Freeciv

Digging into this - it's weird. The struct entry with the gt_marking field is only used when writing a secfile. All () are lost when a ruleset is read in. But for some reason when a ruleset is written out, there's code that sets the gt_marking to TRUE for all struct name_translation names regardless of whether the name had a _() in the input ruleset. Also does the same for the [about] summary="..." in game.ruleset, again regardless of whether it had a _() in the input ruleset. So if I have a custom ruleset with a new unit with an untranslated name ( name="foo", no _() ), ruledit will output that as if it were a name with a translation ( name=("foo") ). Seems wrong - I think the output should either have no _() at all, or put them in for everything in the input ruleset with a _(), which would require a bunch of new coding.

The gt_marking is also used by sg_save_scenario() in savegame3.c when saving scenarios - gets applied to "scenario.name", "scenario.authors" and "scenario.description". That seems legit.

I edited the Subject to be more accurate/general.

More of them than I expected -

actions.ruleset ui_name_*
cities.ruleset short_name
game.ruleset positive_label & negative_label
game.ruleset first_msg & cons_msg
game.ruleset "teams.names"
governments.ruleset ruler_male_title & ruler_female_title
techs.ruleset bonus_message
techs.ruleset control.flags "name" & "helptxt"
terrain.ruleset flags = { "name", "helptxt" ... } & extra_flags = { "name", "helptxt" ... }
terrain.ruleset ui_name_base_*
units.ruleset flags = { "name", "helptxt" ... } & class_flags = { "name", "helptxt" ... }
units.ruleset veteran_names
all helptext =

Had to add another wrapper #define for secfile_insert_str_vec_full() and rewrite secfile_insert_str_vec_full(). That meant changing utility/registry_ini.c/h so there is a problem with #1903, not a code conflict but line numbers don't match so patch won't apply. #1903 needs to be marked blocking #1893 - #1903 patches need to be done first.

'pentry' is often declared in unnecessarily wide scope. Variables should be declared in the innermost block possible.

Fixed.

Seen with user defined flags:

; helptxt       = displayed in the help for unit types with this flag (optional)

flags={"name","helptxt" 
_("Airbase"),_("")
_("Transform"),_("")
_("AirAttacker"),_("Very bad at attacking AEGIS.")

Maybe happen elsewhere too. _() should never (or at least under very special conditions only) be added to an empty string "".

Fixed entry_str_set_gt_marking() to not put gettext marking on empty strings.

Why can't we just not output empty strings at all? I looked into that and it seems possible, but there's a catch indicated by a comment in rulesave.c

      /* Save the user flag help text even when it is undefined. That makes
       * the formatting code happy. The resulting "" is ignored when the
       * ruleset is loaded. */

So would take more work, and be out of scope for this issue. I'm not going to mess with it - it's not broke, don't fix it.