git-push: avoid falling back on pushing "matching" refs.
The underlying "git send-pack remote.host:path" pushes all the
matching refs that both local and remote have, and "git push"
blindly inherits this property. Which probably was a mistake.
A typical cloned repository (e.g. a subsystem repository cloned
from Linus repository) has at least two branches, "master" to
keep the subsystem and "origin" that records tip of Linus
"master" when the repository was cloned. If this is the public
repository for the subsystem, then subsystem developers would
clone it, and then cloned ones have "master" and "origin". When
developers use this public subsystem repository as a shared
repository, pushing into it via "git push subsys:/path/name"
would try to push the matching refs, "master" and "origin", from
the developers' repositories. The "origin" in the public shared
repository does not have much relevance, yet pushing into
"origin" would cause "not a fast forward" checks to be
triggered. Arguably "git push subsys:/path/name master" would
work it around, but having them to say it explicitly to avoid
pushing into "origin" as well is bad.
This commit requires you to give at least one refspec to
git-push. You could "give" by either:
(1) Listing the refspec(s) explicitly on the command line.
E.g. "git push subsys:/path/name master".
(2) Using --all or --tags on the command line.
E.g. "git push --tags subsys:/path/name".
(3) Using a $GIT_DIR/remotes shorthand with 'Push: refspec'
line in it.
Unlike pull that can happen pretty much promiscuously, people
will push into the same set of a limited number of remote
repositories repeatedly over the life of the project, so it is
reasonable to assume they would want to keep a $GIT_DIR/remotes/
entry for those repositories even only to save typing the URL,
so keeping the default 'Push: refspec' line in such is a
sensible thing to do.
It was suggested to further fall back on pushing the current
branch, but this commit does not implement it. If developers
adopt topic branch workflow, pushing to public while on a topic
branch by mistake would expose the topic branch to the public
repository. Not falling back to the current branch prevents
that mistake from happening.
checkout: merge local modifications while switching branches.
* Instead of going interactive, introduce a command line switch
'-m' to allow merging changes when normal two-way merge by
read-tree prevents branch switching.
* Leave the unmerged stages intact if automerge fails, but
reset index entries of cleanly merged paths to that of the
new branch, so that "git diff" (not "git diff HEAD") would
show the local modifications.
* Swap the order of trees in read-tree three-way merge used in
the fallback, so that `git diff` to show the conflicts become
more natural.
* Describe the new option and give more examples in the documentation.
checkout: automerge local changes while switching branches.
When switching branches, if the working tree has a local
modification at paths that are different between current and new
branches, we refused the operation saying "cannot merge." This
attempts to do an automerge for such paths.
The earlier change to separate $(gitexecdir) from $(bindir) had
the installation location of the git wrapper and the rest of the
commands the wrong way (right now, both of them point at the
same location so there is no real harm).
The git suite may not be in PATH (and thus programs such as
git-send-pack could not exec git-rev-list). Thus there is a need for
logic that will locate these programs. Modifying PATH is not
desirable as it result in behavior differing from the user's
intentions, as we may end up prepending "/usr/bin" to PATH.
- git C programs will use exec*_git_cmd() APIs to exec sub-commands.
- exec*_git_cmd() will execute a git program by searching for it in
the following directories:
1. --exec-path (as used by "git")
2. The GIT_EXEC_PATH environment variable.
3. $(gitexecdir) as set in Makefile (default value $(bindir)).
- git wrapper will modify PATH as before to enable shell scripts to
invoke "git-foo" commands.
Ideally, shell scripts should use the git wrapper to become independent
of PATH, and then modifying PATH will not be necessary.
[jc: with minor updates after a brief review.]
Signed-off-by: Michal Ostrowski <mostrows@watson.ibm.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
octopus: allow criss-cross and clarify the message when it rejects
We rejected multi-base merge situations even though we used the
same underlying multi-base git-read-tree as the resolve strategy
uses. This was unneeded and did not add much to ensure the
merge to be truly trivial, so remove this restriction and be
more similar to what resolve does.
Also when the merge did not trivially resolve, we rejected
without stating that octopus strategy does not handle the
situation.
This is not invoked by any other target (most notably, "make
install" does not), but is provided as a convenience for people
who are building from the source.
name-rev: do not omit leading components of ref name.
In a repository with mainto/1.0 (to keep maintaining the 1.0.X
series) and fixo/1.0 (to keep fixes that apply to both 1.0.X
series and upwards) branches, "git-name-rev mainto/1.0" answered
just "1.0" making things ambiguous. Show refnames unambiguously
like show-branch does.
This is based on the patch by Andreas Ericsson, but done slightly
differently, preferring to have separate loops -- one for options
and then arguments.
show-branch: take default arguments from configuration file.
This lets showbranch.default multivalued configuration item to
be used as the default set of parameters to git-show-branch when
none is given on the command line.
I keep many topic branches (e.g. zzz/pack, net/misc) and
branches used only as a reference under subdirectories
(e.g. hold/{html,man,todo} track the same from git.git, but
clutters the show-branch output when shown along with the main
development; ko/master tracks what I have pushed out already and
refetched from the kernel.org server), and often run:
$ git show-branch ko/master heads/*
to view only the ko/master head and branches I keep immediately
under $GIT_DIR/refs/heads. With this change, I can have this in
my $GIT_DIR/config file:
Tommi Virtanen expressed a wish on #git to be able to use short and elegant
git URLs by making git-daemon 'root' in a given directory. This patch
implements this, causing git-daemon to interpret all paths relative to
the given base path if any is given.
Signed-off-by: Petr Baudis <pasky@suse.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
The main loop was prepared to take more than one revs, but the actual
naming logic wad not (it used pop_most_recent_commit while forgetting
that the commit marks stay after it's done).
ls-files --others --directory: give trailing slash
This adds a trailing slash to directory names in the output
when "--others --directory" option shows only untracked
directories and not their contents, to make them stand out.
ls-files --others --directory: fix a bug with index entry ordering
When both howto-index.sh and howto/make-dist.txt exist under
Documentation/ directory, dir_exists() mistakenly checked it
without the trailing slash to see if there was something under
Documentation/howto directory, and did not realize there was,
because '-' sorts earlier than '/' and cache_name_pos() finds
howto-index.sh, which is not under howto/ directory. This
caused --others --directory to show it which was incorrect.
Check the directory name with the trailing slash, because having
an entry that has such as a prefix is what we are looking for.
ls-files -o: optionally skip showing the contents in "untracked" directories
Darrin Thompson notes that git-ls-files -o reports all the unknown
files it finds in a work area. Subversion and probably other systems
"simply ignore all the files and directories inside an unknown
directory and just note the directory as unknown."
With --directory option, ls-files --others shows untracked directories
without descending into them.
I added things to ls-remote so that Cogito can auto-follow tags
easily and correctly a while ago, but git-fetch did not use the
facility. Recently added git-describe command relies on
repository keeping up-to-date set of tags, which made it much
more attractive to automatically follow tags, so we do that as
well.
prune: do not show error from pack-redundant when no packs are found.
When there is no pack yet, git-prune leaked an error message
from "git-pack-redundant --all" which complained that there is
no pack. Squelch the annoying message.
The official maintainer is keeping up-to-date quite well, and now
the older Debian is supported with backports.org, there is no reason
for me to keep debian/ directory around here.
I have not been building and publishing debs since 1.0.4 anyway.
"cvs add" support was already there, but the "unknown" status
returned when querying a file not yet known to cvs caused the
script to abort prematurely.
Make GIT-VERSION-GEN tolerate missing git describe command
I think it is probably a bug that "git non_existent_command"
returns its error message to stdout without an error, where
"git-non_existent_command" behaves differently and does return an
error.
Older versions of git did not implement "git describe" and
GIT-VERSION-GEN produces an empty version string if run on
a system with such a git installed. The consequence
is that "make rpm" fails.
This patch fixes GIT-VERSION-GEN so that it works in the
absence of a working "git describe"
Signed-off-by: John Ellson <ellson@research.att.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
When the other end was prepared with older git and has tags that
do not follow the naming convention (see check-ref-format), do not
barf but simply reject to copy them.
Initial fix by Simon Richter, but done differently.
Not that the stat against open race would matter much in this context,
but that simplifies
the code a bit. Also some diagnostics added (why the open failed)
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
If approxidate ends up with a month that is ahead of the current month, it
decrements the year to last year.
Which is correct, and means that "last december" does the right thing.
HOWEVER. It should only do so if the year is the same as the current year.
Without this fix, "5 days ago" ends up being in 2004, because it first
decrements five days, getting us to December 2005 (correct), but then it
also ends up decrementing the year once more to turn that December into
"last year" (incorrect, since it already _was_ last year).
Duh. Pass me a donut.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Move git-rev-list --merge-order usage check for 'OpenSSL not linked' after
test 1; we cannot trigger this unless we try to actually use --merge-order
by giving some ref, and we do not have any ref until we run the first test
to create commits.
Signed-off-by: Gerrit Pape <pape@smarden.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
send-pack/receive-pack: allow errors to be reported back to pusher.
This updates the protocol between git-send-pack/git-receive-pack
in a backward compatible way to allow failures at the receiving
end to be propagated back to the sender. Most notably, versions
of git-push before this could not notice if the update hook on
the receiving end refused to update the ref for its own policy
reasons.
Makefile: use git-describe to mark the git version.
Note: with this commit, the GIT maintainer workflow must change.
GIT-VERSION-GEN is now the file to munge when the default
version needs to be changed, not Makefile. The tag needs to be
pushed into the repository to build the official tarball and
binary package beforehand.
git-describe: still prefer annotated tag under --all and --tags
Even though --all and --tags can be used to include non
annotated tags in the reference point candidates, prefer to use
annotated tags if there are more than one refs that name the
same commit.
Often there are references other than annotated tags under
refs/tags hierarchy that are used to "keep things just in case".
default to use annotated tags only, still leaving the option to
use any ref with --all flag.
It shows you the most recent tag that is reachable from a particular
commit is.
Maybe this is something that "git-name-rev" should be taught to do,
instead of having a separate command for it. Regardless, I find it useful.
What it does is to take any random commit, and "name" it by looking up the
most recent commit that is tagged and reachable from that commit. If the
match is exact, it will just print out that ref-name directly. Otherwise
it will print out the ref-name, followed by the 8-character "short SHA".
IOW, with something like Junios current tree, I get:
ie the current head of my "parent" branch (ie Junio) is based on v1.0.4,
but since it has a few commits on top of that, it has added the git hash
of the thing to the end: "-g" + 8-char shorthand for the commit 2414721b194453f058079d897d13c4e377f92dc6.
Doing a "git-describe" on a tag-name will just show the full tag path:
unless there are _other_ tags pointing to that commit, in which case it
will just choose one at random.
This is useful for two things:
- automatic version naming in Makefiles, for example. We could use it in
git itself: when doing "git --version", we could use this to give a
much more useful description of exactly what version was installed.
- for any random commit (say, you use "gitk <pathname>" or
"git-whatchanged" to look at what has changed in some file), you can
figure out what the last version of the repo was. Ie, say I find a bug
in commit 39ca371c45b04cd50d0974030ae051906fc516b6, I just do:
Starting from this one I won't be touching debian/ directory
since the official maintainer seems to be reasonably quick to
package up things. The packaging procedure used there seems to
be quite different from what I have, so I'd like to avoid
potential confusion and reduce work by the official maintainer
and myself.
This patch converts a stat() to an lstat() call, thereby fixing the case
when the date of a symlink was not the same as the one recorded in the
index. The included test case demonstrates this.
This is for the case that the symlink points to a non-existing file. If
the file exists, worse things than just an error message happen.
Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <junkio@cox.net>
Johannes found that the test has 1/256 chance of falsely
producing an uncorrupted idx file, causing the check to detect
corruption fail. Now we have 1/2^160 chance of false failure
;-).
Avoid asking for zero bytes when that change simplifies overall
logic. Later we would change the wrapper to ask for 1 byte on
platforms that return NULL for zero byte request.