chr() generating wrong character?

[nsstudios]

Greetings,

Came across this potential bug.
As you'll hopefully be able to reproduce, when the calculation is put directly inside chr(), it generates two characters instead of one, but putting the calculation into a unicode .u variable first and then calling chr() on it works fine.
Tested with 6.01 lts (64/32 bit, c/asm), and 6.03 beta 3 (windows 64/32 bit, c/asm).

Code: Select all

EnableExplicit
Define a.u=(((120000-65536)>>9)+55192)
Debug a; 55298
Define c$=Chr(a)
Debug c$; character 55298 (0xd802)
Debug Len(c$); 1
Define u=PeekU(@c$)
Debug ""+u+" (0x"+Hex(u, #PB_Unicode)+")"
Debug (((120000-65536)>>9)+55192); 55298
c$=Chr((((120000-65536)>>9)+55192))
Debug c$; two 65553 characters (0xfffd)
Debug Len(c$); 2
u=PeekU(@c$)
Debug ""+u+" (0x"+Hex(u, #PB_Unicode)+")"
u=PeekU(@c$+SizeOf(character))
Debug ""+u+" (0x"+Hex(u, #PB_Unicode)+")"
Debug Bin(a, #PB_Unicode); 1101100000000010
Debug Bin((((120000-65536)>>9)+55192), #PB_Unicode); 1101100000000010

[Fred]

Good catch. The value you are using is not a BMP valid character (UCS-2 ranges are 0x0000 to 0xD7FF and 0xE000 to 0xFFFF as specified here https://en.wikipedia.org/wiki/UTF-16#Co ... o_U+10FFFF). I corrected the Chr() command so it returns an empty string
if the character value is not valid so we get consistant result. Added a runtime debugger check and doc is updated as well.

[nsstudios]

Thank you.

[idle]

If you need utf16 you can use this module
https://dnscope.io/idlefiles/UTF16.pb

[nsstudios]

Thank you.