uk: avoid translating fixed strings in protobufs

[mgeisler]

The protobuf exercise asks people to decode a few fixed byte slices:

https://google.github.io/comprehensive-rust/uk/lifetimes/exercise.html

Some of the messages in the slices are fixed English strings, such as “home”. They should thus not be translated into other languages.

This should fix the CI. I guess it wasn’t caught before merging because of a parallel update.

[mgeisler]

Made obsolete by #2506.

@reta, in case you weren't aware, the best way to indicate that a string should remain untranslated is to copy the msgid to the msgstr. That way, it won't show up as untranslated.

[reta]

@reta, in case you weren't aware, the best way to indicate that a string should remain untranslated is to copy the msgid to the msgstr. That way, it won't show up as untranslated.

Thanks @mgeisler , yes I am aware of that but it makes a bit difficult to track new messages that require translation (they are always having empty msgstr) vs the ones where translation is not required. Thanks for the hint anyway!

[mgeisler]

@reta, in case you weren't aware, the best way to indicate that a string should remain untranslated is to copy the msgid to the msgstr. That way, it won't show up as untranslated.

Thanks @mgeisler , yes I am aware of that but it makes a bit difficult to track new messages that require translation (they are always having empty msgstr) vs the ones where translation is not required.

Okay, though I don't understand how it makes difficult to track new messages? You will have

msgid "untranslated"
msgstr ""

msgid "no-op translation"
msgstr "no-op translation"

in the uk.po file. So new messages are those with an empty msgstr, the already translated messages (no-op or not) will have a non-empty msgstr.

If a no-op translation changes, you end up with something like

#, fuzzy
msgid "changed no-op translation"
msgstr "no-op translation"

The fuzzy entry will be ignored when building the book, thus resulting in "changed no-op translation" showing up, which is what you'll want.

[reta]

The fuzzy entry will be ignored when building the book, thus resulting in "changed no-op translation" showing up, which is what you'll want.

Thanks @mgeisler , this is correct, I have difficulties to track this case:

msgid "untranslated-does-not-need-to"
msgstr ""

msgid "new-translation-that-need-to-be-translated"
msgstr ""

May be this is a flaw in my workflow only, but thanks a lot for sharing your thoughts over it.

[reta]

@reta, in case you weren't aware, the best way to indicate that a string should remain untranslated is to copy the msgid to the msgstr. That way, it won't show up as untranslated.

@mgeisler my sincere apologies, I completely misread your initial suggestion, this is exactly the way we (I) should be doing it, sorry about that. Please disregard my previous comment.

[mgeisler]

@reta, in case you weren't aware, the best way to indicate that a string should remain untranslated is to copy the msgid to the msgstr. That way, it won't show up as untranslated.

@mgeisler my sincere apologies, I completely misread your initial suggestion, this is exactly the way we (I) should be doing it, sorry about that. Please disregard my previous comment.

No worries at all! 😄

Since all this is not very obvious to anybody, I guess we should add a line or two about it to the TRANSLATING documentation. Ideally, we should not put these useless strings into the PO files in the first place (and #2478 can help with that), but when they are there, we need to help people get them out of their way.