This allows git-send-pack to push local refs to a destination
repository under different names.
Here is the name mapping rules for refs.
* If there is no ref mapping on the command line:
- if '--all' is specified, it is equivalent to specifying
<local> ":" <local> for all the existing local refs on the
command line
- otherwise, it is equivalent to specifying <ref> ":" <ref> for
all the refs that exist on both sides.
* <name> is just a shorthand for <name> ":" <name>
* <src> ":" <dst>
push ref that matches <src> to ref that matches <dst>.
- It is an error if <src> does not match exactly one of local
refs.
- It is an error if <dst> matches more than one remote refs.
- If <dst> does not match any remote refs, either
- it has to start with "refs/"; <dst> is used as the
destination literally in this case.
- <src> == <dst> and the ref that matched the <src> must not
exist in the set of remote refs; the ref matched <src>
locally is used as the name of the destination.
For example,
- "git-send-pack --all <remote>" works exactly as before;
- "git-send-pack <remote> master:upstream" pushes local master
to remote ref that matches "upstream". If there is no such
ref, it is an error.
- "git-send-pack <remote> master:refs/heads/upstream" pushes
local master to remote refs/heads/upstream, even when
refs/heads/upstream does not exist.
- "git-send-pack <remote> master" into an empty remote
repository pushes the local ref/heads/master to the remote
ref/heads/master.
A template mechanism to populate newly initialized repository
with default set of files is introduced. Use it to ship example
hooks that can be used for update and post update checks, as
Josef Weidendorfer suggests.
When pushing into multi-user repository, or when pushing to a
repository from a local repository that has rebased branches
that has been pruned, the destination repository can have head
commits that are missing from the local repository.
This should not matter as long as the local head of the branch
being pushed is a proper superset of the destination branch, but
we ended up trying to run rev-list telling it to exclude objects
reachable from those heads missing from the local repository,
causing it to barf. Prune those heads from the rev-list
parameter list, and make sure we do not try to push a branch
whose remote head is something we lack.
[PATCH] Add git-send-email-script - tool to send emails from git-format-patch-script
This is based off of GregKH's script, send-lots-of-email.pl, and strives to do
all the nice things a good subsystem maintainer does when forwarding a patch or
50 upstream:
All the prior handlers of the patch, as determined by the Signed-off-by: lines, and/or the author of the commit, are cc:ed on the
email.
All emails are sent as a reply to the previous email, making it easy to
skip a collection of emails that are uninteresting.
Signed-off-by: Ryan Anderson <ryan@michonline.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
This causes ssh-pull to request objects in prefetch() and read then in
fetch(), such that it reduces the unpipelined round-trip time.
This also makes sha1_write_from_fd() support having a buffer of data
which it accidentally read from the fd after the object; this was
formerly not a problem, because it would always get a short read at
the end of an object, because the next object had not been
requested. This is no longer true.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
This processes objects in two simultaneous passes. Each object will
first be given to prefetch(), as soon as it is possible to tell that
it will be needed, and then will be given to fetch(), when it is the
next object that needs to be parsed. Unless an implementation does
something with prefetch(), this should have no effect.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Make sure leading directories exist when pushing refs.
It does not matter if the only refs you push are directly
underneath heads and tags, but we forgot to make sure we have
leading directories so pushing tags/v0.99/1 would not have
worked.
The earlier one conflated update and post-update hooks for no
good reason. Correct that ugly hack. Now post-update hooks
will take the list of successfully updated refs.
Make send-pack --all and explicit ref mutually exclusive.
send-pack had a confusing misfeature that "send-pack --all
master" updated all refs, while "send-pack --all" did not do
anything. Make --all and explicit refs mutually exclusive, and
make sure "send-pack --all" updates all refs.
Things have slowly but surely started to settle down, and the
http transport finally can natively grok packed repositories.
To give Pasky a good anchor point, hoping that he can start
split off the core part from Cogito, here is the 0.99.3, which
will be accompanied with its own tag.
[PATCH] git-merge-cache -q doesn't complain about failing merge program
git-merge-cache reporting failed merge program is undesirable for
Cogito, since it emits its own more appropriate error message in that
case. However, I want to show other possible git-merge-cache error
messages. So -q will just silence this particular error.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] Support downloading packs by HTTP (whitespace fixed)
This adds support to http-pull for finding the list of pack files
available on the server, downloading the index files for those pack
files, and downloading pack files when they contain needed objects not
available individually. It retains the index files even if the pack
files were not needed, but downloads the list of pack files once per
run if an object is not found separately.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] Functions for managing the set of packs the library is using (whitespace fixed)
This adds support for reading an uninstalled index, and installing a
pack file that was added while the program was running, as well as
functions for determining where to put the file.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
is called if executable. The hook can decline the ref to be
updated by exiting with a non-zero status, or allow it to be
updated by exiting with a zero status. The mechanism also
allows e.g sending of a mail with pushed commits on the remote
repository.
Documentation update with an example hook is included.
jc: The credits of the basic idea and initial implementation go
to Josef, but I ended up rewriting major parts of his patch, so
bugs are all mine. Also I changed the semantics for the hook
from his original version (which were post-update hook) so that
the hook can optionally decline to update the ref, and also can
be used to implement the overall cleanups. The latter was
primarily to implement a suggestion from Linus that calling
update-server-info should be made optional.
Introduce a new file $GIT_DIR/info/grafts (or $GIT_GRAFT_FILE)
which is a list of "fake commit parent records". Each line of
this file is a commit ID, followed by parent commit IDs, all
40-byte hex SHA1 separated by a single SP in between. The
records override the parent information we would normally read
from the commit objects, allowing both adding "fake" parents
(i.e. grafting), and pretending as if a commit is not a child of
some of its real parents (i.e. cauterizing).
Implement fetching from a packed repository over http/https
using the dumb server support files.
I consider some parts of the logic should be in a separate C
program, but it appears to work with my simple tests. I have
backburnered it for a bit too long for my liking, so let's throw
it out in the open and see what happens.
Specifically this should fix the following errors:
wrong # args: should be "startdiff ids" (fix from Junio Hamano)
can't read "filelines(....)": no such element in array
can't unset "treepending": no such variable
On Sat, 30 Jul 2005, Linus Torvalds wrote:
>
> Yup, it's git-merge-base, and it is confused by the same thing that
> confused git-rev-list.
Hmm.. Here's a tentative fix. I'm not really happy with it, and maybe
somebody else can come up with a better one. I think this one ends up
being quite a bit more expensive than the old one (it will look up _all_
common parents that have a child that isn't common, and then select the
newest one of the bunch), but I haven't really thought it through very
much.
[PATCH] Making it easier to find which change introduced a bug
This adds a new "git bisect" command.
- "git bisect start"
start bisection search.
- "git bisect bad <rev>"
mark some version known-bad (if no arguments, then current HEAD)
- "git bisect good <revs>..."
mark some versions known-good (if no arguments, then current HEAD)
- "git bisect reset <branch>"
done with bisection search and go back to your work (if
no arguments, then "master").
The way you use it is:
git bisect start
git bisect bad # Current version is bad
git bisect good v2.6.13-rc2 # v2.6.13-rc2 was the last version
# tested that was good
When you give at least one bad and one good versions, it will
bisect the revision tree and say something like:
Bisecting: 675 revisions left to test after this
and check out the state in the middle. Now, compile that kernel, and boot
it. Now, let's say that this booted kernel works fine, then just do
git bisect good # this one is good
which will now say
Bisecting: 337 revisions left to test after this
and you continue along, compiling that one, testing it, and depending on
whether it is good or bad, you say "git bisect good" or "git bisect bad",
and ask for the next bisection.
Until you have no more left, and you'll have been left with the first bad
kernel rev in "refs/bisect/bad".
Oh, and then after you want to reset to the original head, do a
git bisect reset
to get back to the master branch, instead of being in one of the bisection
branches ("git bisect start" will do that for you too, actually: it will
reset the bisection state, and before it does that it checks that you're
not using some old bisection branch).
Not really any harder than doing series of "quilt push" and "quilt pop",
now is it?
[jc: This patch is a rework based on what Linus posted to the
list. The changes are:
- The original introduced four separate commands, which was
three too many, so I merged them into one with subcommands.
- Since the next thing you would want to do after telling it
"bad" and "good" is always to bisect, this version does it
automatically for you.
- I think the termination condition was wrong. The original
version checked if the set of revisions reachable from next
bisection but not rechable from any of the known good ones
is empty, but if the current bisection was a bad one, this
would not terminate, so I changed it to terminate it when
the set becomes a singleton or empty.
- Removed the use of shell array variable.
]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Separate the process of building the commands to compilation and
linkage. This makes it more consistent with the library objects, is the
traditional thing to do, and significantly speeds up the subsequent
rebuilds, especially for us the people who develop git on 300MHz
notebooks.
Ported from Cogito.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
Skip --merge-order test when built with NO_OPENSSL
When built with NO_OPENSSL, rev-list --merge-order does not
work, causing t6001 test to fail. Detect that and skip this
test to allow continuing to the rest of the tests.
Support for completely OpenSSL-less builds. FSF considers distributing GPL
binaries with OpenSSL linked in as a legal problem so this is trouble
e.g. for Debian, or some people might not want to install OpenSSL
anyway. If you
make NO_OPENSSL=1
you get completely OpenSSL-less build, disabling --merge-order and using
Mozilla's SHA1 implementation.
Ported from Cogito.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
The Makefile rules were massively reordered so that they are actually
logically grouped now. Captions were added to separate the sections. No
rule contents was touched during the process.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] Remove the explicit Makefile dependencies description
Remove about one gazillion of explicit dependency rules with few lines
describing the general dependency pattern and then the exceptions. This
noticably shortens the Makefile and makes it easier to touch it.
This is part of the Cogito Makefile changes port.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] Improve the compilation-time settings interface
Describe variables which make itself takes and adjusts compilation
accordingly (MOZILLA_SHA1, NO_OPENSSL, PPC_SHA1), and make adding
defines more convenient through the $DEFINES variable. $COPTS includes
-g as well now and is not overriden if it was already declared in the
environment. Also, $CFLAGS is appended to rather than reset, so that if
there was already a $CFLAGS environment variable, it's appended to. Some
more variables are also made overridable through the environment. Renamed
$bin to $bindir which is the name commonly used for this.
This is part of the Cogito Makefile changes port.
Signed-off-by: Petr Baudis <pasky@ucw.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
I have reviewed all occurrences of mmap() in git and fixed three types
of errors/defects:
1) The result is not checked.
2) The file descriptor is closed if mmap() succeeds, but not when it
fails.
3) Various casts applied to -1 are used instead of MAP_FAILED, which is
specifically defined to check mmap() return value.
[jc: This is a second round of Pavel's patch. He fixed up the problem
that close() potentially clobbering the errno from mmap, which
the first round had.]
Signed-off-by: Pavel Roskin <proski@gnu.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Pasky and others raised many valid points on the problems
initial exclude pattern enhancement work had. Based on the
list discussion, rework the exclude logic to use "last match
determines its fate" rule, and order the list by exclude-from
(the fallback default pattern file), exclude-per-directory
(shallower to deeper, so deeper ones can override), and then
command line exclude patterns.
This corner-case was triggered by a kernel commit that was not in date
order, due to a misconfigured time zone that made the commit appear three
hours older than it was.
That caused git-rev-list to traverse the commit tree in a non-obvious
order, and made it parse several of the _parents_ of the misplaced commit
before it actually parsed the commit itself. That's fine, but it meant
that the grandparents of the commit didn't get marked uninteresting,
because they had been reached through an "interesting" branch.
The reason was that "mark_parents_uninteresting()" (which is supposed to
mark all existing parents as being uninteresting - duh) didn't actually
traverse more than one level down the parent chain.
NORMALLY this is fine, since with the date-based traversal order,
grandparents won't ever even have been looked at before their parents (so
traversing the chain down isn't needed, because the next time around when
we pick out the parent we'll mark _its_ parents uninteresting), but since
we'd gotten out of order, we'd already seen the parent and thus never got
around to mark the grandparents.
Anyway, the fix is simple. Just traverse parent chains recursively.
Normally the chain won't even exist (since the parent hasn't been parsed
yet), so this is not actually going to trigger except in this strange
corner-case.
Add a comment to the simple one-liner, since this was a bit subtle, and I
had to really think things through to understand how it could happen.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Darrin Thompson noticed when he was showing off GIT to others
that the use of filenames "a" and "b" in the tutorial example
was unnecessarily confusing, especially with our "patch -p1"
prefix a/ and b/, without giving us any patch. I was very
tempted to change them back to l/ and k/ prefixes, but decided
to restrain myself and update the tutorial instead ;-).
Improve the merge display when the result differs from all parents.
Now we see if the result is quite similar to one of the parents, and
if it is, display the result as a diff from that parent. If the result
is similar to more than one parent, pick the one that it's most
similar to.
[PATCH] socklen_t needs to be defined and libssl to be linked on old Mac OS X
On older Mac OS X (10.2.8), no socklen_t is defined, and therefore
daemon.c does not compile. However, Mac OS X 10.4 seems to define
socklen_t differently.
Also, linking fails due to some symbols defined in libssl (not just
libcrypto).
[jc: I am tentatively dropping the socklen_t part of the patch
because I am waiting for confirmation on the server side IPV6
patch from Yoshifuji-san]
Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] Make git-apply --stat less butt-ugly with long filenames
When git-apply was printing out long filenames, it used to just truncate
them to show the last "max_len" characters of the filename. Which can be
really quite ugly (note the two filenames that have just been silently
truncated from the beginning - it looks even worse when there are lots
of them, like there were in the current v2.6.13-rc4 cris arch update):
Some places assumed .git is the GIT_DIR, resulting heads and
tags not showing when it was run like "GIT_DIR=. gitk --all".
This is not a contrived example --- I rely on it to verify
my private copy of git.git repository before pushing it out.
Define a single procedure "gitdir" and use it.
Signed-off-by: Junio C Hamano <junkio@cox.net> Signed-off-by: Paul Mackerras <paulus@samba.org>
Display the diffs for a merge in a unified fashion.
Stuff that ended up in the result is shown in bold with a "+" at the
beginning of the line; stuff that didn't is in the normal font with
a "-" at the beginning of the line. The color shows which parent
the stuff was in; red for the first parent, blue for the second, then
green, purple, brown, and the rest are grey. If the result is different
from all of the parents it is shown in black (and bold).
In particular, warn about things like zero-padding of the mode bits,
which is a big no-no, since it makes otherwise identical trees have
different representations (and thus different SHA1 numbers).
The old mode conversion was not only complex, it also refused to change
the length of a mode, which made it fragile. By moving the mode
conversion around a bit, we can not only simplify it, it also ends up
being more powerful.
Also fix a memory leak that made it impossible to convert huge archives
without tons and tons of memory.
git-fsck-cache.c: check commit objects more carefully
We historically used to be very careful in fsck-cache, but when it was
re-written to use "parse_object()" instead of parsing everything by
hand, it lost a bit of the checks. This, together with the previous
commit, should make it do more proper commit object syntax checks.
Also add a "--strict" flag, which warns about the old-style "0664" file
mode bits, which shouldn't exist in modern trees, but that happened
early on in git trees and that the default git-fsck-cache thus silently
accepts.
- use --verify to make sure rev param is a rev, and barf otherwise.
- make it always output to stdout; no funny business with tee.
- take optional branch head name to specify which branch to summarize.
- show baserev in a human readable way.
- do not depend on diffstat; use git-apply --stat instead.
Deb Packaging fixes: Build against Mozilla libs for Debian, conflict with "git"
This patch includes two fixes to the git-core Debian package:
* Conflict with the GNU Interactive Tools package, which _also_
wants to install /usr/bin/git.
* Compile against the unencumbered Mozilla SHA1 code, instead of
the iffy OpenSSL code, as much as possible. This makes it easier to get
the package included for distribution with Debian.
This has been based upon the original patch by Sebastian Kuzminsky
<seb@highlab.com>, but has been fixed up based upon feedback.
Signed-off-by: Ryan Anderson <ryan@michonline.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
Adjust diff-raw tests to the status letter change.
We use 'A' for added files instead of 'N' to make the it
visually easier to distinguish from 'M' now.
While we are at it, make the test scripts executable. Yes, I
know it does not matter because t/Makefile runs them explicitly
with "sh tXXXX-blah.sh", but being consistent is always better.
Use symbolic constants for diff-raw status indicators.
Both Cogito and StGIT prefer to see 'A' for new files. The
current 'N' is visually harder to distinguish from 'M', which is
used for modified files. Prepare the internals to use symbolic
constants to make the change easier.
Add --exclude-per-directory=<name> option that specifies a file
to contain exclude patterns local to that directory and its
subdirectories. Update the exclusion logic to be able to say
"include files that match this more specific pattern, even
though later exclude patterns may match them". Also enhances
that a pattern can contain '/' in which case fnmatch is called
with FNM_PATHNAME flag to match the entire path.
This adds -a (annotate only but not sign) option "A Large Angry
SCM" <gitzilla@gmail.com> sent to the list, after fixing up the
whitespace corruption in the patch, with some of my own fixes.
Namely:
* A new flag '-a' can be used to create an unsigned tag
object;
* The '-f' flag logic did not do the right thing;
* When creating a signed tag, we did not check for GPG failure
as we should;
* Try to use the key for the tagger identity when signing the
tag.