Re: [PATCH 1/1] _gpg: Use explicit UIDs for state = public keys.

[Doron Behar <doron.behar@xxxxxxxxx>]

That didn't work.. I debugged this a little bit and I discovered that
setting IFS=$'\n' is what makes this problem. Perhaps should it be
reported as a bug of `_call_program`?

On Wed, May 30, 2018 at 07:04:14PM +0000, Daniel Shahaf wrote:
> Doron Behar wrote on Tue, May 29, 2018 at 17:11:23 +0300:
> > I've successfully managed to solve address all the comments you've had
> > for my patch. Yet, I'm having trouble with `_call_program`. The line I'm
> > trying to put is this:
> > 
> >     local public_keys=(${(@s.:.)$(_call_program public-keys $words[1] $needed --list-public-keys --list-options no-show-photos --with-colons)})
> > 
> > I debugged this a little bit and the variable `$public_keys` is empty
> > but if I remove the words: `_call_program public-keys` it works good.
> > 
> > I've had a glimpse at `htop` when this function was called when I tested
> > the completion function and I saw these commands running there:
> > 
> > 	pkgfile -b -v -- --list-public-keys
> > 	pkgfile -b -v -- --list-options
> > 	pkgfile -b -v -- no-show-photos
> > 	pkgfile -b -v -- --with-colons
> > 
> > What is _call_program supposed to be doing?
> > 
> 
> _call_program (doc'd in zshcompsys(1)) is supposed to run the command
> "$words[1] $needed --list-public-keys --list-options no-show-photos
> --with-colons".  The purpose of using _call_program is to allow the user to
> override the command to use by setting a zstyle.
> 
> I can't explain the multiple commands you glimpsed in htop.
> 
> Try changing «$words[1] $needed» to «${(q)words[1]} ${(q)needed}».  That adds
> an additional level of quoting, since one of the differences between «foo bar
> baz…» and «_call_program x foo bar baz…» is that in the latter case, the «foo
> bar baz» are eval'd.  I'm not sure whether that would fix the failure mode
> you're seeing, but it's a correct change regardless.
> 
> I would recommend debugging that line inside out, e.g., start by ensuring the
> _call_program by itself gets the expected arguments and produced the expected
> values on its stdout.  Enabling tracing («set -x» or «functions -T _call_program»)
> might help.
> 
> > Besides that, the revised patch is ready. I've also included similar
> > improvements for the other states - `secret-keys` and `public-keyids`.
> 
> I look forward to the revised patch.  However, I'll be busy in the next few
> days so I hope another developer would beat me to reviewing/applying it ;-).
> 
> Cheers,
> 
> Daniel