Re: [PATCH 0/2] "git add -A/--no-all" finishing touches

Junio C Hamano [off-list ref] writes:

Jonathan Nieder [off-list ref] writes:

quoted

Junio C Hamano wrote:

quoted

At the point of calling warn_pathless_add(), it seems that we are
triggering this for paths that are not necessarily modified when run
with "add -n -u".

Do you mean files that were touched but have no content change, or
something more subtle?

I had the change (which by the way needs a fix for the "found a
directory" codepath) on top of master, uncommitted, and no other
change (I also have some cruft that is not ignored).

    cd Documentation && ../git add -n -u

reported GIT-VERSION-GEN which was not touched.  It does not
reproduce, though...

Ahh, I haven't run anything under the debugger yet, but I think I
know what is going on.

Don't we limit our "update-index --refresh" equivalent to the
original pathspec, even though your "-u/-A sans pathspec" warning
detection relies on grabbing the changes from the entire tree?

Junio C Hamano wrote:

quoted

Jonathan Nieder [off-list ref] writes:

quoted

quoted

Do you mean files that were touched but have no content change, or
something more subtle?

[...]

Ahh, I haven't run anything under the debugger yet, but I think I
know what is going on.

Don't we limit our "update-index --refresh" equivalent to the
original pathspec, even though your "-u/-A sans pathspec" warning
detection relies on grabbing the changes from the entire tree?

I think it's more basic than that.  "git add" doesn't bother to
run an "update-index --refresh" equivalent before its main loop
unless you pass --refresh to it, since reading files to compare
them to the index would be duplicated work.  The files hit in
update_callback() are only potentially modified.

Maybe the warning should happen after add_file_to_index() has run,
letting git compare the old and new index entries for that path?

Jonathan