pinentry-curses can be forced to loop endlessly

NIIBE Yutaka gniibe at
Mon Jun 15 14:59:28 CEST 2015

On 06/15/2015 07:54 PM, Steffen Nurpmeso wrote:
> Well no, that is, maybe, but i liked pinentry-curses pretty much,
> it was (working well but then it) just (that it) entered an
> endless loop when i interrupted it once i've tested
> *agent-shell-lookup* implementation.  In my opinion it should
> instead have handled the interrupt (cleanup and whatever), then
> reraise the signal or exit error (or whatever).

Since I'm not sure if I understand your report correctly, please let
me rephrase your problem.

When a program (in your case mail user agent) kicks the chain of:

    (1) gpg --invoke--> gpg-agent --invoke--> pinentry-curses

and a user tries to interrupt, it goes well.

On the other hand, when gpg-agent is there already, it goes:

    (2) gpg --connect-> gpg-agent -invoke-> pinentry-curses

Then, a user tries to interrupt, it goes wrong somehow.

In the original report, the problem was: pinentry-curses's going
endless loop.

Now, pinentry-curses doesn't work well as you expected. --- (*)

Is this description correct?  I don't understand well about the part
of (*).  Do you mean that the screen is wrong?

I think that the cause of the difference between (1) and (2) might be
the controlling terminal of a process.  Let me investigate.

More information about the Gnupg-devel mailing list