Re: [PATCH] diff: alias -q to --quiet
- From
Jeff King <peff@peff.net>
- Date
- Oct 13, 2017, 17:56 UTC
- Message-ID
- <20171013175656.qitarz6lvrppyb5f@sigill.intra.peff.net>
- In-Reply-To
- <20171013164415.6632-1-asottile@umich.edu>
On Fri, Oct 13, 2017 at 09:44:15AM -0700, Anthony Sottile wrote:
> Previously, `-q` was silently ignored:
I'm not sure if is totally ignored. Normally if we have an unknown options we'd complain:
$ git diff -x error: invalid option: -x
but we don't with "-q". Why?
In builtin/diff.c:471, we can see:
else if (!strcmp(argv[1], "-q"))
options |= DIFF_SILENT_ON_REMOVED;So it _does_ do something, just not what you expected.
But wait, that's not the whole story. We convert "-q" into SILENT_ON_REMOVED in git-diff-files and in git-diff (when we're acting like diff-files). But nobody ever seems to check it!
Running "git log -p -SSILENT_ON_REMOVED" turns up two interesting commits:
- 95a7c546b0 (diff: deprecate -q option to diff-files, 2013-07-17)
- c48f6816f0 (diff: remove "diff-files -q" in a version of Git in a
distant future, 2013-07-18).So we dropped "-q" a few years ago and added a deprecation notice. "git tag --contains 95a7c546b0" tells us that happened in v1.8.5, which shipped in Nov 2013.
And then in v2.0.0 (May 2014) we tried to drop "-q" completely. Looking over c48f6816f0, I _think_ it's a mistake that "-q" became a silent noop there. That commit should have ripped out the remaining bits that set the SILENT_ON_REMOVED flag, and "-q" would have become an error.
So there are two separate questions/tasks:
1. Should we remove the special handling of "-q" leftover from this
deprecation? I think the answer is yes. 2. Should we teach the diff machinery as a whole to treat "-q" as a
synonym for "--quiet". Probably yes, but it's less clear to me that this won't have funny
interactions. Are there other commands which use the diff-options
parser via setup_revisions(), but expect "-q" to be left in the
output so that they can handle it themselves? It looks like git-log does so, but it pulls the "-q" out before
handing the remainder to setup_revisions(). And anyway, it just
converts the option into a quiet diff (though it does in a way
that's different than the rest of the diff code -- that might bear
investigating on its own). Grepping for 'q' and OPT__QUIET, I don't see any others, but I
didn't spend very much time digging.-Peff