bug#17836: 24.3; `describe-fontset' confused about e.g. ?\C-@

[Eli Zaretskii]

> From: Samuel Bronson <naesten@gmail.com>
> Date: Sun, 22 Jun 2014 21:57:07 -0400
> 
> Fontset: 
> -misc-fixed-medium-r-semicondensed--13-*-*-*-*-*-fontset-xterm.default
> CHAR RANGE (CODE RANGE)
>     FONT NAME (REQUESTED and [OPENED])
> C-@ .. � (#x43 .. #x10FFFF)
>     -Misc-Fixed-medium-r-semicondensed--13-*-75-75-c-120-ISO10646-1
> --8<---------------cut here---------------end--------------->8---
> 
> Notice how #x43 is NOT a representation of `?\C-@' but, in fact, of
> `?C'?

That's because print-fontset-element does this:

  (beginning-of-line)
  (let ((from (following-char))

IOW, it assumes that there's a single character there, not a
human-readable description of a character, such as "C-@".

How about submitting a patch that uses 'kbd', say?

> Why would you try to extract the codepoints AFTER formatting the
> range as a string ...?

Because the formatting of the codepoints is done by describe-vector,
which doesn't pass the codepoints to print-fontset-element.  So it
needs to reverse-engineer the codepoints from the text that was
already inserted into the buffer.