Regression on Elgg 1.8: with system cache enabled language files with more than 2 characters in filename are not recognized

[iionly]

See discussion: https://community.elgg.org/discussion/view/1852424/elgg-1819-isnt-recognizing-pt-br

If you have system cache enabled, language files don't work if their name has anything else than lowercase letters (e.g. hyphens, underscores, upper case letters).

If the language (for example pt_br) has been selected with system cache disabled, it is still used after system cache is enabled. But you wouldn't be able to re-select it again, if you switched to another language and new users can't select it at all (if the site's language isn't pt_br).

The problem exists on:

• 1.8.19 (since 1.8.14)
• 1.9 (1.9.0-rc.7)

We should choose a language code standard in order to fix this properly. See #6856

[ewinslow]

Thanks so much for your research into this, @iionly. Wouldn't be too much to ask to narrow it down to a specific Elgg version or even a particular commit?

[iionly]

It's broken since Elgg 1.8.14. The problem was introduced by the fix for #4840 (see 58ebe3f).

I think the problem is in engine/lib/languages.php in function reload_all_translations():

if (preg_match('/([a-z]+)\.[^.]+$/', $filename, $matches)) {

Underscores are not taken into account (and also no upper case letters in language file names).

[juho-jaakkola]

Is there a reason why cache handler allows only those characters? Could we just allow more?

Edit: the cache handler already allows those characters: https://github.com/Elgg/Elgg/blob/1.8/engine/handlers/cache_handler.php#L47

So yes, it seems like we should allow those characters also in the language codes.