git/git.git
15 years agogit-convert-cache: fix up mode conversion
Linus Torvalds [Wed, 27 Jul 2005 22:29:38 +0000 (15:29 -0700)]
git-convert-cache: fix up mode conversion

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.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agogit-fsck-cache.c: check commit objects more carefully
Linus Torvalds [Wed, 27 Jul 2005 22:16:03 +0000 (15:16 -0700)]
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.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoMake "parse_commit()" be a lot more careful
Linus Torvalds [Wed, 27 Jul 2005 22:12:48 +0000 (15:12 -0700)]
Make "parse_commit()" be a lot more careful

This was brought on by a bad tree of Thomas Gleixner, where some bogus
commit objects weren't warned about properly

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoTutorial typofix.
Linus Torvalds [Wed, 27 Jul 2005 03:04:22 +0000 (20:04 -0700)]
Tutorial typofix.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoTypofix: usage strings fix.
Junio C Hamano [Wed, 27 Jul 2005 07:04:16 +0000 (00:04 -0700)]
Typofix: usage strings fix.

The *_usage strings should not start with "usage: ", since the
usage() function gives its own.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-log: make sure we have some commit to start from.
Linus Torvalds [Wed, 27 Jul 2005 03:04:22 +0000 (20:04 -0700)]
git-log: make sure we have some commit to start from.

When no usable head/tag is specified, git log barfed with
underlying error message from rev-list, which was not helpful.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agorequest-pull: minor tweaks.
Junio C Hamano [Tue, 26 Jul 2005 18:47:31 +0000 (11:47 -0700)]
request-pull: minor tweaks.

 - 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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Add git-request-pull-script, a short script that generates a summary of pendi...
Ryan Anderson [Tue, 26 Jul 2005 07:30:36 +0000 (03:30 -0400)]
[PATCH] Add git-request-pull-script, a short script that generates a summary of pending changes

A short message requesting a pull from the repository is also included.

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] git-status-script was missed during the conversion from N -> A
Ryan Anderson [Wed, 27 Jul 2005 08:36:08 +0000 (04:36 -0400)]
[PATCH] git-status-script was missed during the conversion from N -> A

git-status-script was missed during the conversion from "N" to "A" as
the new-file marker flag.

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDeb Packaging fixes: Build against Mozilla libs for Debian, conflict with "git"
Ryan Anderson [Sat, 23 Jul 2005 19:26:32 +0000 (15:26 -0400)]
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>
15 years agoFix potential send-pack SIGSEGV
Linus Torvalds [Wed, 27 Jul 2005 03:04:22 +0000 (20:04 -0700)]
Fix potential send-pack SIGSEGV

The check that the source is ahead of the destination incorrectly expects
pop_most_recent_commit() to gracefully handle an empty list.

Fix by just checking the list itself, rather than the return value of the
pop function.

[jc: I did the test script that demonstrated the problem]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoName it 0.99.2 v0.99.2
Junio C Hamano [Wed, 27 Jul 2005 00:21:16 +0000 (17:21 -0700)]
Name it 0.99.2

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdjust diff-raw tests to the status letter change.
Junio C Hamano [Tue, 26 Jul 2005 07:22:43 +0000 (00:22 -0700)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agodiff-raw: Use 'A' instead of 'N' for added files.
Junio C Hamano [Mon, 25 Jul 2005 21:31:19 +0000 (14:31 -0700)]
diff-raw: Use 'A' instead of 'N' for added files.

This actually changes the diff-raw status letter from N to A
for added files.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse symbolic constants for diff-raw status indicators.
Junio C Hamano [Mon, 25 Jul 2005 20:05:44 +0000 (13:05 -0700)]
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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDocumentation: describe git-ls-files --exclude patterns.
Junio C Hamano [Sun, 24 Jul 2005 22:47:26 +0000 (15:47 -0700)]
Documentation: describe git-ls-files --exclude patterns.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-ls-files: --exclude mechanism updates.
Junio C Hamano [Sun, 24 Jul 2005 22:26:09 +0000 (15:26 -0700)]
git-ls-files: --exclude mechanism updates.

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUpdate the documentation for git-tag-script to reflect current behavior.
Ryan Anderson [Mon, 25 Jul 2005 06:17:16 +0000 (02:17 -0400)]
Update the documentation for git-tag-script to reflect current behavior.

[jc: I updated Ryan's patch to mention -a to create an unsigned
tag.]

Signed-off-by: Ryan Anderson <ryan@michonline.com>
15 years agogit-tag-script updates.
Junio C Hamano [Mon, 25 Jul 2005 22:18:35 +0000 (15:18 -0700)]
git-tag-script updates.

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.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix up applymbox script for the addition of "git-" prefix
Linus Torvalds [Sun, 24 Jul 2005 03:28:34 +0000 (20:28 -0700)]
Fix up applymbox script for the addition of "git-" prefix

Ehh, it works much better that way ;)

15 years ago[PATCH] diffcore-pickaxe: switch to "counting" behaviour.
Junio C Hamano [Sat, 23 Jul 2005 23:35:25 +0000 (16:35 -0700)]
[PATCH] diffcore-pickaxe: switch to "counting" behaviour.

Instead of finding old/new pair that one side has and the
other side does not have the specified string, find old/new pair
that contains the specified string as a substring different
number of times.  This would still not catch a case where you
introduce two static variable declarations and remove two static
function definitions from a file with -S"static", but would make
it behave a bit more intuitively.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoAdd a "git rename" to help with - surprise surprise - renames
Linus Torvalds [Sun, 24 Jul 2005 01:52:22 +0000 (18:52 -0700)]
Add a "git rename" to help with - surprise surprise - renames

It's stupid.  We'd want to rename directories too, but this doesn't do
that yet - easy enough to do per se, we just need to carefully list all
the pathnames that got moved (and remember to ignore the files that
weren't tracked but are in the subdirectory that got moved).

Doing the directory case will require a bit more scripting.. Something like

oldfiles=($(git-ls-files | grep '^$src'))
newfiles=($(git-ls-files | sed ':^$src: s:^$src:$dst:'))
mv $src $dst && git-update-cache --add --remove -- "${oldfiles[@]}" "${newfiles[@]}"

might do it, except it needs to be done right, and carefully.  Methinks
perl is probably better at this. Hint hint..

15 years agoRename the "tools" programs to also have the "git-" prefix
Linus Torvalds [Sun, 24 Jul 2005 01:46:27 +0000 (18:46 -0700)]
Rename the "tools" programs to also have the "git-" prefix

Hey, people are using them, and we have an install target for them, so
make sure that we can actually install them sanely without disturbing
the namespace.

15 years ago[PATCH] Support cloning packed repo from dumb http servers.
Junio C Hamano [Sun, 24 Jul 2005 00:55:07 +0000 (17:55 -0700)]
[PATCH] Support cloning packed repo from dumb http servers.

Using the information prepared with update-server-info, a truly
dumb http server can allow cloning with this client side
support.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Document update-server-info.
Junio C Hamano [Sun, 24 Jul 2005 00:54:56 +0000 (17:54 -0700)]
[PATCH] Document update-server-info.

This adds a minimum documentation to the new command.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Add update-server-info.
Junio C Hamano [Sun, 24 Jul 2005 00:54:41 +0000 (17:54 -0700)]
[PATCH] Add update-server-info.

The git-update-server-info command prepares informational files
to help clients discover the contents of a repository, and pull
from it via a dumb transport protocols.  Currently, the
following files are produced.

 - The $repo/info/refs file lists the name of heads and tags
   available in the $repo/refs/ directory, along with their
   SHA1.  This can be used by git-ls-remote command running on
   the client side.

 - The $repo/info/rev-cache file describes the commit ancestry
   reachable from references in the $repo/refs/ directory.  This
   file is in an append-only binary format to make the server
   side friendly to rsync mirroring scheme, and can be read by
   git-show-rev-cache command.

 - The $repo/objects/info/pack file lists the name of the packs
   available, the interdependencies among them, and the head
   commits and tags contained in them.  Along with the other two
   files, this is designed to help clients to make smart pull
   decisions.

The git-receive-pack command is changed to invoke it at the end,
so just after a push to a public repository finishes via "git
push", the server info is automatically updated.

In addition, building of the rev-cache file can be done by a
standalone git-build-rev-cache command separately.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] git-ls-remote: show and optionally store remote refs.
Junio C Hamano [Sun, 24 Jul 2005 00:54:26 +0000 (17:54 -0700)]
[PATCH] git-ls-remote: show and optionally store remote refs.

Retrieve and list the remote refs from git, http, and rsync
repositories, and optionally stores the retrieved refs in the
local repository under the same name.

To access a git URL, git-peek-remote command is used.  An http
URL needs to have an up-to-date info/refs file for discovery,
which will be introduced by a later update-server-info patch.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Documentation: git-peek-remote.
Junio C Hamano [Sun, 24 Jul 2005 00:54:15 +0000 (17:54 -0700)]
[PATCH] Documentation: git-peek-remote.

Add documentation for the git-peek-remote and link it from the
main index.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] git-peek-remote: show tags and heads from a remote repository.
Junio C Hamano [Sun, 24 Jul 2005 00:54:03 +0000 (17:54 -0700)]
[PATCH] git-peek-remote: show tags and heads from a remote repository.

Add a git-peek-remote command that talks with upload-pack the
same way git-fetch-pack and git-clone-pack do, to show the
references the remote side has on the standard output.

A later patch introduces git-ls-remote that implements a UI to
store tag values retrieved using this command.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoUpdate tutorial.txt branches/tags to use the nicer helper syntax
Linus Torvalds [Sat, 23 Jul 2005 22:24:53 +0000 (15:24 -0700)]
Update tutorial.txt branches/tags to use the nicer helper syntax

Teach people to use "git tag <tag-name>" instead of writing the current
HEAD by hand into the .git/refs/tags/<tag-name> file.  Most people
probably don't really want to know about how git does things internally.

15 years agoMake "git tag" create both simple and signed tags
Linus Torvalds [Sat, 23 Jul 2005 22:21:22 +0000 (15:21 -0700)]
Make "git tag" create both simple and signed tags

A simple tag is just a direct pointer to the object, while a signed tag
is a pointer to a "tag object" that has a pgp signature and points to
the object we tagged.

Use "git tag -s tagname" to create a signed tag.

The "-f" flag overwrites any previous tag of that name (useful if you
update a tag to point to a newer version for things like "latest" etc
tags that aren't necessarily static versions).

15 years ago[PATCH] mailinfo: handle folded header.
Junio C Hamano [Sat, 23 Jul 2005 09:10:31 +0000 (02:10 -0700)]
[PATCH] mailinfo: handle folded header.

Some people split their long E-mail address over two lines
using the RFC2822 header "folding".  We can lose authorship
information this way, so make a minimum effort to deal with it,
instead of special casing only the "Subject:" field.

We could teach mailsplit to unfold the folded header, but
teaching mailinfo about folding would make more sense; a single
message can be fed to mailinfo without going through mailsplit.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoFix git protocol connection 'port' override
Linus Torvalds [Sat, 23 Jul 2005 18:10:21 +0000 (11:10 -0700)]
Fix git protocol connection 'port' override

It was broken by the IPv6 patches - we need to remove the ":" part from
the hostname for a successful name lookup.

15 years ago[PATCH] GIT: Try all addresses for given remote name
YOSHIFUJI Hideaki [Thu, 21 Jul 2005 13:10:36 +0000 (09:10 -0400)]
[PATCH] GIT: Try all addresses for given remote name

Try all addresses for given remote name until it succeeds.  Also
supports IPv6.

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Deb packaging needs two more configuration files
Ryan Anderson [Fri, 22 Jul 2005 06:36:07 +0000 (02:36 -0400)]
[PATCH] Deb packaging needs two more configuration files

The deb package building needs these two new files to work correctly.

debian/compat sets the rules under which the debhelper scripts (dh_*)
operate.

debian/git-core.install tells dh_install what files to install in each
package that is generated.  There is only one package being generated,
so all files go into it.

(I missed these in the last patch, mostly because I needed to do this to
find stuff I had missed:
find . -name .git -type d -prune -o -type f -print \
| grep -v -e .tree1 -e .tree2 \
| sed -e "s/^\.\///" \
| sort >.tree1
git-ls-files | grep -v -e .tree1 -e .tree2 \
| sort >.tree2
diff -u .tree1 .tree2
)

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Deb packages should include the binaries
Ryan Anderson [Fri, 22 Jul 2005 05:55:56 +0000 (01:55 -0400)]
[PATCH] Deb packages should include the binaries

The Deb packages were missing a dependency on "build install" from the
binary target - this fixes that, and cleans up some inconsistencies
elsewhere in the rulesets.

Traditionally, Debian packaging uses a file called "build-stamp" (or
"install-stamp", etc) in the main source tree.  The initial deb package
support for Git tried to move this "build-stamp" file into the debian/
directory, but some instances were missed.  That problem, however, was
incidental - the real fix is the missing dependency mentioned above.

Signed-off-by: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoBe more aggressive about marking trees uninteresting
Linus Torvalds [Sat, 23 Jul 2005 17:01:49 +0000 (10:01 -0700)]
Be more aggressive about marking trees uninteresting

We'll mark all the trees at the edges (as deep as we had to go to
realize that we have all the commits needed) as uninteresting.
Otherwise we'll occasionally list a lot of objects that were actually
available at the edge in a commit that we just never ended up parsing
because we could determine early that we had all relevant commits.

NOTE! The object listing is still just a _heuristic_.  It's guaranteed
to list a superset of the actual new objects, but there might be the
occasional old object in the list, just because the commit that
referenced it was much further back in the history.

For example, let's say that a recent commit is a revert of part of the
tree to much older state: since we didn't walk _that_ far back in the
commit history tree to list the commits necessary, git-rev-tree will
never have marked the old objects uninteresting, and we'll end up
listing them as "new".

That's ok.

15 years ago[PATCH] tutorial: mention "git clone" records .git/branches/origin
Junio C Hamano [Sat, 23 Jul 2005 02:13:07 +0000 (19:13 -0700)]
[PATCH] tutorial: mention "git clone" records .git/branches/origin

Update the recommended workflow for individual developers.
While they are tracking the origin, refs/heads/origin is updated
by "git fetch", so there is no need to manually copy FETCH_HEAD
to refs/heads/ anywhere.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] git-clone-script: store where we cloned from in .git/branches/origin
Junio C Hamano [Sat, 23 Jul 2005 02:11:22 +0000 (19:11 -0700)]
[PATCH] git-clone-script: store where we cloned from in .git/branches/origin

A bit more usability enhancement, while retaining Cogito
compatibility (and fixing the "-u" flag).

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Test framework: prettyprint the failed command.
Junio C Hamano [Sat, 23 Jul 2005 02:09:34 +0000 (19:09 -0700)]
[PATCH] Test framework: prettyprint the failed command.

The output from a failure case had the test description message
and the first line of the actual test script concatenated on the
same line, which was ugly.  Correct the output routine a bit to
make it more readable.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Install tools with "make install-tools".
Junio C Hamano [Sat, 23 Jul 2005 02:09:20 +0000 (19:09 -0700)]
[PATCH] Install tools with "make install-tools".

Match the main Makefile by separating COPTS from CFLAGS,
defining INSTALL, prefix, and bin.  Add a new target 'install-tools'
to the main Makefile to install them.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Audit rev-parse users.
Junio C Hamano [Sat, 23 Jul 2005 02:09:05 +0000 (19:09 -0700)]
[PATCH] Audit rev-parse users.

This patch changes rev-parse users that pass a single argument
that is supposed to be a rev parameter to use "--verify".

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] git-branch: avoid getting confused by empty or nonexisting branches.
Junio C Hamano [Sat, 23 Jul 2005 02:08:47 +0000 (19:08 -0700)]
[PATCH] git-branch: avoid getting confused by empty or nonexisting branches.

When not specifying the start point explicitly, we ended up
emitting ^0 in addition to the default HEAD.  Be careful to see
if we have "$2" before finding out which commit to base the new
branch on.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Help scripts that use git-rev-parse to grok args with SP/TAB/LF
Junio C Hamano [Sat, 23 Jul 2005 02:08:32 +0000 (19:08 -0700)]
[PATCH] Help scripts that use git-rev-parse to grok args with SP/TAB/LF

The git-rev-parse command uses LF to separate each argument it
parses, so its users at least need to set IFS to LF to be able
to handle filenames with embedded SPs and TABs.  Some commands,
however, can take and do expect arguments with embedded LF,
notably, "-S" (pickaxe) of diff family, so even this workaround
does not work for them.

When --sq flag to git-rev-parse is given, instead of showing one
argument per line, it outputs arguments quoted for consumption
with "eval" by the caller, to remedy this situation.

As an example, this patch converts git-whatchanged to use this
new feature.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Document "curl" requirements.
Junio C Hamano [Sat, 23 Jul 2005 00:30:26 +0000 (17:30 -0700)]
[PATCH] Document "curl" requirements.

Not just libcurl, but now we require curl executable as well.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Support more http features: https no cert, .netrc -> auth
Darrin Thompson [Wed, 13 Jul 2005 02:12:40 +0000 (21:12 -0500)]
[PATCH] Support more http features: https no cert, .netrc -> auth

Cause setting environment variable GIT_SSL_NO_VERIFY to turn off
curl's ssl peer verification.

Only use curl for http transfers, instead of curl and wget.

Make curl check ~/.netrc for credentials.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Install git-verify-tag-script
Eric W. Biederman [Sat, 16 Jul 2005 20:19:24 +0000 (14:19 -0600)]
[PATCH] Install git-verify-tag-script

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Cleanup: git-verify-tag-script
Eric W. Biederman [Sat, 16 Jul 2005 20:18:40 +0000 (14:18 -0600)]
[PATCH] Cleanup: git-verify-tag-script

- Use git-rev-parse to allow sha1 tags references
- When the tag does not verify set an appropriate exit status
- Use git-sh-setup-script to verify the .git directory

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Fix a typo in git-unpack-objects documentation.
Jan Veldeman [Tue, 19 Jul 2005 21:59:09 +0000 (14:59 -0700)]
[PATCH] Fix a typo in git-unpack-objects documentation.

Fix a typo in git-unpack-objects documentation.

Signed-off-by: Jan Veldeman <jan@mind.be>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] debian/ fixes
Ryan Anderson [Wed, 20 Jul 2005 01:17:47 +0000 (21:17 -0400)]
[PATCH] debian/ fixes

Make debian/rules executable, and correct the spelling of rsync in
debian/control.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] format-patch: --mbox and --check.
Junio C Hamano [Fri, 22 Jul 2005 23:04:47 +0000 (16:04 -0700)]
[PATCH] format-patch: --mbox and --check.

Add --mbox option to export patches in a format resembling UNIX
mbox, so that later they can be concatenated and fed to
applymbox.

Add --check to look for lines that introduce bogus whitespaces.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] git-format-patch-script and mailinfo updates.
Junio C Hamano [Fri, 22 Jul 2005 23:04:34 +0000 (16:04 -0700)]
[PATCH] git-format-patch-script and mailinfo updates.

 - avoid duplicating [PATCH] in the commit message body if the
   original commit has it already (happens for commits done from
   mails via applymbox).

 - check if the commit author is different from the one who is
   running the script, and emit an appropriate "From:" and
   "Date: " lines to the output.

 - with '--date', emit "Date: " line to preserve the original
   author date even for the user's own commit.

 - teach mailinfo to grok not just "From: " but "Date: ".

The patch e-mail output by format-patch starts with the first
line from the original commit message, prefixed with [PATCH],
and optionally a From: line if you are reformatting a patch
obtained from somebody else, a Date: line from the original
commit if (1) --date is specified or (2) for somebody else's
patch, and the rest of the commit message body.

Expected use of this is to move the title line from the commit
to Subject: when sending it via an e-mail, and leave the From:
and the Date: lines as the first lines of your message.

The mailinfo command has been changed to read Date: (in addition
to From: it already understands) and do sensible things when
running applymbox.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] tools/applymbox: allow manual fixing and continuing after a failure.
Junio C Hamano [Fri, 22 Jul 2005 23:04:19 +0000 (16:04 -0700)]
[PATCH] tools/applymbox: allow manual fixing and continuing after a failure.

With "-c .dotest/0002" flag, the applymbox command can be told
to use existing .dotest/patch file after hand-fixing the patch
conflicts for the second patch in the mailbox, and continue on
from that message, skipping the already applied first patch in
.dotest/0001.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] apply.c: --exclude=fnmatch-pattern option.
Junio C Hamano [Fri, 22 Jul 2005 16:56:57 +0000 (09:56 -0700)]
[PATCH] apply.c: --exclude=fnmatch-pattern option.

Adds --exclude=pattern option to the "git-apply" command.  This
was useful while reimporting the BKCVS patchset dump of the
Linux kernel, starting at 2.4.0 and ending at 2.6.12-rc2 Ingo
announced some time ago to exclude BitKeeper directory.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] apply.c: handle incomplete lines correctly.
Junio C Hamano [Fri, 22 Jul 2005 16:56:39 +0000 (09:56 -0700)]
[PATCH] apply.c: handle incomplete lines correctly.

The parsing code had a bug that failed to recognize an
incomplete line at the end of a fragment, and the fragment
application code had a comparison bug to recognize such.  Fix
them to handle incomplete lines correctly.

Add a test script for patches with various combinations of
complete and incomplete lines to make sure the fix works.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoAdd "git commit --all" since everybody seems to want it.
Linus Torvalds [Tue, 19 Jul 2005 11:20:39 +0000 (07:20 -0400)]
Add "git commit --all" since everybody seems to want it.

This replaces

git-diff-files --name-only | xargs git-update-cache
git commit

with a new "--all" argument to "git commit".

15 years agogit-send-pack: verify that sender is a proper superset of receiver
Linus Torvalds [Tue, 19 Jul 2005 11:03:47 +0000 (07:03 -0400)]
git-send-pack: verify that sender is a proper superset of receiver

This should make sure that if you have multiple people pushing to the
same tree, they cannot overwrite each others work, but have to merge
before being able to update the common reference tree.

15 years agoMerge master.kernel.org:/pub/scm/gitk/gitk
Linus Torvalds [Mon, 18 Jul 2005 23:26:00 +0000 (19:26 -0400)]
Merge master.kernel.org:/pub/scm/gitk/gitk

15 years agoMake the diff display less gaudy.
Paul Mackerras [Mon, 18 Jul 2005 18:29:03 +0000 (14:29 -0400)]
Make the diff display less gaudy.

The old style is still available as an option (we still need a
preferences window so we can set/clear it though).

15 years agoAllow short SHA1 IDs in the SHA1 entry field.
Paul Mackerras [Mon, 18 Jul 2005 16:16:35 +0000 (12:16 -0400)]
Allow short SHA1 IDs in the SHA1 entry field.

15 years agoFix display of mode changes in details pane.
Paul Mackerras [Sun, 17 Jul 2005 15:11:44 +0000 (11:11 -0400)]
Fix display of mode changes in details pane.

Also simplified the parsing of the git-diff-tree -p output and got
rid of the unused 'seenfile' variable.

15 years agoFirst cut at displaying the diffs for a merge.
Paul Mackerras [Sun, 17 Jul 2005 07:39:44 +0000 (03:39 -0400)]
First cut at displaying the diffs for a merge.

We display the files where the child differs from the first parent
first in black in the file list window, followed by the second parent
in blue, and so on using different colors for each parent (provided
you don't have more than 6 parents; if you do we cycle around).

15 years agoMake searching in files changed faster, and fix some bugs.
Paul Mackerras [Sun, 17 Jul 2005 01:53:55 +0000 (21:53 -0400)]
Make searching in files changed faster, and fix some bugs.

We now kick off a single git-diff-tree -r --stdin and feed it all
the commit pairs we want to know about, instead of doing a separate
git-diff-tree invocation for each.

15 years agoMerge three separate "fetch refs" functions
Linus Torvalds [Sat, 16 Jul 2005 20:55:50 +0000 (13:55 -0700)]
Merge three separate "fetch refs" functions

It really just boils down to one "get_remote_heads()" function, and a
common "struct ref" structure definition.

15 years agogit-send-pack: add "--all" option to send all refs to the other side
Linus Torvalds [Sat, 16 Jul 2005 20:26:33 +0000 (13:26 -0700)]
git-send-pack: add "--all" option to send all refs to the other side

This affects only refs that the other side doesn't already have. The
ones it has are still filtered by the ref selection.

15 years agogit-fetch-script: fix http:// breakage
Linus Torvalds [Sat, 16 Jul 2005 17:31:38 +0000 (10:31 -0700)]
git-fetch-script: fix http:// breakage

We were trying to fetch using the merge-head name rather than the
merge-head SHA1 that we just got.

Now, http:// is broken anyway right now for packing, but this should
make it work for nonpacked repositories again.

15 years agoFix git-fetch-script breakage
Linus Torvalds [Sat, 16 Jul 2005 17:27:19 +0000 (10:27 -0700)]
Fix git-fetch-script breakage

It had the test for the destination reversed.

15 years agoMerge git-tools repository under "tools" subdirectory
Linus Torvalds [Sat, 16 Jul 2005 17:12:06 +0000 (10:12 -0700)]
Merge git-tools repository under "tools" subdirectory

Thanks to Ryan Anderson for setting me up to do this.  I'd have used his
work, but I wanted to clean up the old git-tools repository before
merging it: it had old-style file modes etc that needed a round of
git-convert-cache to fix up.

15 years agoPrepare git-tools for merging into the main git archive
Linus Torvalds [Sat, 16 Jul 2005 17:05:26 +0000 (10:05 -0700)]
Prepare git-tools for merging into the main git archive

Rename into a "tools" subdirectory, and change name of "dotest" to "applymbox".

Remove stripspace (which was already copied into git) and cvs2git (which
was likewise already copied into git, and then replaced by a much better
perl version).

All of this was brought on by Ryan Anderson shaming me into it.  Thanks.
I guess.

15 years agogit-convert-cache: fix up file modes in trees too
Linus Torvalds [Sat, 16 Jul 2005 16:57:03 +0000 (09:57 -0700)]
git-convert-cache: fix up file modes in trees too

git-fsck-cache complains about some of the odder ones, and is quiet
about the old (S_IFREG | 664) case, but that's wrong too.

Converting the kernel tree is too painful right now, but at least we
know how to do it if we ever want to.

15 years ago[PATCH] Documentation: describe short-hand used in fetch/pull.
Junio C Hamano [Sat, 16 Jul 2005 07:17:42 +0000 (00:17 -0700)]
[PATCH] Documentation: describe short-hand used in fetch/pull.

Describe short-hand for remote repository used in fetch/pull.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] fetch/pull: short-hand notation for remote repositories.
Junio C Hamano [Sat, 16 Jul 2005 07:16:24 +0000 (00:16 -0700)]
[PATCH] fetch/pull: short-hand notation for remote repositories.

Since pull and fetch are done often against the same remote
repository repeatedly, keeping the URL to pull from along with
the name of the head to use in $GIT_DIR/branches/$name makes a
lot of sense.  Adopt that convention from Cogito, and try to be
compatible when possible; storing a partial URL and completing
it with a trailing path may not be understood by Cogito.

While we are at it, fix pulling a tag.  Earlier, we updated only
refs/tags/$tag without updating FETCH_HEAD, and called
resolve-script using a stale (or absent) FETCH_HEAD.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoAdd "Files" and "Pickaxe" to the find menu.
Paul Mackerras [Sat, 16 Jul 2005 11:46:13 +0000 (07:46 -0400)]
Add "Files" and "Pickaxe" to the find menu.

"Files" matches the find string against each of the files modified
by each commit, and can do exact, case-ignoring or regexp matching.

"Pickaxe" uses git-diff-tree -S'string' and can only do exact
matching.  I called it "pickaxe" rather than "find within patch"
since it only finds commits where the string is present in the child
but not the parents or vice versa, and "pickaxe" is what the author
of that feature calls it.

15 years agoBe slightly smarter about git-daemon client shutdown
Linus Torvalds [Sat, 16 Jul 2005 05:53:31 +0000 (22:53 -0700)]
Be slightly smarter about git-daemon client shutdown

Shut down connections that haven't even identified themselves as git
clients first.  That should get rid of people who just connect to the
port and wait for something to happen.

15 years ago[PATCH] Documentation: update recommended workflow when working with others.
Junio C Hamano [Sat, 16 Jul 2005 03:56:12 +0000 (20:56 -0700)]
[PATCH] Documentation: update recommended workflow when working with others.

Clarify that the hierarchy implied by the recommended workflow
is only informal.

Refer readers to nice illustration by Randy Dunlap.

Separate out the step to "push" to own public repository in the
workflow.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Documentation: adjust cvsimport command line.
Junio C Hamano [Sat, 16 Jul 2005 03:54:04 +0000 (20:54 -0700)]
[PATCH] Documentation: adjust cvsimport command line.

The cvsimport example in the cvs migration document was still
using the old syntax for target repository after new and
improved cvsimport-script was merged.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agogit-daemon: actually remember the children we have outstanding
Linus Torvalds [Sat, 16 Jul 2005 04:51:57 +0000 (21:51 -0700)]
git-daemon: actually remember the children we have outstanding

This is using a lockless approach that allows us to handle children
dying without having to block SIGCHLD.

Right now our "solution" to too many kids is pretty damn rough, but it
at least shows what you can do.

15 years agogit-daemon: keep track of children
Linus Torvalds [Sat, 16 Jul 2005 03:42:28 +0000 (20:42 -0700)]
git-daemon: keep track of children

We don't want them as zombies, and eventually we'll want to limit their
number. Right now we just count them.

15 years agoAhh, the heady days of 0.99 patchfiles v0.99.1
Linus Torvalds [Fri, 15 Jul 2005 23:08:01 +0000 (16:08 -0700)]
Ahh, the heady days of 0.99 patchfiles

Now, not all projects can be as refined as Linux.  Before the final 1.0
release, we went through fifteen 0.99 patchfiles, and pl14 alone went
through subreleases 'a' through 'z'. Now _that_ is a release process.

Not to mention the odd-ball releases, like 0.96c+

Sadly, in this day and age of RPM's etc, we have silly limitations, and
I cannot call this release '0.99pl5a or some such awe-inspiring name
just because "rpmbuild" is such a party pooper.  So it's just 0.99.1.

Oh well.  Aspiring to such greatness as the Linux release numbering is
hubris anyway.  You can attain such perfection only once in your life.

15 years agoFix up "make doc"
Linus Torvalds [Fri, 15 Jul 2005 21:57:09 +0000 (14:57 -0700)]
Fix up "make doc"

Fix 'git-var.txt' and use "-b xhtml11" instead of "-b css-embedded" to
make asciidoc 7.0.1 happy.

15 years ago[PATCH] Documentation: pull, push, packing repository and working with others.
Junio C Hamano [Fri, 15 Jul 2005 18:40:56 +0000 (11:40 -0700)]
[PATCH] Documentation: pull, push, packing repository and working with others.

Describe where you can pull from with a bit more detail.
Clarify description of pushing.

Add a section on packing repositories.

Add a section on recommended workflow for the project lead,
subsystem maintainers and individual developers.

Move "Tag" section around to make the flow of example simpler to
follow.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Documentation: update tutorial to talk about push.
Junio C Hamano [Fri, 15 Jul 2005 07:59:34 +0000 (00:59 -0700)]
[PATCH] Documentation: update tutorial to talk about push.

Talk about publishing to a public repository.  Also fixes a
couple of typos.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Initial support for building a debian package (.deb)
Eric W. Biederman [Fri, 15 Jul 2005 01:26:31 +0000 (19:26 -0600)]
[PATCH] Initial support for building a debian package (.deb)

It's not any harder to include debian package support than to include a
spec file so here is the setup to build the equivalent debian package.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Update the spec file so it can build and install the documentation
Eric W. Biederman [Fri, 15 Jul 2005 01:24:23 +0000 (19:24 -0600)]
[PATCH] Update the spec file so it can build and install the documentation

If you don't want the documentation simply build with

make RPMBUILD="rpmbuild --without docs"

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Add doc and install-doc targets to the Makefile
Eric W. Biederman [Fri, 15 Jul 2005 01:21:57 +0000 (19:21 -0600)]
[PATCH] Add doc and install-doc targets to the Makefile

This makes it straightforward for people wanting to build and install
the git man pages and the rest of the documentation to do so.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Use gzip -f when building the git-core tarball
Eric W. Biederman [Fri, 15 Jul 2005 01:20:50 +0000 (19:20 -0600)]
[PATCH] Use gzip -f when building the git-core tarball

This allows rebuilding the tarball when it is already present
without having to answer annoying questions from gzip

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Add a RPMBUILD make variable
Eric W. Biederman [Fri, 15 Jul 2005 01:20:25 +0000 (19:20 -0600)]
[PATCH] Add a RPMBUILD make variable

This allows RPMBUILD to be overridden for people with
old versions of rpm or people who want to pass rpmbuild extra options.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Update tags to record who made them
Eric W. Biederman [Fri, 15 Jul 2005 01:02:10 +0000 (19:02 -0600)]
[PATCH] Update tags to record who made them

And finally what all of this has been leading up to.
The 2 line code change to record who made a tag,
and the 8 line code change to check that we recorded
the tag.

Gosh the error checking is always so much bigger than the code :)

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Update git-tag-script to create the .git/refs/tags if it does not already...
Eric W. Biederman [Fri, 15 Jul 2005 01:00:15 +0000 (19:00 -0600)]
[PATCH] Update git-tag-script to create the .git/refs/tags if it does not already exist

When testing tags I ran into an interesting problem.
git-tag-script dies if .git/refs/tags/ does not exist.
And that directory didn't get created when I build my repository,
so we need to create it if it doesn't exist.

Signed-of-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Update the list of diagnostics for git-commit-tree
Eric W. Biederman [Fri, 15 Jul 2005 00:57:53 +0000 (18:57 -0600)]
[PATCH] Update the list of diagnostics for git-commit-tree

With the recent work on setup_ident() there are
a few more possible diagnostic messages form git-commit-tree

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Add git-var a tool for reading interesting git variables.
Eric W. Biederman [Fri, 15 Jul 2005 00:55:09 +0000 (18:55 -0600)]
[PATCH] Add git-var a tool for reading interesting git variables.

Sharing code between shell scripts and C is a challenge.  The program
git-var allows us to have a set of named values that a shell script can
interrogate and a normal C program can simply call the functions that
compute them.  Allowing sharing when computing plain test values.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] ident.c: Disambiguate the error messages in setup_ident
Eric W. Biederman [Fri, 15 Jul 2005 00:52:31 +0000 (18:52 -0600)]
[PATCH] ident.c: Disambiguate the error messages in setup_ident

If your user name is too long it is your sysadmin who
hates you not your parents!

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
[ Fixed grammar ]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years ago[PATCH] Move git_author_info and git_commiter_info to ident.c
Eric W. Biederman [Fri, 15 Jul 2005 00:50:33 +0000 (18:50 -0600)]
[PATCH] Move git_author_info and git_commiter_info to ident.c

Moving these functions allows all of the logic for figuring out what
these values are to be shared between programs.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
15 years agoAdd "--inetd" flag to git-daemon
Linus Torvalds [Fri, 15 Jul 2005 16:32:16 +0000 (09:32 -0700)]
Add "--inetd" flag to git-daemon

All credit go to Alexey Nezhdanov <snake@penza-gsm.ru>, I just ended up
re-implementing his idea.

15 years agogit-daemon: re-organize code a bit for --inetd flag
Linus Torvalds [Fri, 15 Jul 2005 16:27:05 +0000 (09:27 -0700)]
git-daemon: re-organize code a bit for --inetd flag

Alexey Nezhdanov sent a patch that made git-daemon usable from inetd (ie
where inetd has already done the accept on the new connection, the fork,
and the setup of stdin/stdout).  I wanted to organize the thing slightly
differently, though.

15 years agoSplit up "diff_format" into "format" and "line_termination".
Linus Torvalds [Fri, 15 Jul 2005 00:59:17 +0000 (17:59 -0700)]
Split up "diff_format" into "format" and "line_termination".

This removes the separate "formats" for name and name-with-zero-
termination.

It also removes the difference between HUMAN and MACHINE formats, and
they both become DIFF_FORMAT_RAW, with the difference being just in the
line and inter-filename termination.

It also makes the code easier to understand.

15 years agoMake "ce_match_path()" a generic helper function
Linus Torvalds [Thu, 14 Jul 2005 23:55:06 +0000 (16:55 -0700)]
Make "ce_match_path()" a generic helper function

... and make git-diff-files use it too.  This all _should_ make the
diffcore-pathspec.c phase unnecessary, since the diff'ers now all do the
path matching early interally.

15 years agoMake git-diff-cache skip any comparisons which don't match pathspec
Linus Torvalds [Thu, 14 Jul 2005 23:43:01 +0000 (16:43 -0700)]
Make git-diff-cache skip any comparisons which don't match pathspec

This brings all the same pathspec optimizations that git-diff-tree does
to git-diff-cache.

15 years agoStart using the partial tree reading in "git-diff-cache"
Linus Torvalds [Thu, 14 Jul 2005 20:19:19 +0000 (13:19 -0700)]
Start using the partial tree reading in "git-diff-cache"

The reason I say "start using" is that we really should also limit the
index checking by name - now we limit the tree object accesses by name,
but we still check the whole index.

Still, this should help.

15 years agoFix up read_tree() pathspec matching to use "const char **"
Linus Torvalds [Thu, 14 Jul 2005 18:39:27 +0000 (11:39 -0700)]
Fix up read_tree() pathspec matching to use "const char **"

The same way the other pathspecs work.  Also fix missing success return
from the matching - not that anything actually uses this yet ;)