But UTF-8 should work for saving and loading text files processed by fasm. If you ensure that there is not a UTF-8 signature at the start of the file then there should be no difficulty.

But even if you have the UTF-8 signature then you can make the first line be a colon to make it a label. It would look like this:

Code:
: ; <--- editor hidden UTF-8 signature now becomes a redundant label in fasm
; code goes here
...

:format elf executable
mov eax, 1
int 0x80    

~ hd BOM.asm 
00000000  ef bb bf 3a 66 6f 72 6d  61 74 20 65 6c 66 20 65  |...:format elf e|
00000010  78 65 63 75 74 61 62 6c  65 0a 6d 6f 76 20 65 61  |xecutable.mov ea|
00000020  78 2c 20 31 0a 69 6e 74  20 30 78 38 30           |x, 1.int 0x80|
0000002d
~ fasm BOM.asm && ./BOM 
flat assembler  version 1.73.31  (16384 kilobytes memory)
1 passes, 91 bytes.
~     

: ; BOM
format  ELF64 executable 3
segment executable
entry   $
        call    secondProc
        mov     eax, 60
        xor     dil, dil
        syscall

include "second.asm"    

: ; BOM
secondProc:
        mov     edx, .length
        lea     rsi, [.hello]
        push    1
        pop     rdi
        mov     eax, edi
        syscall
  .hello        db 'Hello world!'
  .length = $ - .hello    

$ fasm -m 102400 main.asm
flat assembler  version 1.73.32  (102400 kilobytes memory, x64)
second.asm [1]:
: ; BOM
processed: :
error: symbol already defined.    

=0 ; BOM
format  ELF64 executable 3
segment executable
entry   $
        call    secondProc
        mov     eax, 60
        xor     dil, dil
        syscall

include "second.asm"    

=0 ; BOM
secondProc:
        mov     edx, .length
        lea     rsi, [.hello]
        push    1
        pop     rdi
        mov     eax, edi
        syscall
  .hello        db 'Hello world!'
  .length = $ - .hello    

$ fasm -m 102400 main.asm
flat assembler  version 1.73.32  (102400 kilobytes memory, x64)
2 passes, 166 bytes.    

$ fasm -m 102400 main.asm
flat assembler  version 1.73.32  (102400 kilobytes memory, x64)
main.asm [1]:
=0 ; BOM
processed: =0
error: illegal instruction.    

BOM=0 ; works whether invisible BOM is present or not
;...    

The fix for BOM and BOM-less sources?

Code:

BOM=0 ; works whether invisible BOM is present or not
;...    

I think thre real fix is to not have the BOM. UTF-8 doesn't need the BOM, it is all bytes anyway, there is no ordering problem to solve.

Some text editors don't display the file in UTF-8 without the BOM and they assume it's ASCII instead. Which to be honest is a sane default because ASCII files definitely don't have any BOM, so it's the most backwards compatible solution.

I think it's quite easy to add BOM support to compiler.

Sure, some editors that are annoying.

The editor I use the most works perfectly fine to detect UTF-8 vs ASCII vs ISO-8859 without any BOM. It isn't hard, it only needs is small amount of logic.

Heuristics are never perfect.

Where to post, here?

        mov     eax,[esi]
        cmp     ax,0FEFFh ; UTF-16 (LE) / UTF-32 (LE)
        je      unsuppoted_bom
        cmp     ax,0FFFEh ; UTF-16 (BE)
        je      unsuppoted_bom
        cmp     eax,0FFFE0000h ; UTF-32 (BE)
        je      unsuppoted_bom
        cmp     eax,3ABFBBEFh ; UTF-8 + colon char
        je      bom_no_skip ; don't skip if colon trick is used (for backward compatibility)
        and     eax,00FFFFFFh
        cmp     eax,00BFBBEFh ; UTF-8
        jne     bom_no_skip
        add     esi,3 ; skip BOM
      bom_no_skip:
        mov     ebx,esi ; moved down by Jin X
    

        cmp     byte [esi+ecx],':' ; BOM + colon trick
        je      bom_no_skip ; don't skip for backward compatibility
        add     esi,ecx ; skip BOM
      bom_no_skip: