In server/services/src/main/java/org/zanata/rest/service/ResourceUtils.java, transferToPotEntryHeader function, you split .po files reference with commas:
In our case, it breaks our workflow:
1. we convert the english version of our strings from JSON to pot, with json2po. The keys of the JSON file are the strings to translate themself, thus become references in pot file
2. we push it to zanata
3. translations are made
4. we pull the translations, in .po format, then convert it to json with po2json. And there comes the problem: some references are splitted, and po2json don't read references on multiples lines (even if it did, the commas are missing) and that results in some strings considered as untranslated.
Is there a good reason for the comma splitting? If there's no good reason, could you consider avoiding the splitting?
NB: we didn't used our JSON file as document because a few versions earlier (4.3.3), Zanata didn't extract well the strings to translate.