Ha! Despite it looking very much like I just copied and resent essentially the same thing Jens did two minutes earlier, I swear I didn't! :)

Tom.

On 1/26/23 12:23 PM, Fraser Gordon via Std-Proposals wrote:

Hi folks,

I was reading the [lex.ccon] section of the current draft and noticed an oddity in the hexadecimal-escape-sequence production:

hexadecimal-escape-sequence:

\x hexadecimal-digit-sequence
hexadecimal-escape-sequence hexadecimal-digit
\x{ simple-hexadecimal-digit-sequence }

The combination of the 2nd and 3rd rules seems to say that \x{20}20 should be parsed as a hex escape sequence (presumably representing u+2020). Octal doesn't have this issue as octal-escape-sequence doesn't have a recursive definition.

I tried this on Godbolt. GCC and Clang interpret '\x{20}20' as a multicharacter literal and "\x{20}{20}" as the string " 20". MSVC and icc appear not to support \x{}.

I'd guess this isn't intentional as it would defeat the purpose of having \x{} sequences? At least two implementations seem to agree.

Good catch!

I think we want:

hexadecimal-escape-sequence:
    \x hexadecimal-digit-sequencesimple-hexadecimal-digit-sequence
    hexadecimal-escape-sequence hexadecimal-digit
    \x{ simple-hexadecimal-digit-sequence }

Would you be so kind as to file a CWG issue at https://github.com/cplusplus/CWG?

Tom.

Fraser