Re: Small patch to enable use of gpg-agent with pgg

[Simon Josefsson <jas@extundo.com>]

Daiki Ueno <ueno@unixuser.org> writes:

>>>>>> In <b4edf9b5-2663-4d5c-981e-ddabca619ad6@well-done.deisui.org> 
>>>>>>	Daiki Ueno <ueno@unixuser.org> wrote:
>> > Maybe you can finish this code, and I can debug why it doesn't work
>> > for a smartcard separately.  It is probably not an important feature.
>
>> The new code mostly finished.
>
> Ok, the attached file is (hopefully) the final version of the new code.
> Differences from the previous one are:
>
> - Passphrase caching now works again.
> - pgg-gpg-use-agent is abolished.  Add "use-agent" to ~/.gnupg/gpg.conf
>   if you want to enable use of gpg-agent, as Miles said.
> - Tested with typical cases.  I also attach some test cases I used.

This works fine except for used with a smartcard.  The current version
works with a smartcard (when you enable pgg-gpg-use-agent), so I think
that should be fixed.

I debugged this, and it seems invoking gpg using the smartcard will
not print any useful [GNUPG:] statements:

/usr/bin/gpg --no-tty --status-fd 1 --command-fd 0 --yes --output /tmp/pgg-output29825FPB --detach-sign --armor --verbose --local-user jas --textmode
...
gpg: using subkey AABB1F7B instead of primary key B565716F
gpg: writing to `/tmp/pgg-output29825FPB'

Then it waits for input.  PGG infloop waiting for GOOD_PASSPHRASE
here.  I think it should wait for 'gpg: writing to' instead, it is the
last printed statement for both smartcard and non-smartcard cases, and
it seem to be printed when gpg is ready to accept input.

For references, here is the non-smartcard case:

jas@latte:~$ /usr/bin/gpg --no-tty --status-fd 1 --command-fd 0 --yes --output /tmp/pgg-output29825FPB --detach-sign --armor --verbose --local-user b565716f! --textmode
[GNUPG:] USERID_HINT EDA21E94B565716F Simon Josefsson <simon@josefsson.org>
[GNUPG:] NEED_PASSPHRASE EDA21E94B565716F EDA21E94B565716F 1 0
<<<hangs here until i type password to the agent>>>
[GNUPG:] GOOD_PASSPHRASE
gpg: writing to `/tmp/pgg-output29825FPB'

Then it hangs waiting for the input to sign.

What do you think?  It might be ugly to depend on the 'gpg: writing'
output instead of the [GNUPG:] stuff, but I see no option here.

I suspect password-less keys may have similar issues.