bug#39058: 26.1; epg-sign-string should report an error if no pinentry is available

[Lars Ingebrigtsen <larsi@gnus.org>]

Arne Babenhauserheide <arne_bab@web.de> writes:

> If you run epg-sign-string without having pinentry available, then the
> result is nil and instead of an error, even though the status-code seen
> for the process returned by epg-start-sign is 2 and the content of the
> process-buffer is 
>
> "
> Process epg exited abnormally with code 2
> [GNUPG:] KEY_CONSIDERED XXXXXXXXXXXXXXX… 2
> [GNUPG:] BEGIN_SIGNING H8
> [GNUPG:] FAILURE sign 67108949
> "
>
> And the buffer *epg-error*-270683 contains:
> gpg: Beglaubigung fehlgeschlagen: Kein Pinentry
> gpg: signing failed: Kein Pinentry

When you say "without having pinentry available", do you mean that you
click "cancel" in the pinentry window, or that there's no pinentry
available at all?

> To test:
>
> (setq context (epg-make-context 'OpenPGP))
> (setq epg-test-proc (epg-start-sign context (epg-make-data-from-string
> "foo") t))
> (let ((print-length nil) (print-level nil))
> (print context))
> (epg-context-result-for context 'sign)
> (print (epg-context-result-for context 'error))
> (with-current-buffer (process-buffer epg-test-proc) (buffer-string))
> (print (epg-context-result-for context 'error)
> ;; prints \nnilnil

I'm able to reproduce this when clicking "cancel" in the pinentry
window -- no error is signalled anywhere, which I think it should be.
I'll try to look into that.

> This currently breaks signing emails for me.

I don't quite get why it breaks signing emails...

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no