Re: [PATCH 1/2] Fix "git diff" setup code

Linus Torvalds [off-list ref] writes:

On Fri, 14 Sep 2007, Junio C Hamano wrote:

quoted

So I tend to think the attached is a better fix.

Ahh, yes, that explains the conditional. 

But whatever gets us to actually verify our options, and fill in the right 
defaults is ok by me!

Sorry, my explanation only explains about missing setup_done()
when --no-index is used, but does not explain _if_ you actually
found that setup_done() was not called for you when you did a
real life test.  Was it only from code inspection, or did you
hit a case where setup_done() is not run?  If the latter then
there is something else going on, as I cannot think of a way to
call setup_revisions() and not have it call setup_done()...

On Fri, 14 Sep 2007, Junio C Hamano wrote:

Sorry, my explanation only explains about missing setup_done()
when --no-index is used, but does not explain _if_ you actually
found that setup_done() was not called for you when you did a
real life test.  Was it only from code inspection, or did you
hit a case where setup_done() is not run?

Hmm. Mea culpa. What seems to have happened is that I ran things under 
gdb, and noticed that the default rename_limit hadn't been correctly set: 
but now that I look more at it, that particular session was probably from 
"git runstatus", not "git diff".

So yeah, ignore my 1/2. It was almost certainly based on a bogus debugging 
session, before I noticed that wt-status.c doesn't use the normal diff 
stuff at all..

			Linus