admit, I've been happy to see translations complete, and generally I take a quick peek to see that they look about right, but I haven't been going through documents in languages I can't read in their entirety.

I looked at a couple of other docs on docs.fp.o and some of the same sections were untranslated. I was about to blame Publican for producing an incomplete pot (sure, blame it on Publican), but then I found some docs where those same sections were translated. That leads me to suspect that something in the interaction between git and Transifex is causing the issue.

As you may or may not be aware, we have been branching liberally to try to keep the translations straight. I suspect what happened is that some of the pos were checked out from transifex when we branched, and the old po overwrote the new po. Normally git would scream bloody murder (and make your life miserable) if you tried to do this sort of thing, but depending on how Transifex interacts with git, it might work around this problem.

I'm not sure the best way to correct this. I suspect we need to identify the problem pos and merge them with the current pot, but I don't know enough about Transifex to know what this might break.

--McD