Re: [PATCH] gpg-interface: reflect stderr to stderr

[Michael J Gruber <git@drmicha.warpmail.net>]

Jeff King venit, vidit, dixit 07.09.2016 10:39:
> On Wed, Sep 07, 2016 at 10:27:34AM +0200, Michael J Gruber wrote:
> 
>> Now, I can't reproduce C on Linux[*], so there is more involved. It
>> could be that my patch just exposes a problem in our start_command()
>> etc.: run-command.c contains a lot of ifdefing, so possibly quite
>> different code is run on different platforms.
> 
> Maybe, though my blind guess is that it is simply that on Linux we can
> open /dev/tty directly, and console-IO on Windows is a bit more
> complicated.
> 
> You might also check your GPG versions; between gpg1.x and gpg2, the
> passphrase input handling has been completely revamped.

That's a good point to note.

gpg1 asks for the passphrase (without use-agent), gpg2 always delegates
to gpg-agent (and starts it on demand).

On Linux, gpg-agent says you should

export GPG_TTY=$(tty)

to make gpg-agent find the tty, and claims it's not necessary on Win.

In fact, it's not necessary on Linux either unless you want to use
pinentry-curses.

Alas, be it gpg1.4.21 or gpg2.1.13, whatever pinentry, I do get the
passphrase prompt, even with curses (if GPG_TTY is set, which was
necessary before any patches already).

I put up a request for more input from the reporters in the github
issue. I guess that's the best way to reach them.

>> It would be great if someone with a Windows environment could help our
>> efforts in resolving issue C, by checking what is actually behind[**]: I
>> can't believe that capturing stderr keeps gpg from reading stdin, but
>> who knows. Maybe Jeff of pipe_command() fame? I'll put him on cc.
> 
> I know nothing about Windows, but I'd be surprised if gpg is reading
> from stdin, as opposed to /dev/tty. It's probably more to do with how
> gpg finds the "tty" on Windows (presumably it looks at stderr for that).
> 
> Anyway, I wrote pipe_command() in such a way as to be prepared for
> exactly this kind of thing, so it would be trivial to extend it to an
> extra descriptor. The trouble is that run_command() doesn't understand
> anything except stdin/stdout/stderr. We can open an extra pipe() before
> calling run_command(), and make sure it is not marked CLOEXEC. I don't
> know if there are other portability concerns, though.

My suggestion to try "--status-fd=3" was meant to test whether the above
could help: If fd=3 helps, then our capturing stderr is not the problem.
(If fd=3 does not help then we still don't know...)

Michael