The latest change to diff-tree -z output adds an extra line
termination after non diff-raw material (the header and the
commit message).  To compensate for this change, stop adding the
output termination of our own.  "diff-tree -v -z" piped to
"diff-helper -z" would give different result from "diff-tree -v"
piped to "diff-helper" without this change.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This attempts to optimize "diff-tree -[CM] --stdin", which
compares successible tree pairs.  This optimization does not
make much sense for other commands in the diff-* brothers.

When reading from --stdin and using rename/copy detection, the
patch makes diff-tree to read the current index file first.
This is done to reuse the optimization used by diff-cache in the
non-cached case.  Similarity estimator can avoid expanding a
blob if the index says what is in the work tree has an exact
copy of that blob already expanded.

Another optimization the patch makes is to check only file sizes
first to terminate similarity estimation early.  In order for
this to work, it needs a way to tell the size of the blob
without expanding it.  Since an obvious way of doing it, which
is to keep all the blobs previously used in the memory, is too
costly, it does so by keeping the filesize for each object it
has already seen in memory.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This changes the way how pathspec is used in the three diff-*
brothers.  Earlier, they tried to grab as much information from
the original input and used pathspec to limit the output.  This
version uses pathspec upfront to narrow the world diffcore
operates in, so "git-diff-* <arguments> some-directory" does not
look at things outside the specified subtree when finding
rename/copy or running pickaxe.

Since diff-tree already takes this view and does not feed
anything outside the specified directotires to begin with, this
patch does not have to touch that command.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Earier version of diffcore-rename used to keep unmodified
filepair in its output so that the last stage of the processing
that tells renames from copies can make all of rename/copy to
copies.  However this had a bad interaction with other diffcore
filters that wanted to run after diffcore-rename, in that such
unmodified filepair must be retained for proper distinction
between renames and copies to happen.

This patch fixes the problem by changing the way diffcore-rename
records the information needed to distinguish "all are copies"
case and "the last one is a rename" case.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

When --pickaxe-all is given in addition to -S, pickaxe shows the
entire diffs contained in the changeset, not just the diffs for
the filepair that touched the sought-after string.  This is
useful to see the changes in context.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Earlier rename/copy detection left unmodified filepair in the
output and forced downstream to keep them even when they are
filtering, and the diff_needs_to_stay() function was used for
the logic.  It is not used anymore, so remove it.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This changes the argument of diff_setup() from an integer that
says if we are feeding reversed diff to a bitmask, so that later
global options can be added more easily.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This change makes the implementation of git-external-diff-script
cleaner.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Instead we can normalize what diff-raw records at the diffcore
side.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This removes a field that is no longer used from diff_score
structure.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Earlier it had a misguided attempt to include paths that matches
either source tree or destination tree after the rename/copy
detection.  The new semantics will be that pathspec defines a
narrowed down world the diffcore operates in, so it should not
even look at where in the source tree the path came from.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This introduces a new function to free a common data structure,
and plugs some leaks.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The math to reject delta that is too big was confused.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Signed-off-by: NThomas Glanzmann <sithglan@stud.uni-erlangen.de>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

check_file_directory_conflict can give the wrong answers. This is
because the wrong length is passed to cache_name_pos. The length
passed should be the length of the whole path from the root, not
the length of each path subcomponent.

	$ git-init-db
	defaulting to local storage area
	$ mkdir path && touch path/file
	$ git-update-cache --add path/file
	$ rm path/file
	$ mkdir path/file && touch path/file/f
	$ git-update-cache --add path/file/f  <-- Conflict ignored
	$
Signed-off-by: NDavid Meybohm <dmeybohmlkml@bellsouth.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

No need to make them multiple lines, in fact we explicitly don't want that.

This also fixes a 64-bit problem pointed out by Markus F.X.J. Oberhumer,
where we gave "%.*s" a "ptrdiff_t" length argument instead of an "int".

The recent rewrite broke "git-whatchanged -v -p drivers/usb/" but
"git-whatchanged -v -p drivers/usb" still works.  Just strip out the
trailing slashes internally to make it work again.

It uses compare-thing-with-number comparison order instead of visual
comparison order ;-).
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

diff-tree does the culling of uninteresting paths internally, and
fundamentally has to do so for performance reasons. So there's no
point in calling the separate pathname culling logic here,
especially as it seems slightly broken.

Actually use GIT_COMMITTER_DATE in git-commit-tree.

(It used to mistakenly re-use the author date)

Add test-case for it.
Signed-off-by: NRene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Use SP as the column separator except the ones before path which
uses TAB, to make the output format consistent across ls-* and
diff-* commands.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Enhance git-ls-tree to allow optional 'match paths' that
restricts the output of git-ls-tree.  This is useful to retrieve
a single file's SHA1 out of a tree without creating an index.

[JC: I added the test case]
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This is a demonstration of GIT_EXTERNAL_DIFF mechanism, and a
testbed for tweaking and enhancing what the built-in diff should
do.  This script is designed to output exactly the same output
as what the built-in diff driver produces when used as the
GIT_EXTERNAL_DIFF command.

I've run this and updated built-in diff on the entire history of
linux-2.6 git repository, and JG's udev.git repository which has
interesting symlink cases to make sure it is equivalent to the
built-in diff driver.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

With the introduction of type 'T' in the diff-raw output, and
the "apply-patch" program Linus has been quietly working on
without much advertisement, it started to make sense to emit
usable information in the "diff --git" patch output format as
well.  Earlier built-in diff driver punted and did not say
anything about a symbolic link changing into a file or vice
versa, but this version represents it as a pair of deletion
and creation.

It also fixes a minor problem dealing with old archive created
with ancient git.  The earlier code was reporting file mode
change between 100664 and 100644 (we shouldn't).  The linux-2.6
git tree has a good example that exposes this problem.  A good
test case is commit ce1dc02f76432a46db149241e015a4f782974623.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

It currently only verifies the index against the working directory,
it doesn't actually verify the diff fragments themselves yet.

This means that filenames are totally unambiguous even if they
have spaces or tabs in them.

Pure mode changes, and deletes or creates of empty files won't have this
information anywhere else.

Slowly this is takign the form of a program that we'd actually
use. Now "git-apply --stat" basically ends up being a perfectly
useful diffstat.

Hey, it's almost free by now, and it's a good way to see that
we parse the patches correctly.

Make it a clear two-phase thing: first a read-only parse of
the patch itself (which is independent of any current index
information), and then the second phase actually uses the patch.

The second phase might not be a real apply, it could be just a
diffstat, for example. Which is trivial to do once the patch is
parsed.

This is the remainder of testcase fix by Mark Allen to make them
work on his Darwin box.  I was using "xargs -r" (GNU) where it
was not needed, sed -ne '/^\(author\|committer\)/s|>.*|>|p'
where some sed does not know what to do with '\|', and also
"cmp - file" to compare standard input with a file, which his
cmp does not support.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

And teach fsck to free it to save memory.

The "end" commit is just faking it right now, it's sorting things
purely by date, so this is _not_ a reachability analysis. Some day.

The "--header" flag causes the commit message to be printed out,
with a NUL character separator after it for parseability. This
allows you to do things like use "grep -z" to grep for certain
authors etc.

object.

A fair number of the users potentially want to look at the
commit objects more closely, and if you worry about memory
leaking in certain applications, you can always do a

	free(commit->buffer);
	commit->buffer = NULL;

by hand after parsing them.

This makes them easier to parse without having to worry about
running off the end, and allows us to treat commits as normal
strings.

This fixes another bug.

 - Mode-only changes were pruned incorrectly from the output.
 - Added test to catch the above problem.
 - Normalize rename/copy similarity score in the diff-raw output
   to per-cent, no matter what scale we internally use.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The interim single-liner '?' fix resulted delete entries that
should not have emitted coming out in the output as an
unintended side effect; I caught this with the "rename" test in
the test suite.  This patch instead fixes the code that assigns
the status code to each filepair.

I verified this does not break the testcase in udev.git tree Kay
Sievers gave us, by running git-diff-tree on that tree which
showed 21 file to symlink changes.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The earlier test was relying on the fact that dotfiles do not
appear in the output to prepare expected test results, which
inevitably got broken when we started handling dotfiles.  Change
the test to be honest about what "--other" file it creates.

The problem was originally pointed out by Mark Allen.
Signed-off-by: NJunio C Hamano <junkio@cox.net>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

This also drops the common ".git" part from the end of the repo
name, and if a non-default head reference is given, makes a nicer
commit message about it.

This change was suggested for my git-switch-tree script, and the same
issues apply to core git's git-resolve-script as well.

Give them an "unknown" status, ie '?'.