Bug#351715: pinentry doesn't work properly on dumb terminals

[Werner Koch]

On Mon, 27 Feb 2006 13:43:04 +0100, Peter Eisentraut said:

> Am Montag, 6. Februar 2006 23:02 schrieb Miles Bader:

>> I often want to use gpg, and thus pinentry, in an Emacs shell buffer or
>> the like where a "graphical" curses-based dialog like pinentry-curses
>> doesn't work; in some cases I do this when logged in remotely via ssh,

What about  M-x ansi-term  then?

>> Pinentry or pinentry-curses should really have a fallback mode that
>> simply turns off echoing and reads from the tty, just like gpg does
>> when gpg-agent isn't used.

If you have this very special requirement you better use 
gpg --no-use-agent
then. 

>> [To tell the truth, I dislike the heavyweight and intrusive
>> pinentry-curses dialog -- it obscures the terminal output which usually
>> tells me exactly why gpg is being run! -- and I'd really like to be able

As of now gpg uses gpg-agent and thus pinenrty only in a very limited
way; i.e. for passphrase caching.  However the greater plan is to move
all operations related to private keys to gpg-agent and thus gpg won't
be able to decide whether a passphrase is required or not.  This has
been implemented in gpg2 (part of gnupg 1.9 but currently not
suggested for use) as well as in gpgsm, gpg's S/MIME cousin.  Further
the ssh-agent included in gpg-agent uses pinentry and you can't
predict when it requires input.



Salam-Shalom,

   Werner