Re: [regression] %2K prompt expansion no longer works

[Mikael Magnusson <mikachu@xxxxxxxxx>]

On 8/30/20, Stephane Chazelas <stephane@xxxxxxxxxxxx> wrote:
> $ print -rnP %2K | hd
> 00000000  1b 5b 34 30 6d                                    |.[40m|
> 00000005
>
> %K{2} is fine:
>
> $ print -rnP '%K{2}' | hd
> 00000000  1b 5b 34 32 6d                                    |.[42m|
> 00000005
>
> Introduced by worker:30496
> (https://www.zsh.org/mla/workers/2012/msg00421.html
> 8a9b141652a0e4157056dc21e36a64ac712a7ee7 git commit).
>
> Oddly, the comment there:
>
>> Well, nobody objected so here's a version with de-duplicated
>> code. The previous patch broke %K, but it works now (passed 1
>> for is_fg on both paths).
>
> Actually describes the bug here. Looks like Mikael just missed a bit in
> that
> fix.
>
> Should be fixed with:
>
>
> diff --git a/Src/prompt.c b/Src/prompt.c
> index bc9734720..997327e18 100644
> --- a/Src/prompt.c
> +++ b/Src/prompt.c
> @@ -258,7 +258,7 @@ parsecolorchar(zattr arg, int is_fg)
>  		bv->fm--;
>  	}
>      } else
> -	arg = match_colour(NULL, 1, arg);
> +	arg = match_colour(NULL, is_fg, arg);
>      return arg;
>  }

Just noticed that these [unposted] testcases don't pass. Added some
debugging prints and it turns out that %F{green} uses a different code
than specifying the number. Eg,

    echo $F1 $F2 $F3 $F4 | cat -v

results in

^[[38;5;2m ^[[38;5;2m ^[[38;5;2m ^[[32m

(the results are equivalent if you explicitly set fg_start_code to
^[[38;5;, but the test case doesn't do that, and of course that might
not be what terminfo contains anyway.)

I think the correct fix is to not include F4 and K4 since they use
these definitions in the source code:
#define TC_COL_FG_START	"\033[3"
#define TC_COL_BG_START	"\033[4"

which will not match the others.

-- 
Mikael Magnusson