On 2019-10-31 at 09:26:16, Doan Tran Cong Danh wrote:
> diff --git a/t/t0028-working-tree-encoding.sh b/t/t0028-working-tree-encoding.sh
> index 7aa0945d8d..bfc4fb9af5 100755
> --- a/t/t0028-working-tree-encoding.sh
> +++ b/t/t0028-working-tree-encoding.sh
> @@ -17,7 +17,7 @@ test_lazy_prereq NO_UTF32_BOM '
>  write_utf16 () {
>  	if test_have_prereq NO_UTF16_BOM
>  	then
> -		printf '\xfe\xff'
> +		printf '\376\377'
>  	fi &&
>  	iconv -f UTF-8 -t UTF-16
>  }
> @@ -25,7 +25,7 @@ write_utf16 () {
>  write_utf32 () {
>  	if test_have_prereq NO_UTF32_BOM
>  	then
> -		printf '\x00\x00\xfe\xff'
> +		printf '\0\0\376\377'
>  	fi &&
>  	iconv -f UTF-8 -t UTF-32
>  }

Yeah, this patch looks obviously correct.  For some reason, I knew that
printf(1) didn't accept hex sequences and yet I still wrote them.

As Peff pointed out, this is because we only trigger this case on musl;
musl is the only known implementation of iconv that produces big-endian
data without a BOM for UTF-16 and UTF-32.
-- 
brian m. carlson: Houston, Texas, US
OpenPGP: https://keybase.io/bk2204