Re: [PATCH 4/4] Implement git commit and status as a builtin commands.

On 2007.11.06 11:46:59 -0500, Kristian Høgsberg wrote:

On Tue, 2007-11-06 at 07:59 +0100, Björn Steinbrink wrote:
...

quoted

Note though, that Kristian had a similar check at the end of his email,
that included "only" (but lacked the bool conversion). The original
reason why I thought that it would be better was that for example
"git commit --all --only foo" didn't care about "only" at all. But that
actually was because the --all + paths usage check was broken. So the
fixed version actually refuses to use accept that, but with a (IMHO) not
so good error message:

$ git commit -a -o file
Paths with -a does not make sense.

Given that some people are used to just pass -a all the time, they might
just automatically pass it together with -o. And I think that we
actually want to tell them that -a + -o makes no sense instead. Just
like we do for -a + -i, which is kind of the complementary usage error.

So I'd go for a correct version of Kristian's suggestion:

if (!!also + !!only + !!all + !!interactive > 1)
	die("Only one of --include/--only/--all/--interactive can be used.");

Good points, I will use that in the next version of the patch.  Just a
note about the !! idiom (which I can't stand, fwiw): my version just
added the variables, which were all integers, initialized to zero and
incremented by the option parser when it sees the corresponding option.
So what I had would work too, with the extra check that:

  $ git commit -a -a

etc would give the error

  Only one of --include/--only/--all/--interactive can be used.

which is acutally accurate.

Hm, why? The user used only one of them. The error message does not say
that you cannot pass the same one multiple times. And I don't think that
passing the same boolean flag twice should be treated as an usage error
either. There's no contradiction in wanting all files and, well, all
files to be committed.

Sidenote: The "or mask" stuff in the option parser would probably
prevent you from catching "-a -a" anyway, because the flag becomes truly
boolean ;-)

Björn