It is careful by default and refuses to overwrite old info, but if you
want to force everything to be re-read, use the "-f" flag.
Some day I'll make it take individual filenames too. Right now
it's all-or-nothing.
Fix permissions, and add trivial "reset" and "add" scripts.
The "reset" script just resets the index back to head, while the "add"
script is just a crutch for people used to do "cvs add".
It's an incredibly cheesy helper that changes human-readable revision
arguments into the git-rev-list argument format.
You can use it to do something like this:
git-rev-list --pretty $(git-rev-parse --default HEAD "$@")
which is what git-log-script will become. Here git-rev-parse will
then allow you to use arguments like "v2.6.12-rc5.." or similar
human-readable ranges.
It's really quite stupid: "a..b" will be converted into "a" and "^b" if
"a" and "b" are valid object pointers. And the "--default" case will be
used if nothing but flags have been seen, so that you can default to a
certain argument if there are no other ranges.
The diff-* brothers acquired a sibling, git-diff-stages. With
an unmerged index file, you specify two stage numbers and it
shows the differences between them.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This gets the "cvs2git" program from the old git-tools
archive, and adds a nice script around it that makes it
much easier to use.
With this, you should be able to import a CVS archive
using just a simple
git cvsimport <cvsroot> <module>
and you're done. At least it worked for my one single test.
NOTE!! This may need tweaking. It currently expects (and
verifies) that cvsps version 2.1 is installed, but you
can't actually set any of the cvsps parameters, like the
time fuzz.
This patch adds code to read a hash out of a specified file under
{GIT_DIR}/refs/, and to write such files atomically and optionally with an
compare and lock.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch linearises the GIT commit history graph into merge order
which is defined by invariants specified in Documentation/git-rev-list.txt.
The linearisation produced by this patch is superior in an objective sense
to that produced by the existing git-rev-list implementation in that
the linearisation produced is guaranteed to have the minimum number of
discontinuities, where a discontinuity is defined as an adjacent pair of
commits in the output list which are not related in a direct child-parent
relationship.
With this patch a graph like this:
a4 ---
| \ \
| b4 |
|/ | |
a3 | |
| | |
a2 | |
| | c3
| | |
| | c2
| b3 |
| | /|
| b2 |
| | c1
| | /
| b1
a1 |
| |
a0 |
| /
root
Sorts like this:
= a4
| c3
| c2
| c1
^ b4
| b3
| b2
| b1
^ a3
| a2
| a1
| a0
= root
Instead of this:
= a4
| c3
^ b4
| a3
^ c2
^ b3
^ a2
^ b2
^ c1
^ a1
^ b1
^ a0
= root
A test script, t/t6000-rev-list.sh, includes a test which demonstrates
that the linearisation produced by --merge-order has less discontinuities
than the linearisation produced by git-rev-list without the --merge-order
flag specified. To see this, do the following:
cd t
./t6000-rev-list.sh
cd trash
cat actual-default-order
cat actual-merge-order
The existing behaviour of git-rev-list is preserved, by default. To obtain
the modified behaviour, specify --merge-order or --merge-order --show-breaks
on the command line.
This version of the patch has been tested on the git repository and also on the linux-2.6
repository and has reasonable performance on both - ~50-100% slower than the original algorithm.
This version of the patch has incorporated a functional equivalent of the Linus' output limiting
algorithm into the merge-order algorithm itself. This operates per the notes associated
with Linus' commit 337cb3fb8d.
This version has incorporated Linus' feedback regarding proposed changes to rev-list.c.
(see: [PATCH] Factor out filtering in rev-list.c)
This version has improved the way sort_first_epoch marks commits as uninteresting.
For more details about this change, refer to Documentation/git-rev-list.txt
and http://blackcubes.dyndns.org/epoch/.
Signed-off-by: Jon Seymour <jon.seymour@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In preparation for 1.0 release, this makes the command names
consistent with others in git-*-pull family.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Somebody finally came through - Jeff Garzik gets a gold
star for writing a shortlog script for git, so that I
can do nice release announcments again.
I added name translations from the current kernel history
(and git, for that matter). Hopefully it won't grow at
nearly the same rate the BK equivalent did, since 99% of
the time git records the full name already.
Usage: just do
git-rev-list --pretty HEAD ^LAST_HEAD | git-shortlog
or, in fact, use any of the other tools (git-diff-tree,
git-whatchanged etc) that use the default "pretty" commit format.
The "git" script is just shorthand: "git xyz <args>" will just execute
"git-xyz-script <args>", which is useful for people used to the CVS
naming convention. So "git log" will run the new git-log-script, which
is just a wrapper around the new pretty-printing git-rev-list.
Cheesy.
A new diffcore filter diffcore-order is introduced. This takes
a text file each of whose line is a shell glob pattern. Patches
that match a glob pattern on an earlier line in the file are
output before patches that match a later line, and patches that
do not match any glob pattern are output last.
A typical orderfile for git project probably should look like
this:
README
Makefile
Documentation
*.h
*.c
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This is meant to make raw git not hugely less usable than something
like raw CVS. I want to make a 1.0 release of the plumbing, and the
actual commit part was just too intimidating.
A new diffcore transformation, diffcore-break.c, is introduced.
When the -B flag is given, a patch that represents a complete
rewrite is broken into a deletion followed by a creation. This
makes it easier to review such a complete rewrite patch.
The -B flag takes the same syntax as the -M and -C flags to
specify the minimum amount of non-source material the resulting
file needs to have to be considered a complete rewrite, and
defaults to 99% if not specified.
As the new test t4008-diff-break-rewrite.sh demonstrates, if a
file is a complete rewrite, it is broken into a delete/create
pair, which can further be subjected to the usual rename
detection if -M or -C is used. For example, if file0 gets
completely rewritten to make it as if it were rather based on
file1 which itself disappeared, the following happens:
The original change looks like this:
file0 --> file0' (quite different from file0)
file1 --> /dev/null
After diffcore-break runs, it would become this:
file0 --> /dev/null
/dev/null --> file0'
file1 --> /dev/null
Then diffcore-rename matches them up:
file1 --> file0'
The internal score values are finer grained now. Earlier
maximum of 10000 has been raised to 60000; there is no user
visible changes but there is no reason to waste available bits.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The second round similarity estimator simply used the size of
the xdelta itself to estimate the extent of damage. This patch
keeps that logic to detect big insertions to terminate the check
early, but otherwise looks at the generated delta in order to
estimate the extent of edit more accurately.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This applies git patches (and old-style unified diffs)
in the index, rather than doing it in the working directory.
That allows for a lot more flexibility, and means that if a
patch fails, we aren't going to mess up the working directory.
NOTE! This is just the first cut at it, and right now it only
parses the incoming patch, it doesn't actually apply it yet.
Thomas Glanzmann points out that it doesn't work well with different
clients accessing the repository over NFS - they have different views
on what the "device" for the filesystem is.
Of course, other filesystems may not even have stable inode numbers.
But we don't care. At least for now.
This moves the path selection logic from individual programs to a new
diffcore transformer (diff-tree still needs to have its own for
performance reasons). Also the header printing code in diff-tree was
tweaked not to produce anything when pickaxe is in effect and there is
nothing interesting to report. An interesting example is the following
in the GIT archive itself:
$ git-whatchanged -p -C -S'or something in a real script'
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This steals the "pickaxe" feature from JIT and make it available
to the bare Plumbing layer. From the command line, the user
gives a string he is intersted in.
Using the diff-core infrastructure previously introduced, it
filters the differences to limit the output only to the diffs
between <src> and <dst> where the string appears only in one but
not in the other. For example:
$ ./git-rev-list HEAD | ./git-diff-tree -Sdiff-tree-helper --stdin -M
would show the diffs that touch the string "diff-tree-helper".
In real software-archaeologist application, you would typically
look for a few to several lines of code and see where that code
came from.
The "pickaxe" module runs after "rename/copy detection" module,
so it even crosses the file rename boundary, as the above
example demonstrates.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This introduces the diff-core, the layer between the diff-tree
family and the external diff interface engine. The calls to the
interface diff-tree family uses (diff_change and diff_addremove)
have not changed and will not change. The purpose of the
diff-core layer is to provide an infrastructure to transform the
set of differences sent from the applications, before sending
them to the external diff interface.
The recently introduced rename detection code has been rewritten
to use the diff-core facility. When applications send in
separate creates and deletes, matching ones are transformed into
a single rename-and-edit diff, and sent out to the external diff
interface as such.
This patch also enhances the rename detection code further to be
able to detect copies. Currently this happens only as long as
copy sources appear as part of the modified files, but there
already is enough provision for callers to report unmodified
files to diff-core, so that they can be also used as copy source
candidates. Extending the callers this way will be done in a
separate patch.
Please see and marvel at how well this works by trying out the
newly added t/t4003-diff-rename-1.sh test script.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This adds the ability to actually create delta objects using a new tool:
git-mkdelta. It uses an ordered list of potential objects to deltafy
against earlier objects in the list. A cap on the depth of delta
references can be provided as well, otherwise the default is to not have
any limit. A limit of 0 will also undeltafy any given object.
Also provided is the beginning of a script to deltafy an entire
repository.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This adds knowledge of delta objects to fsck-cache and various object
parsing code. A new switch to git-fsck-cache is provided to display the
maximum delta depth found in a repository.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With -u flag, git-checkout-cache picks up the stat information
from newly created file and updates the cache. This removes the
need to run git-update-cache --refresh immediately after running
git-checkout-cache.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch adds the basic library functions to create and replay delta
information. Also included is a test-delta utility to validate the
code.
diff-delta was based on LibXDiff written by Davide Libenzi
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The documentation of the test harness still refer to old
numbering and also contains an obvious typo.
Also "make test" should be run after making sure we have built
all binaries, since test is designed to test the newly built
ones.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Petr Baudis <pasky@ucw.cz>
It used to be that diff-tree needed helper support to parse its
raw output to generate diffs, but these days git-diff-* family
produces the same output and the helper is not tied to diff-tree
anymore. Drop "tree" from its name.
This commit is done separately to record just the rename and no
file content changes. The changes in the renamed files are recorded
in the next commit.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Bundled with the changes in the unrenamed files.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
Mark Allen had trouble with building GIT on his Darwin and
posted a patch to link with -lcrypto instead of -lssl on Darwin.
Later Daniel Barkalow suggested to change it for everybody who
uses openssl, because the relevant functionality is in -lcrypto
not in -lssl, and the current linking happens to work only
because -lssl pulls in -lcrypto.
Signed-off-by: Junio C Hamano <junkio@cox.net>
H. Peter Anvin mentioned that using SHA1_whatever as an
environment variable name is not nice and we should instead use
names starting with "GIT_" prefix to avoid conflicts. Here is
what this patch does:
* Renames the following environment variables:
New name Old Name
GIT_AUTHOR_DATE AUTHOR_DATE
GIT_AUTHOR_EMAIL AUTHOR_EMAIL
GIT_AUTHOR_NAME AUTHOR_NAME
GIT_COMMITTER_EMAIL COMMIT_AUTHOR_EMAIL
GIT_COMMITTER_NAME COMMIT_AUTHOR_NAME
GIT_ALTERNATE_OBJECT_DIRECTORIES SHA1_FILE_DIRECTORIES
GIT_OBJECT_DIRECTORY SHA1_FILE_DIRECTORY
* Introduces a compatibility macro, gitenv(), which does an
getenv() and if it fails calls gitenv_bc(), which in turn
picks up the value from old name while giving a warning about
using an old name.
* Changes all users of the environment variable to fetch
environment variable with the new name using gitenv().
* Updates the documentation and scripts shipped with Linus GIT
distribution.
The transition plan is as follows:
* We will keep the backward compatibility list used by gitenv()
for now, so the current scripts and user environments
continue to work as before. The users will get warnings when
they have old name but not new name in their environment to
the stderr.
* The Porcelain layers should start using new names. However,
just in case it ends up calling old Plumbing layer
implementation, they should also export old names, taking
values from the corresponding new names, during the
transition period.
* After a transition period, we would drop the compatibility
support and drop gitenv(). Revert the callers to directly
call getenv() but keep using the new names.
The last part is probably optional and the transition
duration needs to be set to a reasonable value.
Signed-off-by: Junio C Hamano <junkio@cox.net>
On Solaris machines gnu install called ginstall
<JC> Editorial notes. I've also changed it to use $(COPTS), $(prefix),
and $(bin) because I always get confused without compiling it with -O1
when I single step in gdb. The default is left as Linus shipped.
Date: Sat, 7 May 2005 10:41:54 +0200
Signed-off-by: Thomas Glanzmann <sithglan@stud.uni-erlangen.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Separate out the merge resolve from the actual getting of the
data. Also, update the resolve phase to take advantage of the
fact that we don't need to do the commit->tree object lookup
by hand, since all the actors involved happily just act on a
commit object these days.
A new command, git-write-blob, is introduced. This registers
the contents of any file on the filesystem as a blob in the
object database and reports its SHA1 to the standard output.
To implement it, the patch promotes index_fd() from a static
function in update-cache.c to extern and moves it to a library
source, sha1_file.c.
This command is used to update git-merge-one-file-script so that
it does not smudge the work tree.
Signed-off-by: Junio C Hamano <junkio@cox.net>
This adds the git-local-pull command as a smaller brother of
http-pull and rpull.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I said:
- Stop attempting to be compatible with cg-patch, and drop
(mode:XXXXXX) bits from the diff.
- Do keep the /dev/null change for created and deleted case.
- No "Index:" line, no "Mode change:" line, anywhere in the
output. Anything that wants the mode bits and sha1 hash can
do things from GIT_EXTERNAL_DIFF mechanism. Maybe document
suggested usage better.
This adds an example script git-apply-patch-script, that can be
used as the GIT_EXTERNAL_DIFF to apply changes between two trees
directly on the current work tree, like this:
GIT_EXTERNAL_DIFF=git-apply-patch-script git-diff-tree -p <tree> <tree>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The method for deciding what to pull is useful separately from any of the
ways of actually fetching the objects.
So split out "pull" functionality from http-pull and rpull
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
...since everything out there is either strange (libc mktime has issues
with timezones) or introduces unnecessary dependencies for people (libcurl).
This goes back to the old date parsing, but moves it out into a file of
its own, and does the "struct tm" to "seconds since epoch" handling by
hand.
I grepped through the tz-database and it seems there's one "country"
left that has non-60-minute DST: Lord Howe Island. All others dropped
that before 1970.
This switches git-commit-tree to using curl_getdate() for the
AUTHOR_DATE, and thus fixes the problem with "mktime()" parsing dates in
the local timezone. It also ends up being more permissive about the
format of the date.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Here is another. This one belongs to a clean-up category.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This also regularizes the make. The source files themselves don't get
the "git-" prefix, because that's just inconvenient. So instead we just
make the rule that "git-xxxx" depends on "xxxx.c", and do that for
all the core programs (ie the old "git-mktag.c" got renamed to just
"mktag.c" to match everything else).
And "show-diff" got renamed to "git-diff-files" while at it, since
that's what it really should be to match the other git-diff-xxx cases.
This is an improved version of tar-tree, a streaming archive creator for
GIT. The major added feature is blocking; all write(2) calls now have a
size of 10240, just as GNU tar (and tape drives) likes them. The
buffering overhead does not seem to degrade performance because most
files in the repositories I tested this with are smaller than 10KB, so
we need fewer system calls.
File names are still restricted to 500 bytes and the archive format
currently only allows for files up to 8GB. Both restrictions can be
lifted if need be with more pax extended headers.
The archive format used is the pax interchange format, i.e. POSIX tar
format. It can be read by (and created with) GNU tar. If I read the
specs correctly tar-tree should now be standards compliant (modulo
bugs).
Because it streams the archive (think ls-tree merged with cat-file),
tar-tree doesn't need to create any temporary files. That makes it
quite fast.
It accepts tree IDs and commit IDs as first parameter. In the latter
case tar-tree tries to get the commit date out of the committer line.
Else all files in the archive are time-stamped with the current time.
An optional second parameter is used as a path prefix for all files in
the archive. Example:
$ tar-tree a2755a80f40e5794ddc20e00f781af9d6320fafb \
linux-2.6.12-rc3 | bzip9 -9 > linux-2.6.12-rc3.tar.bz2
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This adds preliminary support for tags in the library. It doesn't even
store the signature, however, let alone provide any way of checking it.
Signed-Off-By: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch introduces a new program, diff-tree-helper. It reads
output from diff-cache and diff-tree, and produces a patch file.
The diff format customization can be done the same way the
show-diff uses; the same external diff interface introduced by
the previous patch to drive diff from show-diff is used so this
is not surprising.
It is used like the following examples:
$ diff-cache --cached -z <tree> | diff-tree-helper -z -R paths...
$ diff-tree -r -z <tree1> <tree2> | diff-tree-helper -z paths...
- As usual, the use of the -z flag is recommended in the script
to pass NUL-terminated filenames through the pipe between
commands.
- The -R flag is used to generate reverse diff. It does not
matter for diff-tree case, but it is sometimes useful to get
a patch in the desired direction out of diff-cache.
- The paths parameters are used to restrict the paths that
appears in the output. Again this is useful to use with
diff-cache, which, unlike diff-tree, does not take such paths
restriction parameters.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With this patch, the non-core'ish part of show-diff command that
invokes an external "diff" comand to obtain patches is split
into a separate file. The next patch will introduce a new
command, diff-tree-helper, which uses this common diff interface
to format diff-tree and diff-cache output into a patch form.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch adds three similar and related programs. http-pull downloads
objects from an HTTP server; rpull downloads objects by using ssh and
rpush on the other side; and rpush uploads objects by using ssh and rpull
on the other side.
The algorithm should be sufficient to make the network throughput required
depend only on how much content is new, not at all on how much content the
repository contains.
The combination should enable people to have remote repositories by way of
ssh login for authenticated users and HTTP for anonymous access.
Signed-Off-By: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Here is a SHA1 implementation with the core written in PPC assembly.
On my 2GHz G5, it does 218MB/s, compared to 135MB/s for the openssl
version or 45MB/s for the mozilla version.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This one includes the Mozilla SHA1 implementation sent in by Edgar Toernig.
It's dual-licenced under MPL-1.1 or GPL, so in the context of git, we
obviously use the GPL version.
Side note: the Mozilla SHA1 implementation is about twice as fast as the
default openssl one on my G5, but the default openssl one has optimized
x86 assembly language on x86. So choose wisely.
Use a generic rule for executables that depend only on the corresponding
.o and on $(LIB_FILE).
Signed-Off-By: Andre Noll <maan@systemlinux.org>
Signed-Off-By: Linus Torvalds <torvalds@osdl.org>
the current cache state and/or working directory.
Very useful to see what has changed since the last commit, either in
the index file or in the whole working directory.
Also very possibly very buggy. Matching the two up is not entirely
trivial.
as a set of patches and commentary.
You'd want something like this if you are tracking a git archive
in another SCM format. Notably, we want something like that for
BK users.
This introduces the concept of git "library" objects that
the real programs use, and makes it easier to add such things
to a "libgit.a".
This will also make it trivial to split the current "read-cache.o"
into more aptly named pieces (it does a lot more than just read
the index file).
This switches to my implementation of merge-base, but with the new parsing
library.
Signed-Off-By: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This ports fsck-cache to use parsing functions. Note that performance
could be improved here by only reading each object once, but this requires
somewhat more complicated flow control.
Signed-Off-By: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This ports rev-tree to use the parsing functions introduced in the
previous patches.
Signed-Off-By: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
parent of two commits.
The question of "best" commit can probably be tweaked almost arbitrarily.
In particular, trying to take things like how big the tree differences
are into account migt be a good idea. This one is just very simple.
It's there in the history if somebody wants to resurrect it, but it
seems to have been successfully superceded by the new and improved
index-merge thing, where we do all merging entirely in the index.
Ancient cat-file command used to leave temp_git_file_* and there
was support to remove them in the clean target of Makefile. I
do not think it is needed anymore.
From: Junio C Hamano <junkio@cox.net>
Signed-off-by: Petr Baudis <pasky@ucw.cz>
The nsec field of ctime/mtime is now checked only with -DNSEC defined during
compilation. nsec acts broken since it is stored in the icache but apparently
just gets to zero when flushed to filesystem not supporting it (e.g. ext3),
creating illusions of false changes. At least that's my impression.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
When compiled with -DCOLLISION_CHECK, we will check against SHA1
collisions when writing to the object database.
From: Christopher Li <chrislgit@chrisli.org>
Signed-off-by: Petr Baudis <pasky@ucw.cz>
ls-tree tool provides just a way to export the binary tree objects
to a usable text format. This is bound to be useful in variety
of scripts, although none of those I have currently uses it.
But e.g. the simple script I've sent to HPA for purging the object
database uses it.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
tree graph.
It's quite fast when the commit-objects are cached, but since
it has to walk every single commit-object, it also allows you
to cache an old state and just add on top of that.
