Make xparse not gobble end of line character on optional argument

Question

When using xparse to define a new environment with an optional argument O{}, the first end of line character is gobbled as can be seen in the following example document:

\documentclass{article}
\usepackage{xparse}

\ExplSyntaxOn
\cs_new_protected:Npn \__my_end_of_line: { X }

\NewDocumentEnvironment{mycode}{ !O{} }{
    \char_set_catcode_active:N \^^M
    \char_set_active_eq:nN {`\^^M} \__my_end_of_line:
}{}
\ExplSyntaxOff

\begin{document}
\begin{mycode}
abc
\end{mycode}
\end{document}

which outputs

abcX

where I would expect

XabcX

There already is this question which is similar, but doesn't make use of active characters. Also the comments in that question don't help here. Neither is prefixing the argument specifier with ! working nor is it working in a newer xparse version (my version is xparse 2019-05-28).

Is there a way to work around this problem?

---

EDIT: My use case is to have a special verbatim-like environment with some optional setup at the beginning. The problem is that I need to read ahead until I find the end of the first line/the first ^^M to discard anything that is on that line (it's usually empty). This works fine if an optional argument is present, but without an optional argument the first actual line of code is considered the line to be discarded, which is unwanted.

Answer 1

As requested, my quick mock-up as an answer. You'll want to first change the category codes before you check for the optional argument's present, then switch it back for the argument grabbing and reset the catcodes for the actual environment.

The following example does a pretty simple grab for the optional argument, as it is done in LaTeX2e, but this is inferior to an O{} argument from xparse as there will be no bracket balancing. You might circumvent that by doing a \NewDocumentCommand \__mycode_parse_arg:w { !O{} } { ... }, but \NewDocumentCommand shouldn't be used for code-level macros (hence the name).

I setup the optional argument such that it changes the definition of \__mycode_end_of_line:, just to show that it's working.

\documentclass{article}
\usepackage{xparse}

\ExplSyntaxOn
\cs_new_protected:Npn \__mycode_end_of_line: { X }
\cs_new_protected:Npn \__mycode_real_begin:
  {
    \__mycode_catcode_setup:
    \char_set_active_eq:nN { `\^^M } \__mycode_end_of_line:
  }
\cs_new_protected:Npn \__mycode_parse_arg:w [ #1 ]
  {
    % do whatever with the optional argument
    \cs_set_protected:Npn \__mycode_end_of_line: { #1 }
    \__mycode_real_begin:
  }
\cs_new_protected:Npn \__mycode_catcode_setup:
  {
    \char_set_catcode_active:N \^^M
  }
\NewDocumentEnvironment { mycode } {}
  {
    \group_begin:
    \__mycode_catcode_setup:
    \peek_meaning:NTF [
      {
        \group_end:
        \__mycode_parse_arg:w
      }
      {
        \group_end:
        \__mycode_real_begin:
      }
  }
  {}
\ExplSyntaxOff

\begin{document}
\begin{mycode}
abc
\end{mycode}

\begin{mycode}[Y]
abc
\end{mycode}
\end{document}