[bug-libunistring] Collation on unassigned code points

[Aleksander Morgado]

Hi Bruno,

According to section 7.1.2 in UTS#10, unassigned code points, as the
ones in private planes, have a implicit weight to be used during
collation:
http://unicode.org/reports/tr10/#Unassigned_And_Other

But my tests with u8_strcoll() show that when having for example U
+10FFFD, the collator function in libunistring will actually ignore that
code point. Attached a test program which orders several strings, which
gives the following output:
 Africa
 america
 Ball
 land
 􏿽 last
 lazy
 zzz

When using the same approach with libicu, it orders in the expected way:
 Africa
 america
 Ball
 land
 lazy
 zzz
 􏿽 last


Any reason for this behavior in libunistring?

Cheers!

-- 
Aleksander

libunistring.c
Description: Text Data