From: Stefan Sperling <stsp@stsp.name>
Subject: Re: tog(1) log by default++
To: Martin Pieuchot <mpi@openbsd.org>, gameoftrees@openbsd.org
Date: Fri, 14 Feb 2020 17:15:32 +0100

On Fri, Feb 14, 2020 at 04:59:32PM +0100, Stefan Sperling wrote:
> On Fri, Feb 14, 2020 at 04:39:48PM +0100, Martin Pieuchot wrote:
> > If the second argument on the command line doesn't match a command name
> > let the tool to assume it's a path.  In other words the following:
> > 
> > 	$ tog kern/kern_sync.c
> > 
> > Becomes an alias for:
> > 	
> > 	$ tog log kern/kern_sync.c
> > 
> > This is similar to what tig(1) does and it helps me being lazy :o)
> 
> This behaviour already existed once and was reverted in
> commit 3642c4c6513e3536dc77e8f2b7a2402d1aa916a7
> 
> I don't recall why I reverted it. I suspect it was to keep the
> 'command subcommand' syntax consistent across both got(1) and tog(1).
> And I didn't like the old implementation I had, but yours looks cleaner.

One problem is that 'tog foo' for non-existent foo no longer shows usage
information. It says "tog: no git repository found" or "tog: no such entry
found in tree" depending on the contents of the current working directory,
without any further hints.

Perhaps a custom error message could be added which says something like
"tog: 'foo' is neither a known command nor a path" when it fails, followed
by usage info?