fix map definition / generation

the map is defined with several settings (mapsize, size, tilesperplayer,
xsize, ysize, topology). Now, some combinations are invalid like in this
bug

mapsize == XYSIZE && ysize % 2 != 0 && MAP_IS_ISOMETRIC

I did add the required checks but a framework to check the settings
would be helpful. Check also bug #18673 and patch #2542.

see gna bug #18875

(Browse SVN revision 20506)

fix map definition / generation

the map is defined with several settings (mapsize, size, tilesperplayer,
xsize, ysize, topology). Now, some combinations are invalid like in this
bug

mapsize == XYSIZE && ysize % 2 != 0 && MAP_IS_ISOMETRIC

I did add the required checks but a framework to check the settings
would be helpful. Check also bug #18673 and patch #2542.

see gna bug #18875

(Browse SVN revision 20503)

In settings_game_load(), care is taken to only run setting action functions once all settings are loaded to avoid dependency issues, but setting validation functions are run one at a time during loading. Since some of these functions do cross-checks (xsize/ysize, timeout/unitwaittime), depending on the order of settings in the save file, it is possible for a setting to be rejected even if it would have been valid had other settings been loaded first.

Contrived (unplayable) example attached, with the following settings section:

which generates an error when loaded with trunk (r20250):

but not if ysize is moved before xsize.

This is unlikely to bite anyone in practice at the moment, but if validation functions get more cross-checking, it could become a problem.

This will be fiddly to fix, as the full set of new (or replaced) settings will have to be held somewhere temporarily while the new set is validated, so that settings that fail the validation at the end can be rolled back.

(Also, the non-setting-specific validation is not called, e.g., setting_int_validate() which does the range check.)