Re: [PATCH] Add the diff option --no-defaults

Dscho,

On Sat, 18 Apr 2009, Johannes Schindelin wrote:

On Sat, 18 Apr 2009, Keith Cascio wrote:

quoted

On Fri, 17 Apr 2009, Johannes Schindelin wrote:

quoted

Besides, you still will have a poison:
	git config diff.defaultOptions --no-defaults
which is Russel's paradoxon right there.

I can cleanly modify my v3 to handle this case.

You cannot.  --no-defaults means that diff.defaultOptions should be 
disregarded.  If the diff.defaultOptions say that they should be disregarded 
themselves, then --no-defaults should be disregarded.

--no-defaults *could* mean as you say there.  But a much better meaning for 
--no-defaults is: suppress the values in diff.defaultOptions after options 
processing.  We don't have to disregard them, just suppress them after options 
processing.  In that sense, --no-defaults is a meta-option.  It is an option 
about options.  Even users unfamiliar with set theory would assume the 
suppression semantics.

Nevertheless I applaud the Russell reference.  Very intriguing.

And I still do not like the intrusiveness of your patch.  The last time we did 
something like that with options (some parseoptifications), we had a lot of 
fallout as a consequence.

A reasonable worry!  But blind paranoia is paralyzing.  Peff expressed some 
specific concerns which he and I addressed: (1) whether I'd investigated all 
callsites for possible problems (yes I did), (2) whether we'd have to switch 
every callsite to a macro, rather than direct assignment (no we don't).  
Outside of diff.h/diff.c, my v3 deletes no lines and adds only two.  That 
doesn't really seem "intrusive" to me.  By comparison, your patch adds at least 
ten lines outside of diff.h/diff.c.  I'd rather call my patch "innovative".  
Possible?
                               -- Keith