Address Completion No Longer Working

[Tomi Ollila]

On Mon, Mar 12 2018, David Bremner wrote:

> david wen riccardi-zhu <dwrz at dwrz.net> writes:
>
>> Address completion worked out of the box for me with notmuch and 
>> Emacs. Recently, it's stopped working. I've tried setting 
>> notmuch-address-command to internal, as well as toggling 
>> notmuch-address-toggle-internal-completion, but neither has been 
>> able to return the functionality.
>
> "stopped working" usually suggests some configuration change to me. It
> would be helpful to know what precisely changed.
>
> As far as debugging, I had two ideas to try.
>
> - try running (notmuch-address-matching "dwrz") in *scratch* (or IELM,
>   or M-:) ; this will eliminate company-mode as a suspect, and
>   potentially give you a traceback if something is going wrong.
>
> - have a look at the variable notmuch-address-save-filename. Potentially
>   set it (back) to nil to disable persistent caching. If that fixes it,
>   have a look at the corresponding file, see if something corrupted it.
>
>> I use:
>> Arch Linux X86-64, Kernel 4.15.7-1-ARCH
>> fish shell
>
> fish breaks lots of assumptions for shells, but I guess you didn't just
> switch.

Notmuch Emacs MUA should not rely using any particular $SHELL in any case.
It still uses amdragon's brilliant 
"/bin/sh" "-c" "exec 2>\"$1\"; shift; exec \"$0\" \"$@\"" before emacs 25
and according to make-process' docstring it doesn't use shell (either)...

Also, I tested

$ SHELL=/bin/false ./devel/try-emacs-mua -q

and tried (built-in!) address completion (slow >;), and it worked fine.

>> notmuch 0.26
>
> some subtle things changed with respect to starting external processes
> in notmuch 0.26; if your recent configuration change was upgrading
> notmuch, that might be worth further investigation.

If that is (finally) the case, I'd be interested to know.

Tomi