Just realized that I failed to include the "mkshadow" file in the patch.

On Sun, Aug 6, 2023 at 3:12 PM Bart Schaefer <schaefer@brasslantern.com> wrote:
>
> On Tue, Jul 4, 2023 at 12:06 AM Marlon Richert <marlon.richert@gmail.com> wrote:
> >
> > Perhaps then at least add documentation for _shadow to compsys.yo?
>
> I did that the other day (7/26, ChangeLog "unposted (cf. 51899):
> Doc/Zsh/compsys.yo: document _shadow") and after reading my own
> documentation, decided I didn't like what it said.
>
> Consequently I've produced the attached patch.  The most significant
> visible change is that _unshadow no longer requires any arguments (and
> ignores any that it gets), instead it just unwinds anything done by
> the most recent previous call to _shadow.  This includes a special
> case for more than one call to _shadow with the same -s suffix
> argument, removing my objection to Marlon's _approximate patch in
> workers/51861.
>
> On Thu, Jun 15, 2023 at 7:29 AM Marlon Richert <marlon.richert@gmail.com> wrote:
> >
> > I would find _shadow useful in my own Zsh code, too, not just completions. How
> > about moving it from Completion/ to Functions/?
>
> To address this, this patch adds Functions/Misc/mkshadow, which
> defines wrappers around _shadow and _unshadow.  Although I didn't
> explicitly mention the reasoning in the doc, these are "nofork-ready"
> in that they return the suffix in $REPLY, so once those patches are
> applied one can do
>   local suffix=${| mkshadow ... }
> This is actually implemented via _shadow except that it won't clobber
> an existing $REPLY whereas mkshadow unsets REPLY on the way in.
>
> Other changes under the hood:
> * Use namespaces for the global variables used to track state
> * Keep track of whether the shadowed command is a function, builtin,
> or external, so that the placeholder "command@suffix" function is
> consistently created/removed
> * Handle many pathological cases such as wanting to shadow the "-"
> precommand modifier (although there's still no way to fully duplicate
> the precommand behavior)
>
> And, update the doc for all of this.

Here's the whole patch again so it can have one article number when committed.