Re: [PATCH] pass transport verbosity down to git_connect

From: Eric Wong <normalperson@yhbt.net>
Date: Jan 29, 2016, 00:38 UTC
Message-ID: <20160129003816.GA26200@dcvr.yhbt.net>
In-Reply-To: <20160128235339.GB10308@sigill.intra.peff.net>

Jeff King <peff@peff.net> wrote:

Show 12 quoted lines

> On Thu, Jan 28, 2016 at 10:51:23PM +0000, Eric Wong wrote:
> > -static int connect_setup(struct transport *transport, int for_push, int verbose)
> > +static int connect_setup(struct transport *transport, int for_push)
> >  {
> >  	struct git_transport_data *data = transport->data;
> > +	int flags = transport->verbose > 0 ? CONNECT_VERBOSE : 0;
> 
> Do we want to trigger this only for "transport->verbose > 1"?
> 
> Right now, "git fetch -v" gives us a verbose status table (i.e.,
> includes up-to-date refs), but no more debugging than that. Should we
> reserve more debug-ish information like for "git fetch -vv"?

I'm not sure, I've never used "-v" at all in the past with fetch.

On one hand, I suspect someone who looks up "-v" and uses it is likely wondering: "why is it so slow?" At least, that's what I did before resorting to strace :)

On the other hand, I'm not sure if there's anything parsing the stderr out of "git fetch -v" right now. In that case, perhaps only changing "-vv" (and documenting it) is a better idea. I've always been of the opinion stderr is for humans and test suites, only; and not considered an interface somebody should be parsing.

For reference, "curl -v" includes connection info which I rely on all the time.

Junio:

  I'm leaning towards putting the test into t/t5570-git-daemon.sh
  to avoid potential port conflicts if that's OK with you.
  It doesn't seem like we have a lot of fetch tests on the git://
  at all.