Re: [PATCH v6] regexp-replace and ^, word boundary or look-behind operators (and more).

[Mikael Magnusson <mikachu@xxxxxxxxx>]

On Sun, Jun 7, 2026 at 3:16 PM Stephane Chazelas <stephane@xxxxxxxxxxxx> wrote:
>
> 2026-06-07 12:36:53 +0200, Mikael Magnusson:
> [...]
> >  setopt functionargzero does not take effect when we have already entered the
> > function, so use $funcstack[1] instead.
> [...]
>
> Thanks. I have to admit I'm a bit baffled as to how I ended up
> adding that (and that I would have thought about the case where
> functionargzero is unset in the first place).
>
> Probably by testing in other emulation modes.
>
> If $0 == functionname is not restored by emulate -L zsh, that
> would suggest we'd need to replace $0 with $funcstack[1] in all
> functions we expect to work whatever the user's set of options.
>
> I see $funcstack doesn't work when zsh is *started* in sh/ksh/csh
> emulation, but I wouldn't expect those instances to use
> autoloaded functions anyway.

Hmm, right, you'd have to do zmodload -F zsh/parameter p:funcstack
(this does take effect immediately, because the variable it reads is
maintained always by exec.c). Now that I've looked around in the code,
I see that we do indeed use $0 in a lot of places already. Maybe it's
better to just leave this as $0 too and leave that particular problem
to another cleanup. (It's weird how difficult it is to just
consistently get the name of the currently executing function.)

Maybe a reasonable compromise is just ${funcstack[1]:-$0} or so (but
of course, functionargzero is off in ksh emulation so that'll still
say ksh.) The absolutely simplest solution is to just replace all
instances of $0 with the name of the file :). Argh, if you're in ksh
emulation it's funcstack[0]. You just can't win.

TL;DR I'll just leave it as $0 but remove the setopt.


--
Mikael Magnusson