phoul said we use the pt-BR strings for our stuff (Tortbutton) etc. So, I am curious what is actually wrong with our bundles. Are the Mozilla strings wrong as well? I somehow doubt that as we download the pt-PT bundle for Portugese. So, it seems we mix them and want to have pure pt-BR bundles instead? Or pure pt-PT ones?
pt-br is the most common variant, it should use pt-br in instance of pt-pt because the number of speaker is a determinant of fingerprint, and doesn't have much difference of both.
My preference is to use just pt-br for don't segment user in a lot of fingerprint.
Those are language codes for two different languages: pt_PT is Portuguese from Portugal which is different from Portuguese from Brazil (pt_BR).
We were using translation in PT from Brazil and distributing it with the wrong language code (the one for Portugal). This can be quite frustrating for users who are expecting one type of translation.
So yes, the correct language code for the language we support (translations coming from transifex) should be pt_BR.
Regarding the update, the users of the pt-PT bundle will be updated to the english bundle as the pt-PT one won't exist in the new version. I'm wondering if we should do something to update them to the pt-PT one instead.
Regarding the update, the users of the pt-PT bundle will be updated to the english bundle as the pt-PT one won't exist in the new version. I'm wondering if we should do something to update them to the pt-PT one instead.
Yes, they should probably get the full pt-BR .mar file. I guess there is just some server-side tweaking necessary? If so, could you take care of it?
After renaming the pt-PT mar files to pt-BR and running make update_responses-alpha with this patch, I get an .htaccess file containing new lines such as:
Were we using the pt one before, and this change makes us use the pt-BR one now?
Not sure. I guess what happens is the following: we ship pt-PT bundles now but looking at the above links there is no so locale. Rather, we only have pt and pt-BR. The code is presumably falling back to pt (I have not checked that) which turns out to be pt-BR which we import during translation updates. I guess the same is happening with Tor Launcher. Now, the interesting part is what is going to happen if we switch to pt-BR officially? Then we probably use the pt-BR branch as you mentioned even though we only update the pt one every time. I guess we need to update the locale import scripts as well...
Not ready for 6.5a4, alas. It needs more investigation and probably a patch to our translations import script(s). Keeping it on our November radar, though, as we want to have it in 6.5a5.
Trac: Keywords: TorBrowserTeam201611R deleted, TorBrowserTeam201611 added Status: needs_review to new
are really the ones we want (and that they are up-to-date)? If so, switching to pt-BR should be not problematic. If not, and the real strings are in pt, can we just get them into pt-BR instead and make sure translators are using that one from now on?
Browsing the content on the links above, they are pt-BR translations but they are not complete - some strings are still in English. So I don't really know if they are updated with what is on Transifex (Transifex has pt-BR as 100% complete).
ps: I can't load the PT translations on transifex for some reason
That's probably due to us not importing pt-BR translations during language string updates. Thus, once we adjusted our import script this should be fine.
Okay, we are done here with two patches to tor-browser-bundle.git: commits 06fe66a1d2d366078a63147447389e397cfbf505 and c3e365d6eb07ea07a4187380aebe2c6015a50346 and one to torbutton updating our translations import script: e2fc692ca82ee04eff6d8ffd2313116638d2f6a4. I updated Torbutton master with the latest translations, too. #20901 (moved) is for the website changes.
Trac: Resolution: N/Ato fixed Status: assigned to closed