git-apply: be a lot more careful when writing files
We write them under another name and rename them to their destination,
so that if something bad happens in the middle, we won't have caused any
bigger harm.
Also, this makes the writing be NFS "intr" safe, and as a side effects
makes sure that if the target is hardlinked (or symlinked) we will have
broken the link.
Junio C Hamano [Wed, 13 Jul 2005 19:45:51 +0000 (12:45 -0700)]
[PATCH] git-diff-*: --name-only and --name-only-z.
Porcelain layers often want to find only names of changed files,
and even with diff-raw output format they end up having to pick
out only the filename. Support --name-only (and --name-only-z
for xargs -0 and cpio -0 users that want to treat filenames with
embedded newlines sanely) flag to help them.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
That way we avoid any confusion with "GNU Interactive Tools", and it's
more descriptive anyway (the rpm documentation talks about how git is
split into a "core" part and an "SCM" part, this makes it clear that
this is the core one).
and it creates a new branch of name <branchname>. If a starting point
is specified, that will be where the branch is created, otherwise it
will be created at the current HEAD.
Make "git checkout" verify that the argument refers to a commit
We still need to create a new branch if it didn't refer to an existing
branch, otherwise our HEAD will continue to point to something totally
different than what we just checked out.
I'll need to think about it. Maybe only do it with "-f" and force it to
the "master" branch?
git-rev-parse: Allow a "zeroth" parent of a commit - the commit itself.
This sounds nonsensical, but it's useful to make sure that the result is
a commit.
For example, "git-rev-parse v2.6.12" will return the _tag_ object for
v2.6.12, but "git-rev-parse v2.6.12^0" will return the _commit_ object
associated with that tag (and v2.6.12^1 will return the first parent).
Also, since the "parent" code will actually parse the commit, this,
together with the "--verify" flag, will verify not only that the result
is a single SHA1, but will also have verified that it's a proper commit
that we can see.
Cut-and-paste dup noticed by Junio. It's not even harmless, since a
match also causes that match to be invalidated, so this made it
impossible to update an existing branch by name.
I'd only tested the case of "ref doesn't exist at all on the other end",
which worked fine.
Junio C Hamano [Mon, 11 Jul 2005 20:30:54 +0000 (13:30 -0700)]
[PATCH] git-clone-script local optimization tweaks
- When local optimization is used, the variable repo has
already been passed through get_repo_base so there is no need
to check for .git subdirectory in there.
- Use cpio -l instead of "cp -l".
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Junio C Hamano [Mon, 11 Jul 2005 07:00:55 +0000 (00:00 -0700)]
[PATCH] Check packs and then files.
This reverses the order of object lookup, to check pack index first and
then go to the filesystem to find .git/objects/??/ hierarchy.
When most of the objects are packed, this saves quite many stat() calls
and negative dcache entries; while the price this approach has to pay is
negligible, even when most of the objects are outside pack, because
checking pack index file is quite cheap.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Junio C Hamano [Mon, 11 Jul 2005 06:55:56 +0000 (23:55 -0700)]
[PATCH] Dereference tag repeatedly until we get a non-tag.
When we allow a tag object in place of a commit object, we only
dereferenced the given tag once, which causes a tag that points at a tag
that points at a commit to be rejected. Instead, dereference tag
repeatedly until we get a non-tag.
This patch makes change to two functions:
- commit.c::lookup_commit_reference() is used by merge-base,
rev-tree and rev-parse to convert user supplied SHA1 to that of
a commit.
- rev-list uses its own get_commit_reference() to do the same.
Dereferencing tags this way helps both of these uses.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Junio C Hamano [Sun, 10 Jul 2005 22:40:43 +0000 (15:40 -0700)]
[PATCH] alternate object store and fsck
The location alt_odb[j].name[0..] is filled with ??/?{38} to form a sha1
filename to try, but I was too lazy to allocate a copy, so while
fsck_object_dir() is running for the directory, the filenames ??/?{38}
are filled after NUL (usually and always the location should have '/'),
making them "not found".
git-unpack-objects: show progress report by default
This ends up being very calming for big "git clone"s, since otherwise
you just get very frustrated with a long silence, wondering whether it's
working at all.
Use "-q" to quiet it down.
Now if we could just do the same for the initial "figure out what to
pack" phase, which can also be quite slow if the other end is busy (or
not packed and not in cache)...
Junio C Hamano [Sat, 9 Jul 2005 02:27:15 +0000 (19:27 -0700)]
[PATCH] format-patch: fix skipping of blank-lines
If it is fed a commit with more than one leading blank lines,
the sed scripts git-format-patch-script used looped forever.
Using git-stripspace upfront makes the sed script somewhat
simpler to work around this problem.
Also use git-rev-parse so that we can say
$ git-format-patch-script HEAD^^^^
to prepare the latest four patches for e-mail submission.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Instead of having to cut-and-paste the result, write it to the tag
directory directly. Also, start an editor for the tag message, rather
than just reading it from stdin.
Change git-update-cache --refresh behaviour to use sha1's rather than
comparing byte by byte.
[JC demangled whitespace from the posted patch himself because he
liked it so much. Also adjusted to the index_fd() interface
slightly done differently from the original one.]
Signed-off-by: Bryan Larsen <bryan.larsen@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
[PATCH] Add --info-only option to git-update-cache.
Add --info-only option to git-update-cache.
[JC demangled whitespace from the posted patch himself because he
liked it so much. Also adjusted to the index_fd() interface
slightly done differently from the original one.]
Signed-off-by: Bryan Larsen <bryan.larsen@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch makes the first half of write_sha1_file() and
index_fd() externally visible, to allow callers to compute the
object ID without actually storing it in the object database.
[JC demangled the whitespaces himself because he liked the patch
so much, and reworked the interface to index_fd() slightly,
taking suggestion from Linus and of his own.]
Signed-off-by: Bryan Larsen <bryan.larsen@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This silently adds the ".git" directory component if needed, so you
don't need to state it explicitly for the source. Also, it turns the
source into an absolute pathname when local, so that you can use
relative pathnames without losing sight of the source when we cd into
the destination.
Teach 'git-send-pack' to send new branches and tags.
The protocol always supported it, but send-pack didn't actually know how
to tell the other side about a new branch/tag.
NOTE! You'll have to name it explicitly on the command line: if you
don't name any branches, git-send-pack will default to the branches that
already exist.
Junio C Hamano [Fri, 8 Jul 2005 07:02:52 +0000 (00:02 -0700)]
[PATCH] Use sq_quote() to properly quote the parameter to call shell.
This tries to be more lenient to the users and stricter to the
attackers by quoting the input properly for shell safety,
instead of forbidding certain characters from the input.
Things to note:
- We do not quote "prog" parameter (which comes from --exec).
The user should know what he is doing. --exec='echo foo'
will supply the first two parameters to the resulting
command, while --exec="'echo foo'" will give the first
parameter, a single string with a space inside.
- We do not care too much about leaking the sq_quote() output
just before running exec().
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add "git-sh-setup-script" for common git shell script setup
It sets up the normal git environment variables and a few helper
functions (currently just "die()"), and returns ok if it all looks like
a git archive. So use it something like
. git-sh-setup-script || die "Not a git archive"
to make the rest of the git scripts more careful and readable.
I still worry about just quoting things when passing it off to "ssh" or
"sh -c", so I'm being anal. But _, ^ and , are certainly ok and while
both ~ and @ can have speacial meaning to shell/ssh they are benign.
Junio C Hamano [Thu, 7 Jul 2005 18:39:10 +0000 (11:39 -0700)]
[PATCH] Prevent t6000 series from dropping useless sed.script in t/
The Makefile in the test suite directory considers any file
matching t[0-9][0-9][0-9][0-9]-*.sh as the top-level test script
to be executed. Unfortunately this was not documented, and the
common test library, t6000-lib.sh was named to match that
pattern. This caused t6000-lib.sh to be called from Makefile as
the top-level program, causing it to leave t/sed.script file
behind. Rename it to t6000lib.sh to prevent this, and document
the naming convention a bit more clearly.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Junio C Hamano [Wed, 6 Jul 2005 20:04:21 +0000 (13:04 -0700)]
[PATCH] Short-circuit git-clone-pack while cloning locally (take 2).
When we are cloning a repository on a local filesystem, it is
faster to just create a hard linkfarm of .git/object hierarchy
and copy the .git/refs files. By default, the script uses the
clone-pack method, but it can be told with the -l flag to do the
hard linkfarm (falling back on recursive file copy) to replicate
the .git/object hierarchy.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Chris Wright [Thu, 7 Jul 2005 20:09:50 +0000 (13:09 -0700)]
Infrastructure for git rpm builds. Adds GIT_VERSION to Makefile and new make
targets: git.spec, dist, and rpm. A simple 'make rpm' will build the rpm.
Also adds git.spec.in which is used to generate git.spec.
Jon Seymour [Thu, 7 Jul 2005 00:59:13 +0000 (10:59 +1000)]
[PATCH] Ensure list insertion method does not depend on position of --merge-order argument
This change ensures that git-rev-list --merge-order produces the same result
irrespective of what position the --merge-order argument appears in the argument
list.
Signed-off-by: Jon Seymour <jon.seymour@gmail.com> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Jon Seymour [Thu, 7 Jul 2005 00:50:07 +0000 (10:50 +1000)]
[PATCH] Write sed script directly into temp file, rather than a variable
When sed uses \n rather than ; as a separator (for BSD sed(1) compat),
it is cleaner to use a file directly, rather than an environment
variable containing \n characters.
This change changes t/t6000 write to sed.script directly and changes
the other tests to remove knowledge of sed.script.
Signed-off-by: Jon Seymour <jon.seymour@gmail.com> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make gitk use --topo-order instead of --merge-order
It's cheaper to calculate, and doesn't give different results depending
on the order of the arguments passed in (and is thus more appropriate
for something like gitk that can validly take the unordered "--all" flag
to show all branches).
The previous dup fix seems to have fixed --topo-order. Holler if you
still see problems.
git-rev-list: remove the DUPCHECK logic, use SEEN instead
That's what we should have done in the first place, since it not only
avoids another unnecessary flag, it also protects the commits from
showing up as duplicates later when they show up as parents of another
commit (in the pop_most_recent_commit() path).
Junio C Hamano [Wed, 6 Jul 2005 08:11:52 +0000 (01:11 -0700)]
[PATCH] clone-pack.c:write_one_ref() - Create leading directories.
The function write_one_ref() is passed the list of refs received
from the other end, which was obtained by directory traversal
under $GIT_DIR/refs; this can contain paths other than what
git-init-db prepares and would fail to clone when there is
such.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Jon Seymour [Wed, 6 Jul 2005 16:39:34 +0000 (02:39 +1000)]
[PATCH] Add a topological sort procedure to commit.c
This introduces an in-place topological sort procedure to commit.c.
Given a list of commits, sort_in_topological_order() will perform an in-place
topological sort of that list.
The invariant that applies to the resulting list is:
a reachable from b => ord(b) < ord(a)
This invariant is weaker than the --merge-order invariant, but is cheaper
to calculate (assuming the list has been identified) and will serve any
purpose where only a minimal topological order guarantee is required.
Signed-off-by: Jon Seymour <jon.seymour@gmail.com> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Jon Seymour [Wed, 6 Jul 2005 10:11:29 +0000 (20:11 +1000)]
[PATCH] Change the sed seperator in t/t6000-lib.sh.
This trivial patch removes the semicolon as the sed seperator in the t/t6000-lib.sh test script
and replaces it with white space. This makes BSD sed(1) much happier.
Signed-off-by: Mark Allen <mrallen1@yahoo.com> Signed-off-by: Jon Seymour <jon.seymour@gmail.com> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
GIT_DIR=. ends up being what some of the pack senders use, and we
sometimes messed up when cleaning up the path, ie a ".//HEAD" was
cleaned up into "/HEAD", not "HEAD" like it should be.
We should do some other cleanup, and probably also verify that symlinks
don't point to outside the git area.
Increase the number of possible heads requested from git-upload-pack
Now that git-clone-pack exists, we actually have somebody requesting
more than just a single head in a pack. So allow the Jeff's of this
world to clone things with tens of heads.
Add a "git-show-index" helper that shows the contents of a pack index
This was invaluable for debugging the zero-sized compression issue, and
might be useful for scripting too, if people want to see the contents of
a pack.
zlib actually writes a header for that case, and while ignoring that
header will get us the right data, it will also end up messing up our
stream position. So we actually want zlib to "uncompress" even an empty
object.