All: Fix undefined errors in translations

[wh201906]

Some original strings don't have %s, but the corresponding translations included it. This mismatch causes the text undefined to appear in the UI.

[wh201906]

This could also occurs in other translations. I did a rough search with regular expression ^msgid(?!.*%s).*\nmsgstr.*%s and ^msgid(?!.*%d).*\nmsgstr.*%d. Here are the results.

[laurent22]

That's a good catch, thank you. We have a script to check the translations in validate-translations.js but looks like the msgfmt tool doesn't catch these errors. I've also tried the --check-format flag and that didn't make a difference. I guess we should add your regex to the check to prevent such strings being inserted.

[wh201906]

@laurent22 I tried msgfmt --check-format and found it works for the mismatched string only when the string is marked with c-format. Also, if the string is marked with fuzzy, the mismatch won't be reported.

---

For example, on commit 546c75b, I can get two mismatched strings in packages/tools/locales/hu_HU.po

When running msgfmt -v --check-format, it doesn't show any errors.

---

However, if I add #, c-format to the string, the msgfmt will report the mismatch as an error.

---

If I change #, fuzzy to #, fuzzy, c-format, the msgfmt won't report the mismatch, which is reasonable.

---

Maybe the built-in check of msgfmt is better than the custom regular expressions?

[wh201906]

This could also occurs in other translations.

Based on the facts I just discovered I think we should ignore the strings marked with fuzzy. Here are the new regular expressions for test:
#(?!, fuzzy).*\nmsgid(?!.*%s).*\n(msg.+\n)*msgstr.*%s
#(?!, fuzzy).*\nmsgid(?!.*%d).*\n(msg.+\n)*msgstr.*%d

And here are the results:

[laurent22]

Thanks for looking into this @wh201906. So it means that we should somehow mark the strings as "c-string" to get this msgfmt tool working?

[wh201906]

Yes. But I'm not sure if this is good for Joplin. Normally xgettext could handle the c-format comment automatically, but I found you have switched fromxgettext to gettext-extractor for some reasons.