Re: [PATCH] expanding $'\c^A'

[Chet Ramey]

On 10/13/23 12:36 AM, Grisha Levit wrote:
> Handle $'\c^A' and $'\c^?' correctly when expanded by the parser:
>
> $ bash -c $'printf "%q\n" $\'\\c\001 \\c\177\''
> $'\001\001 \001\177'
>
> ---
> diff --git a/lib/sh/strtrans.c b/lib/sh/strtrans.c
> index d3b27f3b..3df33ade 100644
> --- a/lib/sh/strtrans.c
> +++ b/lib/sh/strtrans.c
> @@ -198,6 +198,8 @@ ansicstr (const char *string, size_t len, int
> flags, int *sawc, size_t *rlen)
>      s++;
>      if ((flags & 2) && c == '\\' && c == *s)
>        s++; /* Posix requires $'\c\\' do backslash escaping */
> +   else if ((flags & 2) && c == CTLESC && (*s == CTLESC || *s == CTLNUL))
> +     s++;

It looks like that should be c = *s++. If there are literal ^A and ^? in
the input, the parser passes the string from parse_matched_pair as
'\c\001\001 \c\001\177' -- it adds the CTLESC escaping. I think your intent
here is to remove it.

Otherwise you change \c^? into ^A when it should end up being ^_.

--
``The lyf so short, the craft so long to lerne.'' - Chaucer
                 ``Ars longa, vita brevis'' - Hippocrates
Chet Ramey, UTech, CWRU    chet@case.edu    http://tiswww.cwru.edu/~chet/