\c_document_cctab does not make upper-half of the 8-bit range active

[stone-zeng]

In l3interface:

• \c_document_cctab:

  Category code table for a standard LaTeX document, as set by the LaTeX kernel. In particular, the upper-half of the 8-bit range will be set to "active" with pdfTeX only. No babel shorthands will be activated.

However, for the following example:

\csname tl_analysis_show:n\endcsname{1:^^70^^80}

\ExplSyntaxOn
\tl_analysis_show:n { 2:^^70^^80 }

\cctab_begin:N \c_document_cctab
\csname tl_analysis_show:n\endcsname{3:^^70^^80}
\cctab_end:

\ExplSyntaxOff

Run with latex or pdflatex (or *-dev), it gives

LaTeX2e <2020-10-01>
L3 programming layer <2020-09-24> xparse <2020-03-03>
The token list contains the tokens:
>  1 (the character 1)
>  : (the character :)
>  p (the letter p)
>  � (active character=macro:->\UTFviii@invalid@err �).
<recently read> }
                 
l.1 ...me tl_analysis_show:n\endcsname{1:^^70^^80}
                                                  
? 
The token list contains the tokens:
>  2 (the character 2)
>  : (the letter :)
>  p (the letter p)
>  � (active character=macro:->\UTFviii@invalid@err �).
<recently read> }
                 
l.4 \tl_analysis_show:n { 2:^^70^^80 }
                                      
? 
The token list contains the tokens:
>  3 (the character 3)
>  : (the character :)
>  p (the letter p)
>  � (the character �).
<recently read> }
                 
l.7 ...me tl_analysis_show:n\endcsname{3:^^70^^80}
                                                  
? 

So after \c_document_cctab, ^^80 becomes a normal character rather than an active one.

BTW, the variables described here should be \c_code_cctab and \c_document_cctab:

latex3/l3kernel/l3cctab.dtx

Lines 769 to 775 in 19ea461

	% \begin{variable}{\c_document_cctab, \c_other_cctab}
	% To pick up document-level category codes, we need to delay set up to the
	% end of the format, where that's possible. Also, as there are a \emph{lot}
	% of category codes to set, we avoid using the official interface and store the
	% document codes using internal code. Depending on whether we are in the hook
	% or not, the catcodes may be code or document, so we explicitly set up both
	% correctly.

[josephwright]

Hopefully now sorted: I was forgetting that we needed to do this

[josephwright]

@PhelypeOleinik (but others too): is there any point in the delay I set up for creating the table? I think we are now ignoring LaTeX2e settings entirely, so we could skip that: correct?

[zauguin]

@josephwright If I'm not missing anything we are only ignoring the LaTeX2e settings in 8-bit engines and therefore still need the delay for Unicode engines.

[josephwright]

@zauguin Ah right yes: the catcodes from UnicodeData have to be loaded. Probably should note that in the code!