[Ada] Improve behavior of -gnaty

[Arnaud Charlet <charlet at adacore dot com>]

Tested on i686-linux, committed on HEAD

As documented and implemented, in the previous versions of GNAT, the
check for line length maximum was based on a byte count, rather than
a logical count of characters. This patch changes the behavior to
count characters instead of bytes, which is a clear improvement for
the wide character case.

Note that the implementation works by counting bytes, and adjusting
for wide character sequences. This avoids any extra overhead in the
common (non-wide character) case.

The following test program:

--  xxxxxxxxxxxxxx Э
--  xxxxxxxxxxxxxx  Э
procedure P1 is
begin
   null;
end P1;

when compiled with:

gcc -c -gnatW8 -gnatyM20 p1.adb

should generate only one diagnostic:

p1.adb:2:21: (style) this line is too long

2005-09-01  Robert Dewar  <dewar@adacore.com>

	* scng.adb (Check_End_Of_Line): Count characters, rather than bytes
	(makes a difference for wide characters)

	* widechar.adb, widechar.ads: 
	Add Wide_Char_Byte_Count feature to count chars vs bytes

Attachment: difs.29
Description: Text document