Following e-mail discussion. git logic for the most recent patch managed to mark a Hungarian translation file (which doesn't exist yet on SVN) as 'branched' from the Albanian file (which does exist, but only has fuzzy translations & hence looks empty to the old logic).
Therefore we need to be careful not to prune from the Weblate mirror those empty .po files which correspond to empty SVN files.
As to how it can happen that an empty file winds up in SVN, this appears to be related to #6 -- on initial check-in, the sq.po file appears to have been generated from R-base.pot, not R.pot:
Following e-mail discussion.
gitlogic for the most recent patch managed to mark a Hungarian translation file (which doesn't exist yet on SVN) as 'branched' from the Albanian file (which does exist, but only has fuzzy translations & hence looks empty to the old logic).Therefore we need to be careful not to prune from the Weblate mirror those empty .po files which correspond to empty SVN files.
As to how it can happen that an empty file winds up in SVN, this appears to be related to #6 -- on initial check-in, the sq.po file appears to have been generated from R-base.pot, not R.pot:
https://github.com/r-devel/r-svn/blob/c23093fd5f0787d08e4c51ddfa4b13f6e6bb476d/src/library/base/po/sq.po
So all of those msgstr, having never been found in R.pot to begin with, would be marked
fuzzyon next update.PS I quickly glanced at R-sq.po, and I think the translations in sq.po have not been lost, having been given in duplicate to begin with.