Re: How to use git attributes to configure server-side checks?

On Thu, Sep 22, 2011 at 1:13 PM, Jeff King [off-list ref] wrote:

No, we definitely don't use in-tree gitattributes. IIRC, there are some
precedence and ordering questions. I think the ordering now is:

 1. Look in $GIT_DIR/info/attributes

 2. If not found, look in per-directory .gitattributes

 3. If not found, look in core.gitattributesfile

Where do in-tree attributes fit in? Between 1 and 2? Or 2 and 3? And
which tree do we look at?

attr.c says:

  a. built-in attributes
  b. $(prefix)/etc/gitattributes unless GIT_ATTR_NOSYSTEM is set
  c. core.attributesfile
  d. per-directory .gitattributes
  e. $GIT_DIR/info/attributes

The mechanics of (d) are established by git_attr_direction:

  GIT_ATTR_CHECKIN: working-copy, then index
  GIT_ATTR_CHECKOUT: index, then working-copy
  GIT_ATTR_INDEX: index-only

Where GIT_ATTR_CHECKIN is the default direction and GIT_ATTR_CHECKOUT
is used while checking-out files from the index. (GIT_ATTR_INDEX is
used only by git-archive.)

Note that (d) only occurs in non-bare repos or if direction is GIT_ATTR_INDEX.

Here are some examples:

 a. If I do "git checkout branch-foo", we should look at branch-foo's
    tree for things like crlf, right?  Do we still fall back to
    per-directory .gitattributes in the working tree? On the one hand,
    they're not really relevant to the commit we're moving to. But they
    are respected in the current code, and can be useful when moving to
    old commits which lack attributes.

    I think this is where the index magic comes in in the current code
    (we do something like "load the index, then respect gitattributes
    from the index").

    So maybe this is solved already.

 b. You're diffing commit $a against commit $b. Whose gitattributes
    have precedence? Is order important? Are gitattributes in the
    working tree and index relevant?

 c. You're diffing commit $a against HEAD. Which gitattributes have
    precedence? Again, is order important (i.e., does "diff -R" look at
    different attributes than "diff")?

I'm sure there are others, too. And I don't think any of these is
insurmountable. But somebody needs to think through a lot of cases and
come up with consistent, sane behavior that does the right thing in each
case (considering both bare repositories and ones with working trees).

I think that diff-index --cached or when there is no working tree,
should set the direction to GIT_ATTR_INDEX so that it may be used with
a bare repo. In such case, the .gitattributes would come from the
index.

Consistent with that, when comparing two commits (diff-tree), I think
you look at the .gitattributes in the second commit.

j.