git/git.git
15 years agoUndef DT_* before redefining them.
Junio C Hamano [Fri, 20 Jan 2006 21:33:20 +0000 (13:33 -0800)]
Undef DT_* before redefining them.

When overriding DT_* macro detection with NO_D_TYPE_IN_DIRENT (recent
Cygwin build problem, which hopefully is already fixed in their CVS
snapshot version), we define DTYPE() macro to return just "we do not
know", but still needed to use DT_* macro to avoid ifdef in the code
we use them.  If the platform defines DT_* macro but with unusable
d_type, this would have resulted in us redefining these preprocessor
symbols.

Admittedly, that would be just a couple of compilation warnings, and
on Cygwin at least this particular problem is transitory (the problem
is already fixed in their CVS snapshot version), so this is a low
priority fix.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agot4011: "sleep 1" is not enough on FAT
Junio C Hamano [Fri, 20 Jan 2006 21:17:59 +0000 (13:17 -0800)]
t4011: "sleep 1" is not enough on FAT

This test depended on "sleep 1" to be enough to dirty the index
entry for a symlink.  Alex noticed that on his Cygwin installation
"sleep 1" was sometimes not enough, and after further discussion with
Christopher Faylor, it was brought up that on FAT filesystem timestamp
granularity is 2 seconds so sleeping 1 second is not enough.

For now this patch takes an easy workaround of sleeping for 3 seconds.

Very strictly speaking, POSIX requires lstat to fill only S_IFMT part
of st_mode and st_size for symlinks, and depending on timestamp might
be considered a bug, but we depend on that anyway, so it is better to
test that.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDT_UNKNOWN: do not fully trust existence of DT_UNKNOWN
Junio C Hamano [Fri, 20 Jan 2006 01:13:57 +0000 (17:13 -0800)]
DT_UNKNOWN: do not fully trust existence of DT_UNKNOWN

The recent Cygwin defines DT_UNKNOWN although it does not have d_type
in struct dirent.  Give an option to tell us not to use d_type on such
platforms.  Hopefully this problem will be transient.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agofsck-objects: support platforms without d_ino in struct dirent.
Junio C Hamano [Fri, 20 Jan 2006 01:13:51 +0000 (17:13 -0800)]
fsck-objects: support platforms without d_ino in struct dirent.

The d_ino field is only used for performance reasons in
fsck-objects.  On a typical filesystem, i-number tends to have a
strong correlation with where the actual bits sit on the disk
platter, and we sort the entries to allow us scan things that
ought to be close together together.

If the platform lacks support for it, it is not a big deal.
Just do not use d_ino for sorting, and scan them unsorted.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMakefile: do not assume lack of IPV6 means no sockaddr_storage.
Junio C Hamano [Fri, 20 Jan 2006 01:13:32 +0000 (17:13 -0800)]
Makefile: do not assume lack of IPV6 means no sockaddr_storage.

Noticed first by Alex, that the latest Cygwin now properly has
sockaddr_storage.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMention install-doc in INSTALL
J. Bruce Fields [Sat, 21 Jan 2006 23:54:12 +0000 (18:54 -0500)]
Mention install-doc in INSTALL

I think most people will want to install the man pages as well.

[jc: incorporated Pasky's comment on not building them as root.
Some people may not want to install asciidoc/xmlto toolchain, so
redirect them to the man and html branches of the git.git
repository as well.]

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocument git-ls-files --directory
Petr Baudis [Sun, 22 Jan 2006 00:22:30 +0000 (01:22 +0100)]
Document git-ls-files --directory

Add the appropriate bit of documentation.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMinor git-reset and git-commit documentation fixes
J. Bruce Fields [Sat, 21 Jan 2006 23:54:11 +0000 (18:54 -0500)]
Minor git-reset and git-commit documentation fixes

Minor copyediting of recent additions to git-commit and git-reset
documentation.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRevert "git-push: avoid falling back on pushing "matching" refs."
Junio C Hamano [Sat, 21 Jan 2006 20:36:12 +0000 (12:36 -0800)]
Revert "git-push: avoid falling back on pushing "matching" refs."

This reverts 9e9b26751a5ca7a257b3e1cfb319fe3e4efc663c commit partially.
When no refspec is specified on the command line and there is no
default refspec to push specified in remotes/ file, just let
send-pack to do its default "matching refs" updates.

Thanks to Greg KH for complaining.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAsciiDoc fixes for the git-svnimport manpage
Florian Weimer [Sat, 21 Jan 2006 19:06:19 +0000 (20:06 +0100)]
AsciiDoc fixes for the git-svnimport manpage

Change "SVN:: Perl" to "SVN::Perl", wrap a long line, and clean up the
description of positional arguments.

Signed-off-by: Florian Weimer <fw@deneb.enyo.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoupdate using-topic-branches
Luck, Tony [Fri, 20 Jan 2006 19:00:18 +0000 (11:00 -0800)]
update using-topic-branches

Update documentation to warn users not to create noise in then Linux
history by creating pointless "Auto-update from upstream" merge
commits.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix generation of "humanish" part of source repo
Uwe Zeisberger [Fri, 20 Jan 2006 06:47:39 +0000 (07:47 +0100)]
Fix generation of "humanish" part of source repo

If repo has the form <host>:<path> and <path> doesn't contain a slash, the
cloned repository is named "<host>:<path>", instead of "<path>" only.

Signed-off-by: Uwe Zeisberger <zeisberg@informatik.uni-freiburg.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDisable USE_SYMLINK_HEAD by default
Pavel Roskin [Tue, 15 Nov 2005 05:59:50 +0000 (00:59 -0500)]
Disable USE_SYMLINK_HEAD by default

Disable USE_SYMLINK_HEAD by default.  Recommend using it only for
compatibility with older software.

Treat USE_SYMLINK_HEAD like other optional defines - check whether it's
defined, not its value.

Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge fixes up to GIT 1.1.4
Junio C Hamano [Fri, 20 Jan 2006 03:57:56 +0000 (19:57 -0800)]
Merge fixes up to GIT 1.1.4

15 years agoGIT 1.1.4 v1.1.4
Junio C Hamano [Fri, 20 Jan 2006 03:56:27 +0000 (19:56 -0800)]
GIT 1.1.4

15 years agoGIT 1.0.12 v1.0.12
Junio C Hamano [Fri, 20 Jan 2006 02:32:54 +0000 (18:32 -0800)]
GIT 1.0.12

15 years agogit-fetch-pack: really do not ask for funny refs
Johannes Schindelin [Wed, 18 Jan 2006 23:24:04 +0000 (00:24 +0100)]
git-fetch-pack: really do not ask for funny refs

If git-fetch-pack was called with out any refspec, it would ask the server
for funny refs. That cannot work, since the funny refs are not marked
as OUR_REF by upload-pack, which just exits with an error.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRevert "check_packed_git_idx(): check integrity of the idx file itself."
Junio C Hamano [Thu, 19 Jan 2006 04:26:14 +0000 (20:26 -0800)]
Revert "check_packed_git_idx(): check integrity of the idx file itself."

This reverts c5ced64578a82b9d172aceb2f67c6fb9e639f6d9 commit.
It turns out that doing this check every time we map the idx file
is quite expensive.  A corrupt idx file is caught by git-fsck-objects,
so this check is not strictly necessary.

In one unscientific test, 0.99.9m spent 10 seconds usertime for
the same task 1.1.3 takes 37 seconds usertime.  Reverting this gives
us the performance of 0.99.9 back.

15 years agoAllow building of RPM from interim snapshot.
Junio C Hamano [Mon, 16 Jan 2006 08:58:45 +0000 (00:58 -0800)]
Allow building of RPM from interim snapshot.

By popular demand.  If you build and install such binary RPMs,
the version numbering will lose monotonicity, so you may have to
later override downgrade warnings from your packaging manager,
but as long as you are aware of that and know how to deal with it,
there is no reason for us to forbid it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-push: fix --tags and document it.
Junio C Hamano [Mon, 16 Jan 2006 07:27:34 +0000 (23:27 -0800)]
git-push: fix --tags and document it.

Previously 'git-push --tags dst', used information from
remotes/dst to determine which refs to push; this patch corrects
it, and also documents the --tags option.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge fixes up to GIT 1.1.3
Junio C Hamano [Mon, 16 Jan 2006 06:25:35 +0000 (22:25 -0800)]
Merge fixes up to GIT 1.1.3

15 years agoGIT 1.1.3 v1.1.3
Junio C Hamano [Mon, 16 Jan 2006 06:22:23 +0000 (22:22 -0800)]
GIT 1.1.3

15 years agodescribe: omit clearing marks on the last one.
Junio C Hamano [Mon, 16 Jan 2006 06:15:37 +0000 (22:15 -0800)]
describe: omit clearing marks on the last one.

When describing more than one, we need to clear the commit marks
before handling the next one, but most of the time we are
running it for only one commit, and in such a case this clearing
phase is totally unnecessary.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fix'
Junio C Hamano [Mon, 16 Jan 2006 05:34:31 +0000 (21:34 -0800)]
Merge branch 'fix'

15 years agocvsimport: ease migration from CVSROOT/users format
Junio C Hamano [Sun, 15 Jan 2006 11:30:30 +0000 (03:30 -0800)]
cvsimport: ease migration from CVSROOT/users format

This fixes a minor bug, which caused the author email to be
doubly enclosed in a <> pair (the code gave enclosing <> to
GIT_AUTHOR_EMAIL and GIT_COMMITTER_EMAIL environment variable).

The read_author_info() subroutine is taught to also understand
the user list in CVSROOT/users format.  This is primarily done
to ease migration for CVS users, who can use the -A option
to read from existing CVSROOT/users file.  write_author_info()
always writes in the git-cvsimport's native format ('='
delimited and value without quotes).

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-cvsimport: Add -A <author-conv-file> option
Andreas Ericsson [Thu, 12 Jan 2006 23:38:59 +0000 (00:38 +0100)]
git-cvsimport: Add -A <author-conv-file> option

This patch adds the option to specify an author name/email conversion
file in the format

exon=Andreas Ericsson <ae@op5.se>
spawn=Simon Pawn <spawn@frog-pond.org>

which will translate the ugly cvs authornames to the more informative
git style.

The info is saved in $GIT_DIR/cvs-authors, so that subsequent
incremental imports will use the same author-info even if no -A
option is specified. If an -A option *is* specified, the info in
$GIT_DIR/cvs-authors is appended/updated appropriately.

Docs updated accordingly.

Signed-off-by: Andreas Ericsson <ae@op5.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fixo/1.0'
Junio C Hamano [Mon, 16 Jan 2006 05:12:12 +0000 (21:12 -0800)]
Merge branch 'fixo/1.0'

15 years agoGIT 1.0.11 v1.0.11
Junio C Hamano [Mon, 16 Jan 2006 05:11:30 +0000 (21:11 -0800)]
GIT 1.0.11

15 years agodiffcore-break/diffcore-rename: integer overflow.
Junio C Hamano [Mon, 16 Jan 2006 05:08:42 +0000 (21:08 -0800)]
diffcore-break/diffcore-rename: integer overflow.

While reviewing the end user tutorial rewrite by J. Bruce
Fields, I noticed that "git-diff-tree -B -C" did not correctly
break the total rewrite of Documentation/tutorial.txt.  It turns
out that we had integer overflow during the break score
computations.

Cop out by using floating point.  This is not a kernel.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branches 'jc/checkout', 'jc/format-patch', 'jc/octopus', 'jc/sb' and 'jc/clone'
Junio C Hamano [Sun, 15 Jan 2006 09:19:09 +0000 (01:19 -0800)]
Merge branches 'jc/checkout', 'jc/format-patch', 'jc/octopus', 'jc/sb' and 'jc/clone'

15 years agoshow-branch: --current includes the current branch.
Junio C Hamano [Wed, 11 Jan 2006 08:16:42 +0000 (00:16 -0800)]
show-branch: --current includes the current branch.

With this, the command includes the current branch to the list
of revs to be shown when it is not given on the command line.
This is handy to use in the configuration file like this:

[showbranch]
default = --current
default = heads/* ; primary branches, not topics under
; subdirectories

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoshow-branch: make the current branch and merge commits stand out.
Junio C Hamano [Wed, 11 Jan 2006 22:02:38 +0000 (14:02 -0800)]
show-branch: make the current branch and merge commits stand out.

This changes the character used to mark the commits that is on the
branch from '+' to '*' for the current branch, to make it stand out.
Also we show '-' for merge commits.

When you have a handful branches with relatively long diversion, it
is easier to see which one is the current branch this way.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocumentation: show-branch.
Junio C Hamano [Wed, 11 Jan 2006 08:09:16 +0000 (00:09 -0800)]
Documentation: show-branch.

Describe showbranch.default configuration item and give an example.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agooctopus: allow manual resolve on the last round.
Junio C Hamano [Sat, 14 Jan 2006 00:45:42 +0000 (16:45 -0800)]
octopus: allow manual resolve on the last round.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] format-patch: always --mbox and show sane Date:
Junio C Hamano [Fri, 13 Jan 2006 23:56:44 +0000 (15:56 -0800)]
[PATCH] format-patch: always --mbox and show sane Date:

Make --mbox, --author, and --date options a no-op, and always
use --mbox output, and rewrite the commit log formatting in
Perl.  This makes it easier to output Date: header in RFC 2822
format, so do that as well.

Inspiration for this patch came from Andreas Ericsson's earlier
patch.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoclone: --naked option.
Junio C Hamano [Sun, 15 Jan 2006 00:00:32 +0000 (16:00 -0800)]
clone: --naked option.

The new option --naked is to help creating a naked repository
for public consumption.

$ git clone -l -s --naked \
  /pub/scm/.../torvalds/linux-2.6.git subproj-2.6.git

is equivalent to this sequence:

$ git clone -l -s -n /pub/scm/.../torvalds/linux-2.6.git temp
$ mv temp/.git subproj-2.6.git
$ rmdir temp

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] checkout: show dirty state upon switching branches.
Junio C Hamano [Sat, 14 Jan 2006 20:31:18 +0000 (12:31 -0800)]
[PATCH] checkout: show dirty state upon switching branches.

This shows your working file state when you switch branches.  As
a side effect, "git checkout" without any branch name (i.e. stay
on the current branch) becomes a more concise shorthand for the
"git status" command.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-push: avoid falling back on pushing "matching" refs.
Junio C Hamano [Thu, 12 Jan 2006 23:29:12 +0000 (15:29 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agocheckout: merge local modifications while switching branches.
Junio C Hamano [Thu, 12 Jan 2006 22:04:36 +0000 (14:04 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agocheckout: automerge local changes while switching branches.
Junio C Hamano [Thu, 12 Jan 2006 07:07:27 +0000 (23:07 -0800)]
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.

This is still experimental.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge fixes up to GIT 1.1.2
Junio C Hamano [Sat, 14 Jan 2006 00:51:21 +0000 (16:51 -0800)]
Merge fixes up to GIT 1.1.2

15 years agoFix the installation location.
Junio C Hamano [Fri, 13 Jan 2006 05:37:17 +0000 (21:37 -0800)]
Fix the installation location.

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).

Also gitk needs to be installed in $(bindir).

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoExec git programs without using PATH.
Michal Ostrowski [Wed, 11 Jan 2006 02:12:17 +0000 (21:12 -0500)]
Exec git programs without using PATH.

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>
15 years agoGIT 1.1.2 v1.1.2
Junio C Hamano [Sat, 14 Jan 2006 00:47:05 +0000 (16:47 -0800)]
GIT 1.1.2

15 years agoGIT 1.0.10 v1.0.10
Junio C Hamano [Sat, 14 Jan 2006 00:39:17 +0000 (16:39 -0800)]
GIT 1.0.10

15 years agoDocumentation: git-reset - interrupted workflow.
Junio C Hamano [Fri, 13 Jan 2006 21:17:55 +0000 (13:17 -0800)]
Documentation: git-reset - interrupted workflow.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocumentation: git-commit -a
Junio C Hamano [Fri, 13 Jan 2006 20:59:21 +0000 (12:59 -0800)]
Documentation: git-commit -a

A bit more elaboration on what "update all paths" means.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agooctopus: allow criss-cross and clarify the message when it rejects
Junio C Hamano [Fri, 13 Jan 2006 09:37:09 +0000 (01:37 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocumentation: clarify fetch parameter descriptions.
J. Bruce Fields [Fri, 13 Jan 2006 03:08:37 +0000 (22:08 -0500)]
Documentation: clarify fetch parameter descriptions.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMakefile: add 'strip' target
Junio C Hamano [Fri, 13 Jan 2006 05:42:25 +0000 (21:42 -0800)]
Makefile: add 'strip' target

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoshow-branch: handle [] globs as well.
Junio C Hamano [Wed, 11 Jan 2006 08:20:25 +0000 (00:20 -0800)]
show-branch: handle [] globs as well.

Earlier only '?' and '*' signalled the command that what the
user has given is a glob pattern.  This prevented us to say:

$ git show-branch 'v0.99.[0-3]'

Now we notice '[' as well, so the above would work.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoname-rev: do not omit leading components of ref name.
Junio C Hamano [Wed, 11 Jan 2006 22:20:09 +0000 (14:20 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-describe: default to HEAD
Junio C Hamano [Wed, 11 Jan 2006 21:57:42 +0000 (13:57 -0800)]
git-describe: default to HEAD

This is based on the patch by Andreas Ericsson, but done slightly
differently, preferring to have separate loops -- one for options
and then arguments.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agodescribe: do not silently ignore indescribable commits
Junio C Hamano [Wed, 11 Jan 2006 21:41:25 +0000 (13:41 -0800)]
describe: do not silently ignore indescribable commits

We silently ignored indescribable commits without complaining.
Complain and die instead.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoupdate-index: work with c-quoted name
Junio C Hamano [Wed, 11 Jan 2006 21:36:45 +0000 (13:36 -0800)]
update-index: work with c-quoted name

update-index --stdin did not work with c-style quoted names even though
update-index --index-info did.  This fixes the inconsistency.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd git-describe to .gitignore.
Tom Prince [Wed, 11 Jan 2006 06:40:33 +0000 (23:40 -0700)]
Add git-describe to .gitignore.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd --keep option to keep downloaded packs to git-fetch.
Tom Prince [Wed, 11 Jan 2006 01:50:19 +0000 (18:50 -0700)]
Add --keep option to keep downloaded packs to git-fetch.

Signed-off-by: Tom Prince <tom.prince@ualberta.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge fixes up to GIT 1.1.1
Junio C Hamano [Wed, 11 Jan 2006 00:29:17 +0000 (16:29 -0800)]
Merge fixes up to GIT 1.1.1

15 years agoGIT 1.0.9 v1.0.9
Junio C Hamano [Wed, 11 Jan 2006 00:20:22 +0000 (16:20 -0800)]
GIT 1.0.9

15 years agoGIT 1.1.1 v1.1.1
Junio C Hamano [Wed, 11 Jan 2006 00:19:11 +0000 (16:19 -0800)]
GIT 1.1.1

15 years agoglossary: explain "master" and "origin"
Johannes Schindelin [Tue, 10 Jan 2006 21:26:46 +0000 (22:26 +0100)]
glossary: explain "master" and "origin"

If you are a long time git user/developer, you forget that to a new git
user, these words have not the same meaning as to you.

[jc: with updates from J. Bruce Fields.]

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoshow-branch: take default arguments from configuration file.
Junio C Hamano [Mon, 9 Jan 2006 21:29:23 +0000 (13:29 -0800)]
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:

[showbranch]
default = ko/master
default = heads/*

and say

$ git show-branch

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoGIT-VERSION-GEN: detect dirty tree and mark the version accordingly.
Junio C Hamano [Mon, 9 Jan 2006 22:25:10 +0000 (14:25 -0800)]
GIT-VERSION-GEN: detect dirty tree and mark the version accordingly.

If we are building from a working tree with local modifications,
mark the version accordingly.

Deliberately uses '-' to prevent RPM from being built from such
a tree.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFor release tarballs, include the proper version
H. Peter Anvin [Tue, 10 Jan 2006 02:07:01 +0000 (18:07 -0800)]
For release tarballs, include the proper version

When producing a release tarball, include a "version" file, which
GIT-VERSION-GEN can then use to do the right thing when building from a
tarball.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoGIT 1.1.0 v1.1.0
Junio C Hamano [Sun, 8 Jan 2006 22:22:19 +0000 (14:22 -0800)]
GIT 1.1.0

15 years agoAdd a test for rebase when a change was picked upstream
Yann Dirson [Sun, 8 Jan 2006 00:39:48 +0000 (01:39 +0100)]
Add a test for rebase when a change was picked upstream

This test exercises the standard feature that makes rebase useful.

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd a minimal test for git-cherry
Yann Dirson [Sun, 8 Jan 2006 00:38:15 +0000 (01:38 +0100)]
Add a minimal test for git-cherry

This test checks that git-cherry finds the expected number of patches
in two simple cases, and then tests the new limit arguments.

[jc: collapsed two patches into one and added sleep to make sure
 the two commits would get different timestamps]

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd an optional limit to git-cherry
Yann Dirson [Sun, 8 Jan 2006 00:40:33 +0000 (01:40 +0100)]
Add an optional limit to git-cherry

This allows to use another commit than the merge base as a limit for
scanning patches.

[jc: part about t3500 test omitted.]

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-daemon --base-path
Petr Baudis [Fri, 23 Dec 2005 01:27:40 +0000 (02:27 +0100)]
git-daemon --base-path

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>
15 years agodescribe: allow more than one revs to be named.
Junio C Hamano [Sun, 8 Jan 2006 02:52:42 +0000 (18:52 -0800)]
describe: allow more than one revs to be named.

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).

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agols-files --others --directory: test
Junio C Hamano [Sat, 7 Jan 2006 22:17:58 +0000 (14:17 -0800)]
ls-files --others --directory: test

Add a test to run with --directory option.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-status: use ls-files --others --directory for untracked list.
Junio C Hamano [Sat, 7 Jan 2006 22:32:07 +0000 (14:32 -0800)]
git-status: use ls-files --others --directory for untracked list.

This shortens "Untracked files" list by using --directory option
when running ls-files --others.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agols-files --others --directory: give trailing slash
Junio C Hamano [Sat, 7 Jan 2006 22:31:56 +0000 (14:31 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agols-files --others --directory: fix a bug with index entry ordering
Junio C Hamano [Sat, 7 Jan 2006 22:32:09 +0000 (14:32 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agols-files -o: optionally skip showing the contents in "untracked" directories
Linus Torvalds [Wed, 4 Jan 2006 21:31:25 +0000 (13:31 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-fetch: auto-following tags.
Junio C Hamano [Sat, 7 Jan 2006 08:48:04 +0000 (00:48 -0800)]
git-fetch: auto-following tags.

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoGIT 1.0.8 v1.0.8
Junio C Hamano [Sun, 8 Jan 2006 05:32:48 +0000 (21:32 -0800)]
GIT 1.0.8

15 years agomailsplit: allow empty input from stdin
Junio C Hamano [Sun, 8 Jan 2006 05:26:30 +0000 (21:26 -0800)]
mailsplit: allow empty input from stdin

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agorevert/cherry-pick: handle single quote in author name.
Junio C Hamano [Sun, 8 Jan 2006 04:45:11 +0000 (20:45 -0800)]
revert/cherry-pick: handle single quote in author name.

The same fix as aa66c7ec77d474b737da607d6cb2d07f56628def is
needed here.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix git-format-patch usage string wrt output modes.
Yann Dirson [Sun, 8 Jan 2006 00:39:16 +0000 (01:39 +0100)]
Fix git-format-patch usage string wrt output modes.

--stdout was not mentionned, and the description for the case where -o
was not given was thus incomplete.

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix typo in debug stanza of t2001
Yann Dirson [Sun, 8 Jan 2006 00:38:43 +0000 (01:38 +0100)]
Fix typo in debug stanza of t2001

Signed-off-by: Yann Dirson <ydirson@altern.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agotar-tree: finish honoring extractor's umask in git-tar-tree.
Junio C Hamano [Sat, 7 Jan 2006 23:11:11 +0000 (15:11 -0800)]
tar-tree: finish honoring extractor's umask in git-tar-tree.

Earlier commit 38ec15a973a1f075f0d94d130b0ef279562921cd forgot
to apply the same principle of not forcing go-w to the base
directory when specified.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Compilation: zero-length array declaration.
Junio C Hamano [Sat, 7 Jan 2006 09:33:54 +0000 (01:33 -0800)]
[PATCH] Compilation: zero-length array declaration.

ISO C99 (and GCC 3.x or later) lets you write a flexible array
at the end of a structure, like this:

struct frotz {
int xyzzy;
char nitfol[]; /* more */
};

GCC 2.95 and 2.96 let you to do this with "char nitfol[0]";
unfortunately this is not allowed by ISO C90.

This declares such construct like this:

struct frotz {
int xyzzy;
char nitfol[FLEX_ARRAY]; /* more */
};

and git-compat-util.h defines FLEX_ARRAY to 0 for gcc 2.95 and
empty for others.

If you are using a C90 C compiler, you should be able
to override this with CFLAGS=-DFLEX_ARRAY=1 from the
command line of "make".

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoprune: do not show error from pack-redundant when no packs are found.
Junio C Hamano [Sat, 7 Jan 2006 08:29:48 +0000 (00:29 -0800)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRetire debian/ directory.
Junio C Hamano [Sat, 7 Jan 2006 03:18:12 +0000 (19:18 -0800)]
Retire debian/ directory.

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agounpack-objects: default to quiet if stderr is not a tty.
Junio C Hamano [Sat, 7 Jan 2006 02:53:16 +0000 (18:53 -0800)]
unpack-objects: default to quiet if stderr is not a tty.

This would help cron/at jobs that run send-pack to mirror
repositories.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoSubstitute "/" with $opt_s in tag names as well as branch names
Joe English [Fri, 6 Jan 2006 20:52:27 +0000 (12:52 -0800)]
Substitute "/" with $opt_s in tag names as well as branch names

In 'git cvsimport' changes "/" to "-" (or $opt_s) in branch names,
but not in tag names, which is inconsistent.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoTeach cvsexportcommit to add new files
Yann Dirson [Fri, 6 Jan 2006 20:54:41 +0000 (21:54 +0100)]
Teach cvsexportcommit to add new files

"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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMake GIT-VERSION-GEN tolerate missing git describe command
John Ellson [Fri, 30 Dec 2005 16:23:17 +0000 (11:23 -0500)]
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>
15 years agoMerge fixes up to GIT 1.0.7
Junio C Hamano [Fri, 6 Jan 2006 04:54:42 +0000 (20:54 -0800)]
Merge fixes up to GIT 1.0.7

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoGIT 1.0.7 v1.0.7
Junio C Hamano [Fri, 6 Jan 2006 04:52:50 +0000 (20:52 -0800)]
GIT 1.0.7

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix git-symbolic-ref typo in git.txt.
Jon Loeliger [Tue, 3 Jan 2006 14:38:15 +0000 (08:38 -0600)]
Fix git-symbolic-ref typo in git.txt.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit: grok 'help' to mean '--help'.
Andreas Ericsson [Tue, 3 Jan 2006 09:53:54 +0000 (10:53 +0100)]
git: grok 'help' to mean '--help'.

Most other scm's understand it, most users expect it and it's an easy fix.

Signed-off-by: Andreas Ericsson <ae@op5.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocumentation/git-svnimport: document -T and -t switches correctly
Eric Wong [Sun, 1 Jan 2006 21:32:53 +0000 (13:32 -0800)]
Documentation/git-svnimport: document -T and -t switches correctly

The -T and -t switches are swapped in the documentation and actual
code.  I've made the documentation match the code.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agosvnimport: support repositories requiring SSL authentication
Eric Wong [Sun, 1 Jan 2006 21:25:47 +0000 (13:25 -0800)]
svnimport: support repositories requiring SSL authentication

I looked at svn-mirror to see how it did this, seems about right.
"It works for me" when using it against https://svn.musicpd.org

tested command-line: git-svnimport -C mpc -i -m -v  \
-T mpc/trunk -b mpc/branches -t mpc/tags https://svn.musicpd.org

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agot3300: skip when filesystem does not like TAB in filenames.
Junio C Hamano [Fri, 6 Jan 2006 04:14:40 +0000 (20:14 -0800)]
t3300: skip when filesystem does not like TAB in filenames.

Instead of checking Cygwin explicitly, see if the filesystem lets us
create funny filenames.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoformat-patch/commit: Quote single quote in the author name properly.
Junio C Hamano [Fri, 6 Jan 2006 03:50:37 +0000 (19:50 -0800)]
format-patch/commit: Quote single quote in the author name properly.

Noticed by Kyle McMartin.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-fetch --tags: reject malformed tags.
Junio C Hamano [Fri, 6 Jan 2006 03:42:12 +0000 (19:42 -0800)]
git-fetch --tags: reject malformed tags.

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoWrap synopsis lines and use [verse] to keep formatting
Jonas Fonseca [Sat, 31 Dec 2005 17:37:15 +0000 (18:37 +0100)]
Wrap synopsis lines and use [verse] to keep formatting

In addition, also fixes a few synopses to be more consistent and a gitlink.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agouse GIT_DIR instead of /var/tmp
Alex Riesen [Thu, 5 Jan 2006 11:52:07 +0000 (12:52 +0100)]
use GIT_DIR instead of /var/tmp

Not every system (will not one microsoft windows system) have /var/tmp,
whereas using GIT_DIR for random temporary files is more or less established.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>