Uppercase first letter of notify messages if current locale is Finnish.

See patch #3028

(Browse SVN revision 22596)

This patch goes nowhere in resolving the uppercasing problem. It does uppercasing of the translated format string, so in the very case translator cannot uppercase her/himself, i.e., when the (translated) format string begins with "%s", subject to uppercasing is character "%". There's no way to cause uppercasing of the parameter string at this level (before the actual string has been constructed from format and parameters).

Ah so this isn't so much about forcing an uppercase but rather forcing a lowercase in the middle of a sentence? So that something like
Mountains
becomes
Your engineers have converted this terrain to mountains

Is that it? That way round, it shouldn't be a problem for Gaelic, we use lowercase throughout anyway except for names.

I'm not certain I totally understand the meaning of the suggested change but in Irish and Gaelic usually anything that forces uppercase initials is a terrible idea because we have words with obligatory lowercase initial irrespective of where they appear (e.g. h-Èireann, h-ÈIREANN, t-aran, t-ARAN).

We sometimes want to have same translation to appear with lowercase first letter in some places, and uppercase first letter in other places. patch #3027 may eventually lead to good solution for this, but here's quick hack for urgent short-term (2.4, 2.3?) needs.

It introduces new macro C_() that returns string with first character uppercased in some cases. It works only if that said first character is 7-bit ascii character (single-byte UTF-8 character) This is not completely useless, as translators still control what word they push there. If all unit names in their translation start with ascii-character, they can put them via "%s" to the beginning of this kind of string. If some unit name does not begin with ascii-character, they cannot put it to the beginning, but neither were they able to do so before.

Note that unlike other xx_() macros, this does not return same statically allocated constant string for all identical calls, but returned string is constructed for each call. For this reason it has to be also freed with separate function call when string is no longer needed.

Uppercasing any other characters than first one in UTF-8 text would be hard, even if that particular character itself is ascii one. Since characters before it are variable width, it's not easy to tell which bytes belong to which character (or even if certain byte looking like ascii character is indeed ascii character or latter part of multibyte character)