git/git.git
15 years agoDisable hooks during tests.
Junio C Hamano [Sun, 16 Oct 2005 07:24:34 +0000 (00:24 -0700)]
Disable hooks during tests.

Individual tests for hooks would want to have their own tests when
written.  Also we should not pick up from random templates the user
happens to have.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoSparse fixes for http-fetch
Peter Hagervall [Sun, 16 Oct 2005 07:01:08 +0000 (00:01 -0700)]
Sparse fixes for http-fetch

This patch cleans out all sparse warnings from http-fetch.c

I'm a bit uncomfortable with adding extra #ifdefs to avoid either
'mixing declaration with code' or 'unused variable' warnings, but I
figured that since those functions are already littered with #ifdefs I
might just get away with it. Comments?

[jc: I adjusted Peter's patch to address uncomfortableness issues.]

Signed-off-by: Peter Hagervall <hager@cs.umu.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agowhatchanged: document -m option from git-diff-tree.
Junio C Hamano [Sun, 16 Oct 2005 06:49:27 +0000 (23:49 -0700)]
whatchanged: document -m option from git-diff-tree.

The documentation for git-whatchanged is meant to describe only
the most frequently used options from git-diff-tree.  Because "why
doesn't it show merges" was asked more than once, we'd better
describe '-m' option there.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoShow peeled onion from upload-pack and server-info.
Junio C Hamano [Fri, 14 Oct 2005 01:57:40 +0000 (18:57 -0700)]
Show peeled onion from upload-pack and server-info.

This updates git-ls-remote to show SHA1 names of objects that are
referred by tags, in the "ref^{}" notation.

This would make git-findtags (without -t flag) almost trivial.

    git-peek-remote . |
    sed -ne "s:^$target "'refs/tags/\(.*\)^{}$:\1:p'

Also Pasky could do:

    git-ls-remote --tags $remote |
    sed -ne 's:\( refs/tags/.*\)^{}$:\1:p'

to find out what object each of the remote tags refers to, and
if he has one locally, run "git-fetch $remote tag $tagname" to
automatically catch up with the upstream tags.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoIntroduce notation "ref^{type}".
Junio C Hamano [Fri, 14 Oct 2005 01:57:40 +0000 (18:57 -0700)]
Introduce notation "ref^{type}".

Existing "tagname^0" notation means "dereference tag zero or more
times until you cannot dereference it anymore, and make sure it is a
commit -- otherwise barf".  But tags do not necessarily reference
commit objects.

This commit introduces a bit more generalized notation, "ref^{type}".
Existing "ref^0" is a shorthand for "ref^{commit}".  If the type
is empty, it just dereferences tags until it hits a non-tag object.

With this, "git-rev-parse --verify 'junio-gpg-pub^{}'" shows the blob
object name -- there is no need to manually read the tag object and
find out the object name anymore.

"git-rev-parse --verify 'HEAD^{tree}'" can be used to find out the
tree object name of the HEAD commit.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoIgnore funny refname sent from remote
Junio C Hamano [Fri, 14 Oct 2005 01:57:40 +0000 (18:57 -0700)]
Ignore funny refname sent from remote

This allows the remote side (most notably, upload-pack) to show
additional information without affecting the downloader.  Peek-remote
does not ignore them -- this is to make it useful for Pasky's
automatic tag following.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRefuse to create funny refs in clone-pack, git-fetch and receive-pack.
Junio C Hamano [Fri, 14 Oct 2005 01:57:39 +0000 (18:57 -0700)]
Refuse to create funny refs in clone-pack, git-fetch and receive-pack.

Using git-check-ref-format, make sure we do not create refs with
funny names when cloning from elsewhere (clone-pack), fast forwarding
local heads (git-fetch), or somebody pushes into us (receive-pack).

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-check-ref-format: reject funny ref names.
Junio C Hamano [Fri, 14 Oct 2005 01:57:39 +0000 (18:57 -0700)]
git-check-ref-format: reject funny ref names.

Update check_ref_format() function to reject ref names that:

 * has a path component that begins with a ".", or
 * has a double dots "..", or
 * has ASCII control character, "~", "^", ":" or SP, anywhere, or
 * ends with a "/".

Use it in 'git-checkout -b', 'git-branch', and 'git-tag' to make sure
that newly created refs are well-formed.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fixes'
Junio C Hamano [Sat, 15 Oct 2005 18:19:09 +0000 (11:19 -0700)]
Merge branch 'fixes'

15 years agoShow curl error a bit better.
Junio C Hamano [Sat, 15 Oct 2005 18:13:55 +0000 (11:13 -0700)]
Show curl error a bit better.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoSome curl versions lack curl_easy_duphandle()
Johannes Schindelin [Sat, 15 Oct 2005 18:10:46 +0000 (11:10 -0700)]
Some curl versions lack curl_easy_duphandle()

Hi,

On Fri, 14 Oct 2005, Junio C Hamano wrote:

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
>
> > This patch looks bigger than it really is: The code to get the
> > default handle was refactored into a function, and is called
> > instead of curl_easy_duphandle() if that does not exist.
>
> I'd like to take Nick's config file patch first, which
> unfortunately interferes with your patch.  I'd hate to ask you
> this, but could you rebase it on top of Nick's patch, [...]

No need to hate it. Here comes the rebased patch, and this time, I
actually tested it a bit.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUnlocalized isspace and friends
Linus Torvalds [Thu, 13 Oct 2005 18:03:18 +0000 (11:03 -0700)]
Unlocalized isspace and friends

Do our own ctype.h, just to get the sane semantics: we want
locale-independence, _and_ we want the right signed behaviour. Plus we
only use a very small subset of ctype.h anyway (isspace, isalpha,
isdigit and isalnum).

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse config file settings for http
Nick Hengeveld [Sat, 15 Oct 2005 00:16:01 +0000 (17:16 -0700)]
Use config file settings for http

Use "http." config file settings if they exist.  Environment variables
still work, and they will override config file settings.

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-http-fetch: Remove size limit for objects/info/{packs,alternates}
Sergey Vlasov [Thu, 13 Oct 2005 17:49:53 +0000 (10:49 -0700)]
git-http-fetch: Remove size limit for objects/info/{packs,alternates}

git-http-fetch received objects/info/packs into a fixed-size buffer
and started to fail when this file became larger than the buffer.
Change it to grow the buffer dynamically, and do the same thing for
objects/info/alternates.  Also add missing free() calls for these
buffers.

Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agodebian packaging
Aneesh Kumar [Sat, 15 Oct 2005 00:15:11 +0000 (17:15 -0700)]
debian packaging

Make it build with stable testing and unstable.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoSupport git+ssh:// and ssh+git:// URL
Linus Torvalds [Sat, 15 Oct 2005 00:14:56 +0000 (17:14 -0700)]
Support git+ssh:// and ssh+git:// URL

It seemed to be such a stupid syntax. It's both what "ssh://" means,
and it's what not specifying a protocol at _all_ means.

But hey, since we already have two ways of saying "use ssh with
pack-files", here's two more.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd new programs and stamp file to .gitignore.
Tom Prince [Sat, 15 Oct 2005 00:13:49 +0000 (17:13 -0700)]
Add new programs and stamp file to .gitignore.

Signed-off-by: Tom Prince <tom.prince@ualberta.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoformat-patch: take sequence of ranges.
Junio C Hamano [Fri, 14 Oct 2005 06:18:46 +0000 (23:18 -0700)]
format-patch: take sequence of ranges.

This enhances set of revs you can give format-patch.

Originally, format-patch took either one rev, or two revs:

    format-patch rev1
    format-patch rev1 rev2

The first format was a short-hand for "format-patch rev1 HEAD"
(i.e. rev2==HEAD).  What this meant was to find commits that are
in branch rev2 that has not been merged to branch rev1.

The above notation is still supported, but now it takes sequence
of "from1..to1 from2..to2 ...".  In short, the second format has
become a short-hand for "format-patch rev1..rev2".  Commits in
to1 but not in from1, to2 but not in from2, ... are formatted as
emailable patches.

With this, cherry-picking from other branch can be written as:

    git-format-patch -k --stdout master..branch1 master..branch2 |
    git-am -k -3

which is generally faster than traditional cherry-pick (which
always did 3-way merge) if patches apply cleanly, and still
falls back on 3-way merge if some of them do not.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd "-l" flag for repacking only local packs
Linus Torvalds [Thu, 13 Oct 2005 22:38:29 +0000 (15:38 -0700)]
Add "-l" flag for repacking only local packs

This uses the new "--local" flag to git-pack-objects.  It currently only
makes a difference together with "-a", since a normal incremental repack
won't pack any packed objects at all (whether local or remote).

Eventually, it might end up skipping any objects that aren't local to
the current object directory, but for now it only knows to skip packed
objects.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd support for "local" packing
Linus Torvalds [Thu, 13 Oct 2005 22:38:28 +0000 (15:38 -0700)]
Add support for "local" packing

This adds the "--local" flag to git-pack-objects, which acts like
"--incremental", except that instead of ignoring all packed objects, it
only ignores objects that are packed and in an alternate object tree.

As a result, it effectively only does a local re-pack: any remote-packed
objects will stay in the alternate object directories.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoKeep track of whether a pack is local or not
Linus Torvalds [Thu, 13 Oct 2005 22:38:28 +0000 (15:38 -0700)]
Keep track of whether a pack is local or not

If we want to re-pack just local packfiles, we need to know whether a
particular object is local or not.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agotutorial: update the initial commit example.
Junio C Hamano [Thu, 13 Oct 2005 18:57:05 +0000 (11:57 -0700)]
tutorial: update the initial commit example.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-am: do not fail if 3-way fallback succeeds.
Junio C Hamano [Thu, 13 Oct 2005 18:46:43 +0000 (11:46 -0700)]
git-am: do not fail if 3-way fallback succeeds.

The current one incorrectly stops there without committing.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdapt tutorial to cygwin and add test case
Johannes Schindelin [Thu, 13 Oct 2005 18:36:37 +0000 (11:36 -0700)]
Adapt tutorial to cygwin and add test case

Lacking reliable symlinks, the instructions in the tutorial did not work
in a cygwin setup. Also, a few outputs were not correct.

This patch fixes these, and adds a test case which follows the
instructions of the tutorial (except git-clone, -fetch and -push, which I
have not done yet).

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd findtags - reworked
Martin Langhoff [Thu, 13 Oct 2005 17:47:42 +0000 (10:47 -0700)]
Add findtags - reworked

A short perl script that will walk the tag refs, tag objects, and even commit
objects in its quest to figure out whether the given SHA1 (for a commit or
tree) was ever tagged.

This version is reworked incorporating sanity, feature and style fixes from
Junio.

Usage: git-findtags.perl [ -t ] <commit-or-tree-sha1>

Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRevert "Also use 'track_object_refs = 0' in update-server-info."
Sergey Vlasov [Thu, 13 Oct 2005 17:01:57 +0000 (10:01 -0700)]
Revert "Also use 'track_object_refs = 0' in update-server-info."

This reverts d119e3de13ea1493107bd57381d0ce9c9dd90976 commit.

Object references are used in server-info.c:find_pack_info_one() to
find out which objects in the pack are heads, therefore tracking of
references cannot be disabled.

Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agorsh.c: typo fix
Ralf Baechle [Thu, 13 Oct 2005 17:01:38 +0000 (10:01 -0700)]
rsh.c: typo fix

Example in a comment used a wrong environment variable.

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoclone-pack: new option --keep tells it not to explode the pack.
Junio C Hamano [Thu, 13 Oct 2005 01:12:27 +0000 (18:12 -0700)]
clone-pack: new option --keep tells it not to explode the pack.

With new option --keep, or a configuration item clone.keeppack (we
need a better name, or start allowing dash,"clone.keep-pack"), the packed
data downloaded while cloning is saved as a pack in .git/objects/pack/
locally, with index generated for it with git-index-pack.

Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoclone-pack: new option --keep tells it not to explode the pack.
Junio C Hamano [Thu, 13 Oct 2005 01:12:27 +0000 (18:12 -0700)]
clone-pack: new option --keep tells it not to explode the pack.

With new option --keep, or a configuration item clone.keeppack (we
need a better name, or start allowing dash,"clone.keep-pack"), the packed
data downloaded while cloning is saved as a pack in .git/objects/pack/
locally, with index generated for it with git-index-pack.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix packname hash generation.
Junio C Hamano [Wed, 12 Oct 2005 23:54:19 +0000 (16:54 -0700)]
Fix packname hash generation.

This changes the generation of hash packfiles have in their names, from
"hash of object names as fed to us" to "hash of object names in the
resulting pack, in the order they appear in the index file".  The new
"git-index-pack" command is taught to output the computed hash value
to its standard output.

With this, we can store downloaded pack in a temporary file without
knowing its final name, run git-index-pack to generate idx for it
while finding out its final name, and then rename the pack and idx to
their final names.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd git-index-pack utility
Sergey Vlasov [Wed, 12 Oct 2005 19:01:31 +0000 (12:01 -0700)]
Add git-index-pack utility

git-index-pack builds a pack index file for an existing packed
archive.  With this utility a packed archive which was transferred
without the corresponding pack index can be added to objects/pack/
without repacking.

Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-am: interactive should fail gracefully.
Junio C Hamano [Thu, 13 Oct 2005 01:31:41 +0000 (18:31 -0700)]
git-am: interactive should fail gracefully.

When feeding patches from standard input, and --interactive is specified,
quit, so that the user can re-run the command, instead of infinitely
looping.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-fetch --tags: deal with tags with spaces in them.
Junio C Hamano [Tue, 11 Oct 2005 06:22:02 +0000 (23:22 -0700)]
git-fetch --tags: deal with tags with spaces in them.

"git-fetch --tags" can get confused with tags with spaces in their names,
it used to use shell IFS to split the list of tags and also used curl
which insists the URL to be escaped.  Fix it so it can work with Martin's
moodle repository http://locke.catalyst.net.nz/git/moodle.git/.

We still reserve characters like leading plus-sign '+' and colon
':' anywhere to represent refspec src-dst pair, and obviously we
cannot use LF (that terminates Pull: line in .git/remotes
files), but now you can have spaces with this patch.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoQuote reference names while fetching with curl.
Junio C Hamano [Tue, 11 Oct 2005 06:22:02 +0000 (23:22 -0700)]
Quote reference names while fetching with curl.

curl_escape ought to do this, but we should not let it quote
slashes (nobody said refs/tags cannot have subdirectories), so
we roll our own safer version.  With this, the last part of
git-clone from Martin's moodle repository that used to fail now
works, which reads:

 $ git-http-fetch -v -a -w 'tags/MOODLE_15_MERGED **INVALID**' \
   'tags/MOODLE_15_MERGED **INVALID**' \
   http://locke.catalyst.net.nz/git/moodle.git/

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fixes'
Junio C Hamano [Wed, 12 Oct 2005 05:04:11 +0000 (22:04 -0700)]
Merge branch 'fixes'

15 years ago[PATCH] cvsimport: don't pass --cvs-direct if user options contradict us
Martin Langhoff [Wed, 12 Oct 2005 04:57:04 +0000 (21:57 -0700)]
[PATCH] cvsimport: don't pass --cvs-direct if user options contradict us

Detecting if the user passed --no-cvs-direct and don't force the mode.
It allows us to support all the protocol that the standard cvs client
supports at the snail speed you should expect.

This only affects the rlog reading stage.

Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
15 years agoMake git config variable names case-insensitive
Linus Torvalds [Wed, 12 Oct 2005 01:47:34 +0000 (18:47 -0700)]
Make git config variable names case-insensitive

They always were meant to be case-insensitive, but I had missed one
"tolower()", making that not true.

The actual _values_ aren't case-insensitive, of course, although some uses
of them may be (ie boolean parsing uses "strcasecmp()" to match against
the strings "true" and "false").

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse git config file for committer name and email info
Linus Torvalds [Wed, 12 Oct 2005 01:47:34 +0000 (18:47 -0700)]
Use git config file for committer name and email info

This starts using the "user.name" and "user.email" config variables if
they exist as the default name and email when committing.  This means
that you don't have to use the GIT_COMMITTER_EMAIL environment variable
to override your email - you can just edit the config file instead.

The patch looks bigger than it is because it makes the default name and
email information non-static and renames it appropriately.  And it moves
the common git environment variables into a new library file, so that
you can link against libgit.a and get the git environment without having
to link in zlib and libcrypt.

In short, most of it is renaming and moving, the real change core is
just a few new lines in "git_default_config()" that copies the user
config values to the new base.

It also changes "git-var -l" to list the config variables.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoExport template_dir in Makefile.
Tom Prince [Wed, 12 Oct 2005 01:47:33 +0000 (18:47 -0700)]
Export template_dir in Makefile.

If somebody set template_dir in config.mak. Then git-init-db would be
compiled with the correct location but the templates would be installed
in the default location. Fix it.

Signed-off-by: Tom Prince <tom.prince@ualberta.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse core.filemode.
Junio C Hamano [Wed, 12 Oct 2005 01:45:33 +0000 (18:45 -0700)]
Use core.filemode.

With "[core] filemode = false", you can tell git to ignore
differences in the working tree file only in executable bit.

 * "git-update-index --refresh" does not say "needs update" if index
   entry and working tree file differs only in executable bit.

 * "git-update-index" on an existing path takes executable bit
   from the existing index entry, if the path and index entry are
   both regular files.

 * "git-diff-files" and "git-diff-index" without --cached flag
   pretend the path on the filesystem has the same executable
   bit as the existing index entry, if the path and index entry
   are both regular files.

If you are on a filesystem with unreliable mode bits, you may need to
force the executable bit after registering the path in the index.

 * "git-update-index --chmod=+x foo" flips the executable bit of the
   index file entry for path "foo" on.  Use "--chmod=-x" to flip it
   off.

Note that --chmod only works in index file and does not look at nor
update the working tree.

So if you are on a filesystem and do not have working executable bit,
you would do:

 1. set the appropriate .git/config option;

 2. "git-update-index --add new-file.c"

 3. "git-ls-files --stage new-file.c" to see if it has the desired
   mode bits.  If not, e.g. to drop executable bit picked up from the
   filesystem, say "git-update-index --chmod=-x new-file.c".

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoImprove config file escape sanity checking
Linus Torvalds [Tue, 11 Oct 2005 22:24:11 +0000 (15:24 -0700)]
Improve config file escape sanity checking

I had meant to disallow unknown escape characters in the config file
parser, but instead an unknown escaped character would silently pass
through as itself. That's correct for some cases (notably '\' itself), but
wasn't correct in general.

This fixes it, and makes the parser write a nice error message if the
config file contains bogus escaped characters.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoshow-branch: optionally use unique prefix as name.
Junio C Hamano [Tue, 11 Oct 2005 22:22:48 +0000 (15:22 -0700)]
show-branch: optionally use unique prefix as name.

git-show-branch acquires two new options. --sha1-name to name
commits using the unique prefix of their object names, and
--no-name to not to show names at all.

This was outlined in <7vk6gpyuyr.fsf@assigned-by-dhcp.cox.net>

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoSupport custom build options in config.mak
Johannes Schindelin [Tue, 11 Oct 2005 22:22:47 +0000 (15:22 -0700)]
Support custom build options in config.mak

With this patch, it is possible to store configuration options like
NO_CURL=YesPlease or NO_OPENSSL=YesPlease into a file named
config.mak, which will be included in the Makefile.

[jc: redone with suggestion from Daniel Barkalow to just use -include]

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse git-update-ref and git-symbolic-ref in tests
Johannes Schindelin [Tue, 11 Oct 2005 22:15:15 +0000 (15:15 -0700)]
Use git-update-ref and git-symbolic-ref in tests

This makes all tests pass on cygwin.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agosvn import: copy directories
Matthias Urlichs [Tue, 11 Oct 2005 17:42:27 +0000 (19:42 +0200)]
svn import: copy directories

Import SVN-copied and -referenced directories correctly.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: Add direct HTTP access
Matthias Urlichs [Tue, 11 Oct 2005 16:13:30 +0000 (18:13 +0200)]
svn import: Add direct HTTP access

Some SVN repositories that are accessible through HTTP don't like when I
retrieve files using SVN methods ("internal server error").

Therefore, I added an option to get the contents using (persistent) HTTP
directly. This also reduces round-trip time, from two or three requests
down to one.

Also corrected error handling a bit.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: remove some CVS cruft
Matthias Urlichs [Tue, 11 Oct 2005 15:02:45 +0000 (17:02 +0200)]
svn import: remove some CVS cruft

Some remains of CVS still lingered. Removed.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: make -s option actually optional
Matthias Urlichs [Tue, 11 Oct 2005 14:22:03 +0000 (16:22 +0200)]
svn import: make -s option actually optional

The -s option was accidentally not optional.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: typo fix
Matthias Urlichs [Tue, 11 Oct 2005 12:18:01 +0000 (14:18 +0200)]
svn import: typo fix

Fixed a minor typo

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agoRemove unused 'got_alternates' variable.
Junio C Hamano [Tue, 11 Oct 2005 06:22:02 +0000 (23:22 -0700)]
Remove unused 'got_alternates' variable.

The function fetch_alternates() is called only once from the
main now.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoUse the same move_temp_to_file in git-http-fetch.
Junio C Hamano [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
Use the same move_temp_to_file in git-http-fetch.

The http commit walker cannot use the same temporary file
creation code because it needs to use predictable temporary
filename for partial fetch continuation purposes, but the code
to move the temporary file to the final location should be
usable from the ordinary object creation codepath.

Export move_temp_to_file from sha1_file.c and use it, while
losing the custom relink_or_rename function from http-fetch.c.

Also the temporary object file creation part needs to make sure
the leading path exists, in preparation of the really lazy
fan-out directory creation.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRestore functionality to allow proxies to cache objects
Nick Hengeveld [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
Restore functionality to allow proxies to cache objects

The parallel request changes didn't properly implement the previous patch to
allow caching of retrieved objects by proxy servers.  Restore the previous
functionality such that by default requests include the "Pragma: no-cache"
header, and this header is removed on requests for pack indexes, packs, and
objects.

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] Don't fetch objects that exist in the local repository
Nick Hengeveld [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
[PATCH] Don't fetch objects that exist in the local repository

Be sure not to fetch objects that already exist in the local repository.
The main process loop no longer performs this check, http-fetch now checks
prior to starting a new request queue entry and when fetch_object() is called,
and local-fetch now checks when fetch_object() is called.

As discussed in this thread: http://marc.theaimsgroup.com/?t=112854890500001

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
15 years agoSet the parallel HTTP request limit via an environment variable
Nick Hengeveld [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
Set the parallel HTTP request limit via an environment variable

Use an environment variable rather than a command-line argument to set the
parallel HTTP request limit.  This allows the setting to work whether
git-http-fetch is run directly or via git-fetch.

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoOnly compile parallel HTTP support with CURL >= 7.9.8
Nick Hengeveld [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
Only compile parallel HTTP support with CURL >= 7.9.8

Only compile parallel HTTP support with CURL >= 7.9.8

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd support for parallel HTTP transfers
Nick Hengeveld [Tue, 11 Oct 2005 06:22:01 +0000 (23:22 -0700)]
Add support for parallel HTTP transfers

Add support for parallel HTTP transfers.  Prefetch populates a queue of
objects to transfer and starts feeding requests to an active request
queue for processing; fetch_object keeps the active queue moving
while the specified object is being transferred.  The size of the active
queue can be restricted using -r and defaults to 5 concurrent transfers.
Requests for objects that are not prefetched are also processed via the
active queue.

Signed-off-by: Nick Hengeveld <nickh@reactrix.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRemove empty directories after read-tree -u.
Junio C Hamano [Tue, 11 Oct 2005 00:34:08 +0000 (17:34 -0700)]
Remove empty directories after read-tree -u.

This fixes everybody's favorite gripe that switching branche with
'git checkout' leaves empty directories.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoAdd ".git/config" file parser
Linus Torvalds [Mon, 10 Oct 2005 23:31:08 +0000 (16:31 -0700)]
Add ".git/config" file parser

This is a first cut at a very simple parser for a git config file.

The format of the file is a simple ini-file like thing, with simple
variable/value pairs. You can (and should) make the variables have a
simple single-level scope, ie a valid file looks something like this:

#
# This is the config file, and
# a '#' or ';' character indicates
# a comment
#

; core variables
[core]
; Don't trust file modes
filemode = false

; Our diff algorithm
[diff]
external = "/usr/local/bin/gnu-diff -u"
renames = true

which parses into three variables: "core.filemode" is associated with the
string "false", and "diff.external" gets the appropriate quoted value.

Right now we only react to one variable: "core.filemode" is a boolean that
decides if we should care about the 0100 (user-execute) bit of the stat
information. Even that is just a parsing demonstration - this doesn't
actually implement that st_mode compare logic itself.

Different programs can react to different config options, although they
should always fall back to calling "git_default_config()" on any config
option name that they don't recognize.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fixes'
Junio C Hamano [Mon, 10 Oct 2005 23:03:43 +0000 (16:03 -0700)]
Merge branch 'fixes'

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoThe synopsis of the manpages should use the hyphenated version
Christian Meder [Mon, 10 Oct 2005 23:01:32 +0000 (16:01 -0700)]
The synopsis of the manpages should use the hyphenated version

The synopsis of the manpages should use the hyphenated version of the git
commands. Adapt the remaining offenders.

Signed-off-by: Christian Meder <chris@absolutegiganten.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoConvert usage of GIT and Git into git
Christian Meder [Mon, 10 Oct 2005 23:01:31 +0000 (16:01 -0700)]
Convert usage of GIT and Git into git

Convert usage of GIT and Git into git.

Signed-off-by: Christian Meder <chris@absolutegiganten.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoRemove the version tags from the manpages
Junio C Hamano [Mon, 10 Oct 2005 21:49:52 +0000 (14:49 -0700)]
Remove the version tags from the manpages

Signed-off-by: Christian Meder <chris@absolutegiganten.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMake rsh.c use sq_quote_buf()
H. Peter Anvin [Mon, 10 Oct 2005 21:46:14 +0000 (14:46 -0700)]
Make rsh.c use sq_quote_buf()

Make rsh.c use sq_quote_buf()

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoTrivial optimization
H. Peter Anvin [Mon, 10 Oct 2005 21:46:12 +0000 (14:46 -0700)]
Trivial optimization

GIT_DIR_ENVIRONMENT is always a string literal

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoEnhanced sq_quote()
H. Peter Anvin [Mon, 10 Oct 2005 21:46:10 +0000 (14:46 -0700)]
Enhanced sq_quote()

Create function to sq_quote into a buffer
Handle !'s for csh-based shells

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agot5400-send-pack relies on a working cpio
Johannes Schindelin [Mon, 10 Oct 2005 21:14:22 +0000 (14:14 -0700)]
t5400-send-pack relies on a working cpio

Since cygwin does not install cpio by default, t5400 results in a very
cryptic failure. So, test for cpio explicitely.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit.sh: quote all paths
Jonas Fonseca [Mon, 10 Oct 2005 21:14:06 +0000 (14:14 -0700)]
git.sh: quote all paths

This makes it handle spaces in paths.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoTeach git-status about spaces in file names also on MacOSX
Johannes Schindelin [Mon, 10 Oct 2005 20:58:52 +0000 (13:58 -0700)]
Teach git-status about spaces in file names also on MacOSX

Not all sed understands '\t' and consequently cuts off every
file name at the first "t" (or backslash...).

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoDeal with $(bindir) and friends with whitespaces.
Junio C Hamano [Mon, 10 Oct 2005 20:50:01 +0000 (13:50 -0700)]
Deal with $(bindir) and friends with whitespaces.

... using HPA's shellquote macro.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoOpenBSD needs the strcasestr replacement.
Junio C Hamano [Mon, 10 Oct 2005 18:51:11 +0000 (11:51 -0700)]
OpenBSD needs the strcasestr replacement.

Noticed by Randal L. Schwartz.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agosvn import: get all revisions
Matthias Urlichs [Mon, 10 Oct 2005 18:10:48 +0000 (20:10 +0200)]
svn import: get all revisions

Not skipping the last revision is generally seen as Good Thing. ;-)

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: unlink downlaoded files
Matthias Urlichs [Mon, 10 Oct 2005 16:54:53 +0000 (18:54 +0200)]
svn import: unlink downlaoded files

Actually removing the files that have been checked out of SVN,
after checking them into git of course, is a good idea...

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: Add a loop limit option
Matthias Urlichs [Mon, 10 Oct 2005 16:45:00 +0000 (18:45 +0200)]
svn import: Add a loop limit option

The svn library has a serious memory leak.
Added a new option (-l NUM) which causes git-svnimport to exit cleanly
after fetching that many changes, in order to \1a.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: incremental imports
Matthias Urlichs [Mon, 10 Oct 2005 13:28:00 +0000 (15:28 +0200)]
svn import: incremental imports

Incremental imports skipped a revision.

Also improve interrupt safety -- ^C while writing a tag caused the tag
to be skipped.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: skip initial revisions
Matthias Urlichs [Mon, 10 Oct 2005 13:14:21 +0000 (15:14 +0200)]
svn import: skip initial revisions

Add a flag to skip initial revisions: some SVN repositories have
initial setup cruft in their logs which we might want to ignore.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: Do not create empty tags
Matthias Urlichs [Mon, 10 Oct 2005 12:51:13 +0000 (14:51 +0200)]
svn import: Do not create empty tags

If a tag is "clean", do not create a commit for it.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: Fix tagging.
Matthias Urlichs [Mon, 10 Oct 2005 12:42:59 +0000 (14:42 +0200)]
svn import: Fix tagging.

Tagging was 100% broken. :-/

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: remove debugging
Matthias Urlichs [Mon, 10 Oct 2005 12:19:15 +0000 (14:19 +0200)]
svn import: remove debugging

Removed debugging output used to identify the too-many-connections problem.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: wrong file open mode
Matthias Urlichs [Mon, 10 Oct 2005 12:18:38 +0000 (14:18 +0200)]
svn import: wrong file open mode

There are multiple | characters in Unicode. Don't use the wrong one ...

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: avoid reconnecting
Matthias Urlichs [Mon, 10 Oct 2005 12:14:44 +0000 (14:14 +0200)]
svn import: avoid reconnecting

Perl's eval() sets $@ to empts, not undef, when it succeeds.
That caused excessive reconnect attempts.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn improt needs SVN::Core 1.2.1 or better
Matthias Urlichs [Mon, 10 Oct 2005 11:42:48 +0000 (13:42 +0200)]
svn improt needs SVN::Core 1.2.1 or better

Die with a warning if Perl's svn module is too old.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: add libsvn-core-perl to Debian's control file
Matthias Urlichs [Mon, 10 Oct 2005 11:10:01 +0000 (13:10 +0200)]
svn import: add libsvn-core-perl to Debian's control file

Added libsvn-core-perl to debian/control, "Recommends:" section.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: fixed two pipe open calls.
Matthias Urlichs [Mon, 10 Oct 2005 10:45:46 +0000 (12:45 +0200)]
svn import: fixed two pipe open calls.

Perl's magic "different semantics of open() based on the number of
arguments" is really annoying at times...

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: copy path information
Matthias Urlichs [Mon, 10 Oct 2005 10:41:15 +0000 (12:41 +0200)]
svn import: copy path information

Due to a bug in the SVN library, path information is freed as soon as the
callback returns, even if it still refers to the data.

Workaround: Copy it. (Also fix a wrong-method-name bug while we're at it.)

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agoSVN import: No modes
Matthias Urlichs [Mon, 10 Oct 2005 10:34:32 +0000 (12:34 +0200)]
SVN import: No modes

svn doesn't seem to save file modes:
removed the code that analyzes them.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agosvn import: add eval()
Matthias Urlichs [Mon, 10 Oct 2005 10:33:22 +0000 (12:33 +0200)]
svn import: add eval()

Trying to downlaod a file that's really a subdirectory doesn't work too well.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agoNew: git-svnimport.
Matthias Urlichs [Mon, 10 Oct 2005 09:40:43 +0000 (11:40 +0200)]
New: git-svnimport.

As the name suggests, this script imports from SVN.

Only "normal" SVN repositories (with single trunk/, branches/, and tags/
subdrectories) are supported. Incremental imports require preserving
the file .git/svn2git.

Signed-Off-by: Matthias Urlichs <smurf@smurf.noris.de>
15 years agocvsimport: report merge parents
Matthias Urlichs [Mon, 10 Oct 2005 09:15:09 +0000 (11:15 +0200)]
cvsimport: report merge parents

Matching and reporting merge parents happens in a subprocess.
Re-open stdout before redirecting stdout to the pipe, so that printing
verbose messages doesn't go to the wrong place.

Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
15 years agoAlso force LC_ALL in test scripts.
Junio C Hamano [Mon, 10 Oct 2005 04:58:02 +0000 (21:58 -0700)]
Also force LC_ALL in test scripts.

Noticed by Junichi Uekawa.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMerge branch 'fixes'
Junio C Hamano [Mon, 10 Oct 2005 02:16:33 +0000 (19:16 -0700)]
Merge branch 'fixes'

with minor hand resolving on git-tag.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agogit-tag: update usage string and documentation.
Junio C Hamano [Mon, 10 Oct 2005 02:13:47 +0000 (19:13 -0700)]
git-tag: update usage string and documentation.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoWrite .editmsg in GIT_DIR to avoid being in git-status.
Santi_Béjar [Mon, 10 Oct 2005 00:30:19 +0000 (17:30 -0700)]
Write .editmsg in GIT_DIR to avoid being in git-status.

It appears in the git-status output during a git-commit if you have
something in info/exclude.

Also for .cmitmsg and .cmitchk to make git-commit work
in read-only working trees.

[jc: while we are at it, I removed the use of .cmitchk temporary
     file which was not necessary, and renamed them -- they are out
     of way now and do not have to be dotfiles anymore.]

Signed-off-by: Santi Béjar <sbejar@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoignore new git-diff index header when computing patch ids
Kai Ruemmler [Sun, 9 Oct 2005 23:52:50 +0000 (16:52 -0700)]
ignore new git-diff index header when computing patch ids

Two else equal patches should not result in different checksums, only
because they were applied to different versions of the file.

Signed-off-by: Kai Ruemmler <kai.ruemmler@gmx.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoFix git-verify-tag for light-weight tags
Paolo 'Blaisorblade' Giarrusso [Sun, 9 Oct 2005 23:50:34 +0000 (16:50 -0700)]
Fix git-verify-tag for light-weight tags

It currently exits printing "git-cat-file SHA1: bad file", while
instead we must just abort the verification for light-weight
tags (e.g. referring to commit objects).

[jc: tag objects can tag anything not just commits, so I fixed
 up the original patch slightly.  you should be able to validate
 a signed tag that points at a blob object. ]

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMakefile: avoid error message from 'uname -o'
Junio C Hamano [Sun, 9 Oct 2005 19:52:35 +0000 (12:52 -0700)]
Makefile: avoid error message from 'uname -o'

The platform specific tweaking part was using 'uname -o' which
is not always available.  Squelch error message from it.

It was suggested to chain the if..else, but I chose not to, because
maintaining the nested if..else if..else..endif endif to match is a
pain.  If we had "elif", things would have been different, though.
While we are at it, try not to invoke 'uname -s' for each platform
candidate.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoMake sure 'make install' does not have to rebuild templates.
Junio C Hamano [Sun, 9 Oct 2005 01:01:24 +0000 (18:01 -0700)]
Make sure 'make install' does not have to rebuild templates.

The dependency rule in templates directory forced 'make install'
that immediately followed 'make all' to rebuild boilerplates.
This was problematic for a workflow that built first as yourself
and then installed as root, from a working tree that is on an
NFS mounted filesystem that is unwritable by root.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agomake $prefix available for sub-makefiles
Kai Ruemmler [Sat, 8 Oct 2005 22:54:37 +0000 (15:54 -0700)]
make $prefix available for sub-makefiles

exports $prefix and makes Documentation/Makefile following it also.

Signed-off-by: Kai Ruemmler <kai.ruemmler@gmx.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years agoYank writing-back support from gitfakemmap.
Junio C Hamano [Sat, 8 Oct 2005 22:54:36 +0000 (15:54 -0700)]
Yank writing-back support from gitfakemmap.

We do not write through our use of mmap(), so make sure callers pass
MAP_PRIVATE and remove support for writing changes back.

Signed-off-by: Junio C Hamano <junkio@cox.net>
15 years ago[PATCH] If NO_MMAP is defined, fake mmap() and munmap()
Johannes Schindelin [Sat, 8 Oct 2005 22:54:36 +0000 (15:54 -0700)]
[PATCH] If NO_MMAP is defined, fake mmap() and munmap()

Since some platforms do not support mmap() at all, and others do only just
so, this patch introduces the option to fake mmap() and munmap() by
malloc()ing and read()ing explicitely.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
15 years agoAlso use 'track_object_refs = 0' in update-server-info.
Junio C Hamano [Sat, 8 Oct 2005 22:54:35 +0000 (15:54 -0700)]
Also use 'track_object_refs = 0' in update-server-info.

Signed-off-by: Junio C Hamano <junkio@cox.net>