[ANNOUNCE] GIT 1.5.0-rc3

[ANNOUNCE] GIT 1.5.0-rc3

[Junio C Hamano]

It's been a week and a half, and here comes the -rc3

  http://www.kernel.org/pub/software/scm/git/

  git-1.5.0-rc3.tar.{gz,bz2}			(tarball)
  git-htmldocs-1.5.0-rc3.tar.{gz,bz2}		(preformatted docs)
  git-manpages-1.5.0-rc3.tar.{gz,bz2}		(preformatted docs)
  testing/git-*-1.5.0-rc3-1.$arch.rpm		(RPM)

Hopefully this will be the last -rc before the final 1.5.0.

This contains the "git user's manual" by J. Bruce Fields.  While
it still has its own "To Do" section, it has already been in a
very good shape for some time, giving information with coherence
and guiding a new user in a carefully chosen order of topics
into much more pleasant git experience than before.

----------------------------------------------------------------

Changes since v1.5.0-rc2 are as follows:

Alex Riesen (5):
  Insert ACTIVESTATE_STRING in Git.pm
  Force Activestate Perl to tie git command pipe handle to a handle class
  Cleanup uninitialized value in chomp
  Allow default core.logallrefupdates to be overridden with template's config
  Fix git-update-index to work with relative pathnames.

Alexandre Julliard (1):
  vc-git.el: Take into account the destination name in vc-checkout.

Andy Parkins (5):
  New files in git weren't being downloaded during CVS update
  If abbrev is set to zero in git-describe, don't add the unique suffix
  Allow the tag signing key to be specified in the config file
  UNIX reference time of 1970-01-01 00:00 is UTC timezone, not local time zone
  Heavily expanded update hook to send more useful emails than the old hook

Aneesh Kumar K.V (2):
  blameview: Use git-cat-file to read the file content.
  Update git-cat-file documentation

Bill Lear (1):
  Document --check option to git diff.

David Kågedal (1):
  Improved error message from git-rebase

Eric Wong (1):
  git-svn: remove leading slash when printing removed directories

J. Bruce Fields (31):
  Documentation: add git user's manual
  Documentation: expand preface and todo's
  Documentation: git-rebase discussion, miscellaneous user-manual updates
  Documentation: more user-manual todo's
  Documentation: reorder development section, todo's
  Documentation: begin discussion of git-remote in user manual
  Documentation: rev-list -> rev-parse, other typos, start examples
  user manual: answer some comments from Junio
  user-manual: replace init-db by init
  user-manual: reindent
  user-manual:  rewrap, fix heading levels
  user-manual: add "quick start" as chapter 1
  user-manual: minor quickstart reorganization
  user-manual: clarify difference between tag and branch
  user-manual: update references discussion
  user-manual: update git-gc discussion
  user-manual: stub discussion of fsck and reflog
  user-manual: reorganize fetch discussion, add internals, etc.
  user-manual: git-fsck, dangling objects
  user-manual: fsck-objects -> fsck
  user-manual: repo-config -> config
  user-manual: add references to git-config man page
  user-manual: typo fix
  user-manual: fix a header level
  user-manual: reflogs, other recovery
  user-manual: rewrap a few long lines
  user-manual: minor "TODO" updates
  user-manual: document git-show-branch example
  user-manual: SHA1 -> object name
  user-manual: point to README for gitweb information
  user-manual: todo's

Jakub Narebski (3):
  Documentation/config.txt: Document config file syntax better
  t/t1300-repo-config.sh: value continued on next line
  Documentation/config.txt: Correct info about subsection name

Jason Riedy (1):
  Use inttypes.h rather than stdint.h.

Jeff King (4):
  format-patch: fix bug with --stdout in a subdirectory
  contrib/vim: update syntax for changed commit template
  diffcore-pickaxe: fix infinite loop on zero-length needle
  Add a sample program 'blameview' to show how to use git-blame --incremental

Johannes Schindelin (2):
  annotate: use pager
  reflog inspection: introduce shortcut "-g"

Johannes Sixt (1):
  Add a missing fork() error check.

Junio C Hamano (43):
  User manual: fix typos in examples
  Documentation/tutorial-2: Fix interesting typo in an example.
  Revert "prune: --grace=time"
  Make sure git_connect() always give two file descriptors.
  is_repository_shallow(): prototype fix.
  shallow repository: disable unsupported operations for now.
  git-gc: do not run git-prune by default.
  cvsimport: activate -a option, really.
  .mailmap: fix screw-ups in Uwe's name
  honor --author even with --amend, -C, and -c.
  reflog gc: a tag that does not point at a commit is not a crime.
  git-checkout -m: fix merge case
  git-daemon documentation on enabling services.
  ls-remote and clone: accept --upload-pack=<path> as well.
  Refactor the pack header reading function out of receive-pack.c
  Allow fetch-pack to decide keeping the fetched pack without exploding
  fetch-pack: remove --keep-auto and make it the default.
  Consolidate {receive,fetch}.unpackLimit
  Allow non-developer to clone, checkout and fetch more easily.
  parse-remote: do not barf on a remote shorthand without any refs to fetch.
  show-branch -g: default to HEAD
  Documentation: pack-refs --all vs default behaviour
  Make sure we do not write bogus reflog entries.
  git-merge: leave sensible reflog message when used as the first level UI.
  create_symref: check error return from open().
  write_in_full: size_t is unsigned.
  Don't force everybody to call setup_ident().
  git-blame --porcelain: quote filename in c-style when needed.
  Update describe documentation.
  Document 'git-blame --incremental'
  git-fsck-objects is now synonym to git-fsck
  Two small typofixes.
  lock_any_ref_for_update(): do not accept malformatted refs.
  git-commit -s: no extra space when sign-offs appear at the end already.
  git-blame: somewhat better commenting.
  git-send-email: remove debugging output.
  Fix git-tag -u
  Documentation: "git-checkout <tree> <path>" takes any tree-ish
  t9200: Re-code non-ascii path test in UTF-8
  t9200: do not test -x bit if the filesystem does not support it.
  git main documentation: point at the user's manual.
  Do not use hardcoded path to xhmtl.xsl to generate user's manual
  GIT v1.5.0-rc3

Linus Torvalds (5):
  fsck-objects: refactor checking for connectivity
  Fix seriously broken "git pack-refs"
  Add dangling objects tips.
  git-blame --incremental
  git-push through git protocol

Mark Wooding (3):
  wt-status: Actually accept `color.status.BLAH' configuration variables.
  Documentation/config.txt: Fix documentation of colour config tweaks.
  Make fsck and fsck-objects be builtins.

Nicolas Pitre (3):
  fix suggested branch creation command when detaching head
  git-log -g --pretty=oneline should display the reflog message
  tone down the detached head warning

Peter Eriksen (2):
  sha1_file.c: Avoid multiple calls to find_pack_entry().
  Documentation: --amend cannot be combined with -c/-C/-F.

René Scharfe (1):
  git-blame --incremental: don't use pager

Sam Vilain (1):
  contrib/emacs/vc-git.el: support vc-version-other-window

Santi Béjar (1):
  git-fetch: Allow fetching the remote HEAD

Shawn O. Pearce (6):
  Remove unnecessary found variable from describe.
  Teach git-describe to display distances from tags.
  Compute accurate distances in git-describe before output.
  Teach for-each-ref about a little language called Tcl.
  Don't coredump on bad refs in update-server-info.
  Escape --upload-pack from expr.

Simon 'corecode' Schubert (1):
  Replace perl code with pure shell code

Tom Prince (1):
  Rename git-repo-config to git-config.

Uwe Kleine-König (2):
  rename --exec to --upload-pack for fetch-pack and peek-remote
  make --upload-pack option to git-fetch configurable

Re: [ANNOUNCE] GIT 1.5.0-rc3

[Bill Lear]

On Wednesday, January 31, 2007 at 16:26:36 (-0800) Junio C Hamano writes:
>It's been a week and a half, and here comes the -rc3
>
>  http://www.kernel.org/pub/software/scm/git/
>
>  git-1.5.0-rc3.tar.{gz,bz2}			(tarball)
>  git-htmldocs-1.5.0-rc3.tar.{gz,bz2}		(preformatted docs)
>  git-manpages-1.5.0-rc3.tar.{gz,bz2}		(preformatted docs)
>  testing/git-*-1.5.0-rc3-1.$arch.rpm		(RPM)

There are no such files .*1.5.0-rc3.* on this server.  However, there
are .*1.5.0.rc3.* files.


Bill

Re: [ANNOUNCE] GIT 1.5.0-rc3

[Robin Rosenberg]

torsdag 01 februari 2007 01:26 skrev Junio C Hamano:
> Junio C Hamano (43):
>   t9200: Re-code non-ascii path test in UTF-8

I sent a patch that converts to "visible" characters when looked
at with iso-8859-1/windows-1252 googles. It also tests the file system before 
executing the test. 

>   t9200: do not test -x bit if the filesystem does not support it.

The description is misleading. Git on cygwin chooses to ignore the executable 
flag by default, not because the filesystem won't handle it, but rather that 
enough many *other* applications sets it needlessly, that ignoring it becomes 
less problematic than honouring it.

Adding "git config core.filemode true" at the start of this test verifiies 
that it works on NTFS too. 

-- robin

What's in git.git (stable)

[Junio C Hamano]

Thanks everybody for helping to tie the loose ends for the next
release.

Thanks to Nico's handful updates, I think his "reflog on HEAD
itself" series is ready for inclusion, although tonight's update
still has it parked in 'next' (I did try out 'next' with his
latest for a couple of hours today).  I've already updated the
draft release notes for 1.5.0 (v1.5.0.txt in todo branch), with
its inclusion in mind.

I've re-read the three tutorials and tried to fix them up
somewhat to match the recent reality.  I'd appreciate somebody
can look at JBF's user manual.

Several things to note (in the following, substitute $gmane with
http://article.gmane.org/gmane.comp.version-control.git):

 - Working around Tk geometry problem, especially on non Linux+X
   platforms.  I've forwarded Mark Levedahl's patches
   ($gmane/38361) to Paul Mackerras for his blessing; hopefully
   he can Ack and/or suggest improvements.  I'd really like to
   have them in 1.5.0 in some form.

 - Nico's "reflog on HEAD".  I'll merge this tomorrow (I just
   ran out of time today).  I will re-try to break its "git
   reflog expire --stale-fix --all" before actually merging it,
   though.

 - Reverting the patch to allow tracking branch names as the
   value of branch.*.merge ($gmane/38621).  I think it is a good
   idea to revert this before 1.5.0 gets out; just haven't got
   around to do so.

 - Teaching "git blame" to also use the working tree files
   and/or index.  I actually think defaulting to working tree
   when an explicit HEAD is not given (and --cached to use the
   one in the index) makes a lot of sense, but I haven't got
   around to code the latter yet.  Not defaulting to HEAD
   changes semantics, so if we ever are going to do it, I think
   we should do so before 1.5.0.

 - Preventing push from updating the current branch of non-bare
   repository.  I think doing so unconditionally is a bad idea
   (and I have Linus's veto to back it up $gmane/38592), but I
   suspect most people would want the default to be less
   confusing to new people.  If we are ever going to forbid by
   default and allow pusher to force, that would be a behaviour
   change and it would be better to do so before 1.5.0.

 - We might want to allow git-push to use the wildcard refspec,
   like git-fetch does, for symmetry.  It would make the
   mothership-satellite arrangement much more natural
   ($gname/38549).  Unfortunately I haven't had a chance to
   start coding it.  I think however this could be added post
   1.5.0.

 - "git remote add -t -f -m" and rewrite of "git clone" based on
   it ($gmane/38470, $gmane/38545).  While I think this leads us
   in right direction, I do not think 1.5.0 needs to wait for
   it.  Certainly I do not want to reimplement "git clone"
   before 1.5.0, although I think the additions to "git remote add"
   are relatively safe.

 - Catering to filesystems whose readdir() returns pathnames
   that are different from what are used when they were creat()ed
   will not happen ($gmane/38620).


* The 'master' branch has these since v1.5.0-rc3.

 Andy Parkins (1):
  doc: hooks.txt said post-commit default sends an email, it doesn't

 Eric Wong (2):
  git-svn: do not let Git.pm warn if we prematurely close pipes
  Disallow invalid --pretty= abbreviations

 Johannes Schindelin (2):
  Teach the '@{...}' notation to git-log -g
  Update the documentation for the new '@{...}' syntax

 Junio C Hamano (9):
  detached HEAD -- finishing touches
  Use "git checkout -q" in git-bisect
  Tutorial: fix asciidoc formatting of "git add" section.
  Tutorial-2: Adjust git-status output to recent reality.
  core-tutorial: http reference link fix
  fix reflog entries for "git-branch"
  honor GIT_REFLOG_ACTION in git-commit
  Why is it bad to rewind a branch that has already been pushed out?
  combine-diff: special case --unified=0

 Mike Coleman (1):
  Fix some documentation typos and grammar

 Nicolas Pitre (4):
  reword the detached head message a little again
  add a quiet option to git-checkout
  prevent HEAD reflog to be interpreted as current branch reflog
  provide a nice @{...} syntax to always mean the current branch reflog

 Pavel Roskin (2):
  git-config --rename-section could rename wrong section
  Assorted typo fixes

 Shawn O. Pearce (18):
  Pull out remote listing functions in git-remote.
  Teach 'git remote' how to cleanup stale tracking branches.
  Cleanup prepare_packed_git_one to reuse install_packed_git.
  Correct comment in prepare_packed_git_one.
  Refactor open_packed_git to return an error code.
  Don't find objects in packs which aren't available anymore.
  Don't leak file descriptors from unavailable pack files.
  Cleanup subcommand documentation for git-remote.
  Keep untracked files not involved in a merge.
  Default GIT_MERGE_VERBOSITY to 5 during tests.
  bash: Remove short option completions for branch/checkout/diff.
  bash: Classify cat-file and reflog as plumbing.
  bash: Complete long options to git-add.
  bash: Add space after unique command name is completed.
  bash: Classify more commends out of completion.
  bash: Support unique completion on git-config.
  bash: Support unique completion when possible.
  bash: Support internal revlist options better.

Re: What's in git.git (stable)

[Jeff King]

On Sun, Feb 04, 2007 at 01:36:29AM -0800, Junio C Hamano wrote:

>  - Teaching "git blame" to also use the working tree files
>    and/or index.  I actually think defaulting to working tree
>    when an explicit HEAD is not given (and --cached to use the
>    one in the index) makes a lot of sense, but I haven't got
>    around to code the latter yet.  Not defaulting to HEAD
>    changes semantics, so if we ever are going to do it, I think
>    we should do so before 1.5.0.

Just a thought, but it might be useful to blame the contents of an
arbitrary file (but starting the history at a given pathname). Something
like "git blame --contents /tmp/foo.c file.c", with contents defaulting
to "file.c". There's much discussion of editor interfaces, and this
leaves the possibility of git-blaming the contents of the editor buffer
(after writing it out to a temp file) without having to save changes to
the working tree file.

Admittedly, I think this will be rare, but if you git-blame from an
editor, it seems awkward to either be inaccurate (by blaming the last
saved working tree file, but comparing it to the buffer) or to save the
file as a side effect.

-Peff

Re: What's in git.git (stable)

[Linus Torvalds]

On Sun, 4 Feb 2007, Jeff King wrote:
> 
> Just a thought, but it might be useful to blame the contents of an
> arbitrary file (but starting the history at a given pathname). Something
> like "git blame --contents /tmp/foo.c file.c", with contents defaulting
> to "file.c". There's much discussion of editor interfaces, and this
> leaves the possibility of git-blaming the contents of the editor buffer
> (after writing it out to a temp file) without having to save changes to
> the working tree file.

I agree, that probably would make most sense. If we do this at all. On the 
other hand, I suspect that most editors would probably want to pipe the 
contents to the program, not write it to a temp-file.

(I think it's a worthy feature, but Junio's patch wasn't exactly pretty, 
so the question boils down to whether it's really worth it).

		Linus

Re: What's in git.git (stable)

[Theodore Tso]

On Sun, Feb 04, 2007 at 11:12:34AM -0800, Linus Torvalds wrote:
> 
> 
> On Sun, 4 Feb 2007, Jeff King wrote:
> > 
> > Just a thought, but it might be useful to blame the contents of an
> > arbitrary file (but starting the history at a given pathname). Something
> > like "git blame --contents /tmp/foo.c file.c", with contents defaulting
> > to "file.c". There's much discussion of editor interfaces, and this
> > leaves the possibility of git-blaming the contents of the editor buffer
> > (after writing it out to a temp file) without having to save changes to
> > the working tree file.
> 
> I agree, that probably would make most sense. If we do this at all. On the 
> other hand, I suspect that most editors would probably want to pipe the 
> contents to the program, not write it to a temp-file.

... and use it with --incremental, as well.  In emacs you can have the
annotation take place as it is being written out relatively easily, by
arranging to have a callback function get called each time more
information is handed back to emacs via a pipe.

						- Ted

Re: What's in git.git (stable)

[Jakub Narebski]

Theodore Tso wrote:
> On Sun, Feb 04, 2007 at 11:12:34AM -0800, Linus Torvalds wrote:
>> On Sun, 4 Feb 2007, Jeff King wrote:
>>> 
>>> Just a thought, but it might be useful to blame the contents of an
>>> arbitrary file (but starting the history at a given pathname). Something
>>> like "git blame --contents /tmp/foo.c file.c", with contents defaulting
>>> to "file.c". There's much discussion of editor interfaces, and this
>>> leaves the possibility of git-blaming the contents of the editor buffer
>>> (after writing it out to a temp file) without having to save changes to
>>> the working tree file.
>> 
>> I agree, that probably would make most sense. If we do this at all. On the 
>> other hand, I suspect that most editors would probably want to pipe the 
>> contents to the program, not write it to a temp-file.
> 
> ... and use it with --incremental, as well.  In emacs you can have the
> annotation take place as it is being written out relatively easily, by
> arranging to have a callback function get called each time more
> information is handed back to emacs via a pipe.

So perhaps instead of "git blame --contents /tmp/foo.c file.c" we should
have "cat /tmp/foo.c | git blame --stdin file.c", hmmm? Editor would then
pipe current contents of the buffer to "git blame --stdin --incremental
file.c" (where file.c is the name in tree/in HEAD).

-- 
Jakub Narebski
Warsaw, Poland
ShadeHawk on #git

Re: What's in git.git (stable)

[David Kågedal]

Jakub Narebski <jnareb@gmail.com> writes:

> Theodore Tso wrote:
>> On Sun, Feb 04, 2007 at 11:12:34AM -0800, Linus Torvalds wrote:
>>> On Sun, 4 Feb 2007, Jeff King wrote:
>>>> 
>>>> Just a thought, but it might be useful to blame the contents of an
>>>> arbitrary file (but starting the history at a given pathname). Something
>>>> like "git blame --contents /tmp/foo.c file.c", with contents defaulting
>>>> to "file.c". There's much discussion of editor interfaces, and this
>>>> leaves the possibility of git-blaming the contents of the editor buffer
>>>> (after writing it out to a temp file) without having to save changes to
>>>> the working tree file.
>>> 
>>> I agree, that probably would make most sense. If we do this at all. On the 
>>> other hand, I suspect that most editors would probably want to pipe the 
>>> contents to the program, not write it to a temp-file.
>> 
>> ... and use it with --incremental, as well.  In emacs you can have the
>> annotation take place as it is being written out relatively easily, by
>> arranging to have a callback function get called each time more
>> information is handed back to emacs via a pipe.
>
> So perhaps instead of "git blame --contents /tmp/foo.c file.c" we should
> have "cat /tmp/foo.c | git blame --stdin file.c", hmmm? Editor would then
> pipe current contents of the buffer to "git blame --stdin --incremental
> file.c" (where file.c is the name in tree/in HEAD).

But if we allow the standard convention of using - to mean stdin, the
--contents option would be more flexible, since "--contents -" would
just be a special case.

-- 
David Kågedal

[PATCH 1/2] Rename get_ident() to fmt_ident() and make it available to outside

[Junio C Hamano]

This makes the functionality of ident.c::get_ident() available to
other callers.

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

 * This by itself is totally uninteresting, but the second one
   depends on it.

 cache.h |    1 +
 ident.c |    8 ++++----
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/cache.h b/cache.h
index 201704b..38a9bc0 100644
--- a/cache.h
+++ b/cache.h
@@ -321,6 +321,7 @@ unsigned long approxidate(const char *);
 
 extern const char *git_author_info(int);
 extern const char *git_committer_info(int);
+extern const char *fmt_ident(const char *name, const char *email, const char *date_str, int);
 
 struct checkout {
 	const char *base_dir;
diff --git a/ident.c b/ident.c
index a6fc7b5..bb03bdd 100644
--- a/ident.c
+++ b/ident.c
@@ -185,8 +185,8 @@ static const char *env_hint =
 "Add --global to set your account\'s default\n"
 "\n";
 
-static const char *get_ident(const char *name, const char *email,
-			     const char *date_str, int error_on_no_name)
+const char *fmt_ident(const char *name, const char *email,
+		      const char *date_str, int error_on_no_name)
 {
 	static char buffer[1000];
 	char date[50];
@@ -233,7 +233,7 @@ static const char *get_ident(const char *name, const char *email,
 
 const char *git_author_info(int error_on_no_name)
 {
-	return get_ident(getenv("GIT_AUTHOR_NAME"),
+	return fmt_ident(getenv("GIT_AUTHOR_NAME"),
 			 getenv("GIT_AUTHOR_EMAIL"),
 			 getenv("GIT_AUTHOR_DATE"),
 			 error_on_no_name);
@@ -241,7 +241,7 @@ const char *git_author_info(int error_on_no_name)
 
 const char *git_committer_info(int error_on_no_name)
 {
-	return get_ident(getenv("GIT_COMMITTER_NAME"),
+	return fmt_ident(getenv("GIT_COMMITTER_NAME"),
 			 getenv("GIT_COMMITTER_EMAIL"),
 			 getenv("GIT_COMMITTER_DATE"),
 			 error_on_no_name);
-- 
1.5.0.rc3.40.g1f7d

[PATCH 2/2] git-blame: no rev means start from the working tree file.

[Junio C Hamano]

Warning: this changes the semantics.

This makes "git blame" without any positive rev to start digging
from the working tree copy, which is made into a fake commit
whose sole parent is the HEAD.

It also adds --contents <file> option to pretend as if the
working tree copy has the contents of the named file.  You can
use '-' to make the command read from the standard input.

If you want the command to start annotating from the HEAD
commit, you need to explicitly give HEAD parameter.

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

 Theodore Tso <tytso@mit.edu> writes:

 > On Sun, Feb 04, 2007 at 11:12:34AM -0800, Linus Torvalds wrote:
 >> 
 >> On Sun, 4 Feb 2007, Jeff King wrote:
 >> > 
 >> > Just a thought, but it might be useful to blame the contents of an
 >> > arbitrary file (but starting the history at a given pathname). Something
 >> > like "git blame --contents /tmp/foo.c file.c", with contents defaulting
 >> > to "file.c". There's much discussion of editor interfaces, and this
 >> > leaves the possibility of git-blaming the contents of the editor buffer
 >> > (after writing it out to a temp file) without having to save changes to
 >> > the working tree file.
 >> 
 >> I agree, that probably would make most sense. If we do this at all. On the 
 >> other hand, I suspect that most editors would probably want to pipe the 
 >> contents to the program, not write it to a temp-file.
 >
 > ... and use it with --incremental, as well.  In emacs you can have the
 > annotation take place as it is being written out relatively easily, by
 > arranging to have a callback function get called each time more
 > information is handed back to emacs via a pipe.

 * By popular request, this is the second round of the patch.
   I've splitted out a small change to make ident.c::get_ident()
   available to this one into a separate commit, which this
   patch depends on.

 builtin-blame.c |  209 ++++++++++++++++++++++++++++++++++++++++++++++++-------
 diff-lib.c      |   44 +++++++++++-
 diff.h          |    1 +
 3 files changed, 228 insertions(+), 26 deletions(-)

diff --git a/builtin-blame.c b/builtin-blame.c
index 1c21204..897323a 100644
--- a/builtin-blame.c
+++ b/builtin-blame.c
@@ -15,9 +15,10 @@
 #include "revision.h"
 #include "quote.h"
 #include "xdiff-interface.h"
+#include "cache-tree.h"
 
 static char blame_usage[] =
-"git-blame [-c] [-l] [-t] [-f] [-n] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [commit] [--] file\n"
+"git-blame [-c] [-l] [-t] [-f] [-n] [-p] [-L n,m] [-S <revs-file>] [-M] [-C] [-C] [--contents <filename>] [commit] [--] file\n"
 "  -c, --compatibility Use the same output mode as git-annotate (Default: off)\n"
 "  -b                  Show blank SHA-1 for boundary commits (Default: off)\n"
 "  -l, --long          Show long commit SHA1 (Default: off)\n"
@@ -29,6 +30,7 @@ static char blame_usage[] =
 "  -L n,m              Process only line range n,m, counting from 1\n"
 "  -M, -C              Find line movements within and across files\n"
 "  --incremental       Show blame entries as we find them, incrementally\n"
+"  --contents file     Use <file>'s contents as the final image\n"
 "  -S revs-file        Use revisions from revs-file instead of calling git-rev-list\n";
 
 static int longest_file;
@@ -333,9 +335,13 @@ static struct origin *find_origin(struct scoreboard *sb,
 	diff_tree_setup_paths(paths, &diff_opts);
 	if (diff_setup_done(&diff_opts) < 0)
 		die("diff-setup");
-	diff_tree_sha1(parent->tree->object.sha1,
-		       origin->commit->tree->object.sha1,
-		       "", &diff_opts);
+
+	if (is_null_sha1(origin->commit->object.sha1))
+		do_diff_cache(parent->tree->object.sha1, &diff_opts);
+	else
+		diff_tree_sha1(parent->tree->object.sha1,
+			       origin->commit->tree->object.sha1,
+			       "", &diff_opts);
 	diffcore_std(&diff_opts);
 
 	/* It is either one entry that says "modified", or "created",
@@ -402,9 +408,13 @@ static struct origin *find_rename(struct scoreboard *sb,
 	diff_tree_setup_paths(paths, &diff_opts);
 	if (diff_setup_done(&diff_opts) < 0)
 		die("diff-setup");
-	diff_tree_sha1(parent->tree->object.sha1,
-		       origin->commit->tree->object.sha1,
-		       "", &diff_opts);
+
+	if (is_null_sha1(origin->commit->object.sha1))
+		do_diff_cache(parent->tree->object.sha1, &diff_opts);
+	else
+		diff_tree_sha1(parent->tree->object.sha1,
+			       origin->commit->tree->object.sha1,
+			       "", &diff_opts);
 	diffcore_std(&diff_opts);
 
 	for (i = 0; i < diff_queued_diff.nr; i++) {
@@ -1047,9 +1057,12 @@ static int find_copy_in_parent(struct scoreboard *sb,
 	    (!porigin || strcmp(target->path, porigin->path)))
 		diff_opts.find_copies_harder = 1;
 
-	diff_tree_sha1(parent->tree->object.sha1,
-		       target->commit->tree->object.sha1,
-		       "", &diff_opts);
+	if (is_null_sha1(target->commit->object.sha1))
+		do_diff_cache(parent->tree->object.sha1, &diff_opts);
+	else
+		diff_tree_sha1(parent->tree->object.sha1,
+			       target->commit->tree->object.sha1,
+			       "", &diff_opts);
 
 	if (!diff_opts.find_copies_harder)
 		diffcore_std(&diff_opts);
@@ -1336,9 +1349,9 @@ static void get_commit_info(struct commit *commit,
 	tmp += 2;
 	endp = strchr(tmp, '\n');
 	if (!endp)
-		goto error_out;
+		endp = tmp + strlen(tmp);
 	len = endp - tmp;
-	if (len >= sizeof(summary_buf))
+	if (len >= sizeof(summary_buf) || len == 0)
 		goto error_out;
 	memcpy(summary_buf, tmp, len);
 	summary_buf[len] = 0;
@@ -1910,6 +1923,137 @@ static int git_blame_config(const char *var, const char *value)
 	return git_default_config(var, value);
 }
 
+static struct commit *fake_working_tree_commit(const char *path, const char *contents_from)
+{
+	struct commit *commit;
+	struct origin *origin;
+	unsigned char head_sha1[20];
+	char *buf;
+	const char *ident;
+	int fd;
+	time_t now;
+	unsigned long fin_size;
+	int size, len;
+	struct cache_entry *ce;
+	unsigned mode;
+
+	if (get_sha1("HEAD", head_sha1))
+		die("No such ref: HEAD");
+
+	time(&now);
+	commit = xcalloc(1, sizeof(*commit));
+	commit->parents = xcalloc(1, sizeof(*commit->parents));
+	commit->parents->item = lookup_commit_reference(head_sha1);
+	commit->object.parsed = 1;
+	commit->date = now;
+	commit->object.type = OBJ_COMMIT;
+
+	origin = make_origin(commit, path);
+
+	if (!contents_from || strcmp("-", contents_from)) {
+		struct stat st;
+		const char *read_from;
+
+		if (contents_from) {
+			if (stat(contents_from, &st) < 0)
+				die("Cannot stat %s", contents_from);
+			read_from = contents_from;
+		}
+		else {
+			if (lstat(path, &st) < 0)
+				die("Cannot lstat %s", path);
+			read_from = path;
+		}
+		fin_size = st.st_size;
+		buf = xmalloc(fin_size+1);
+		mode = canon_mode(st.st_mode);
+		switch (st.st_mode & S_IFMT) {
+		case S_IFREG:
+			fd = open(read_from, O_RDONLY);
+			if (fd < 0)
+				die("cannot open %s", read_from);
+			if (read_in_full(fd, buf, fin_size) != fin_size)
+				die("cannot read %s", read_from);
+			break;
+		case S_IFLNK:
+			if (readlink(read_from, buf, fin_size+1) != fin_size)
+				die("cannot readlink %s", read_from);
+			break;
+		default:
+			die("unsupported file type %s", read_from);
+		}
+	}
+	else {
+		/* Reading from stdin */
+		contents_from = "standard input";
+		buf = NULL;
+		fin_size = 0;
+		mode = 0;
+		while (1) {
+			ssize_t cnt = 8192;
+			buf = xrealloc(buf, fin_size + cnt);
+			cnt = xread(0, buf + fin_size, cnt);
+			if (cnt < 0)
+				die("read error %s from stdin",
+				    strerror(errno));
+			if (!cnt)
+				break;
+			fin_size += cnt;
+		}
+		buf = xrealloc(buf, fin_size + 1);
+	}
+	buf[fin_size] = 0;
+	origin->file.ptr = buf;
+	origin->file.size = fin_size;
+	write_sha1_file(buf, fin_size, blob_type, origin->blob_sha1);
+	commit->util = origin;
+
+	/*
+	 * Read the current index, replace the path entry with
+	 * origin->blob_sha1 without mucking with its mode or type
+	 * bits; we are not going to write this index out -- we just
+	 * want to run "diff-index --cached".
+	 */
+	discard_cache();
+	read_cache();
+
+	len = strlen(path);
+	if (!mode) {
+		int pos = cache_name_pos(path, len);
+		if (0 <= pos)
+			mode = ntohl(active_cache[pos]->ce_mode);
+		else
+			/* Let's not bother reading from HEAD tree */
+			mode = S_IFREG | 0644;
+	}
+	size = cache_entry_size(len);
+	ce = xcalloc(1, size);
+	hashcpy(ce->sha1, origin->blob_sha1);
+	memcpy(ce->name, path, len);
+	ce->ce_flags = create_ce_flags(len, 0);
+	ce->ce_mode = create_ce_mode(mode);
+	add_cache_entry(ce, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE);
+
+	/*
+	 * We are not going to write this out, so this does not matter
+	 * right now, but someday we might optimize diff-index --cached
+	 * with cache-tree information.
+	 */
+	cache_tree_invalidate_path(active_cache_tree, path);
+
+	commit->buffer = xmalloc(400);
+	ident = fmt_ident("Not Committed Yet", "not.committed.yet", NULL, 0);
+	sprintf(commit->buffer,
+		"tree 0000000000000000000000000000000000000000\n"
+		"parent %s\n"
+		"author %s\n"
+		"committer %s\n\n"
+		"Version of %s from %s\n",
+		sha1_to_hex(head_sha1),
+		ident, ident, path, contents_from ? contents_from : path);
+	return commit;
+}
+
 int cmd_blame(int argc, const char **argv, const char *prefix)
 {
 	struct rev_info revs;
@@ -1924,6 +2068,7 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 	const char *final_commit_name = NULL;
 	char type[10];
 	const char *bottomtop = NULL;
+	const char *contents_from = NULL;
 
 	git_config(git_blame_config);
 	save_commit_buffer = 0;
@@ -1968,6 +2113,11 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 				die("More than one '-L n,m' option given");
 			bottomtop = arg;
 		}
+		else if (!strcmp("--contents", arg)) {
+			if (++i >= argc)
+				usage(blame_usage);
+			contents_from = argv[i];
+		}
 		else if (!strcmp("--incremental", arg))
 			incremental = 1;
 		else if (!strcmp("--score-debug", arg))
@@ -2087,7 +2237,7 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 	argv[unk] = NULL;
 
 	init_revisions(&revs, NULL);
-	setup_revisions(unk, argv, &revs, "HEAD");
+	setup_revisions(unk, argv, &revs, NULL);
 	memset(&sb, 0, sizeof(sb));
 
 	/*
@@ -2114,16 +2264,14 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 	if (!sb.final) {
 		/*
 		 * "--not A B -- path" without anything positive;
-		 * default to HEAD.
+		 * do not default to HEAD, but use the working tree
+		 * or "--contents".
 		 */
-		unsigned char head_sha1[20];
-
-		final_commit_name = "HEAD";
-		if (get_sha1(final_commit_name, head_sha1))
-			die("No such ref: HEAD");
-		sb.final = lookup_commit_reference(head_sha1);
-		add_pending_object(&revs, &(sb.final->object), "HEAD");
+		sb.final = fake_working_tree_commit(path, contents_from);
+		add_pending_object(&revs, &(sb.final->object), ":");
 	}
+	else if (contents_from)
+		die("Cannot use --contents with final commit object name");
 
 	/*
 	 * If we have bottom, this will mark the ancestors of the
@@ -2132,11 +2280,22 @@ int cmd_blame(int argc, const char **argv, const char *prefix)
 	 */
 	prepare_revision_walk(&revs);
 
-	o = get_origin(&sb, sb.final, path);
-	if (fill_blob_sha1(o))
-		die("no such path %s in %s", path, final_commit_name);
+	if (is_null_sha1(sb.final->object.sha1)) {
+		char *buf;
+		o = sb.final->util;
+		buf = xmalloc(o->file.size + 1);
+		memcpy(buf, o->file.ptr, o->file.size + 1);
+		sb.final_buf = buf;
+		sb.final_buf_size = o->file.size;
+	}
+	else {
+		o = get_origin(&sb, sb.final, path);
+		if (fill_blob_sha1(o))
+			die("no such path %s in %s", path, final_commit_name);
 
-	sb.final_buf = read_sha1_file(o->blob_sha1, type, &sb.final_buf_size);
+		sb.final_buf = read_sha1_file(o->blob_sha1, type,
+					      &sb.final_buf_size);
+	}
 	num_read_blob++;
 	lno = prepare_lines(&sb);
 
diff --git a/diff-lib.c b/diff-lib.c
index 2c9be60..91cd877 100644
--- a/diff-lib.c
+++ b/diff-lib.c
@@ -7,6 +7,7 @@
 #include "diff.h"
 #include "diffcore.h"
 #include "revision.h"
+#include "cache-tree.h"
 
 /*
  * diff-files
@@ -271,7 +272,7 @@ static int diff_cache(struct rev_info *revs,
 				break;
 			}
 			/* Show difference between old and new */
-			show_modified(revs,ac[1], ce, 1,
+			show_modified(revs, ac[1], ce, 1,
 				      cached, match_missing);
 			break;
 		case 1:
@@ -372,3 +373,44 @@ int run_diff_index(struct rev_info *revs, int cached)
 	diff_flush(&revs->diffopt);
 	return ret;
 }
+
+int do_diff_cache(const unsigned char *tree_sha1, struct diff_options *opt)
+{
+	struct tree *tree;
+	struct rev_info revs;
+	int i;
+	struct cache_entry **dst;
+	struct cache_entry *last = NULL;
+
+	/*
+	 * This is used by git-blame to run diff-cache internally;
+	 * it potentially needs to repeatedly run this, so we will
+	 * start by removing the higher order entries the last round
+	 * left behind.
+	 */
+	dst = active_cache;
+	for (i = 0; i < active_nr; i++) {
+		struct cache_entry *ce = active_cache[i];
+		if (ce_stage(ce)) {
+			if (last && !strcmp(ce->name, last->name))
+				continue;
+			cache_tree_invalidate_path(active_cache_tree,
+						   ce->name);
+			last = ce;
+			ce->ce_mode = 0;
+			ce->ce_flags &= ~htons(CE_STAGEMASK);
+		}
+		*dst++ = ce;
+	}
+	active_nr = dst - active_cache;
+
+	init_revisions(&revs, NULL);
+	revs.prune_data = opt->paths;
+	tree = parse_tree_indirect(tree_sha1);
+	if (!tree)
+		die("bad tree object %s", sha1_to_hex(tree_sha1));
+	if (read_tree(tree, 1, opt->paths))
+		return error("unable to read tree %s", sha1_to_hex(tree_sha1));
+	return diff_cache(&revs, active_cache, active_nr, revs.prune_data,
+			  1, 0);
+}
diff --git a/diff.h b/diff.h
index 7a347cf..eece65d 100644
--- a/diff.h
+++ b/diff.h
@@ -222,6 +222,7 @@ extern int run_diff_files(struct rev_info *revs, int silent_on_removed);
 
 extern int run_diff_index(struct rev_info *revs, int cached);
 
+extern int do_diff_cache(const unsigned char *, struct diff_options *);
 extern int diff_flush_patch_id(struct diff_options *, unsigned char *);
 
 #endif /* DIFF_H */
-- 
1.5.0.rc3.40.g1f7d

Quick status updates

[Junio C Hamano]

Junio C Hamano <junkio@cox.net> writes:

> Several things to note (in the following, substitute $gmane with
> http://article.gmane.org/gmane.comp.version-control.git):
>
>  - Working around Tk geometry problem, especially on non Linux+X
>    platforms.  I've forwarded Mark Levedahl's patches
>    ($gmane/38361) to Paul Mackerras for his blessing; hopefully
>    he can Ack and/or suggest improvements.  I'd really like to
>    have them in 1.5.0 in some form.

Mark and I got a response from Paul; unfortunately he is too
busy right now, immediately after post 2.6.20 merge window has
opened.  I've parked Mark's patches and another one suggested by
Linus in response to Johannes Sixt's MinGW work in 'next' for
now.  Depending on Paul's availablity, I might end up applying
them in my tree first.

>  - Nico's "reflog on HEAD".
>  - Reverting the patch to allow tracking branch names as the
>    value of branch.*.merge ($gmane/38621).
>  - Teaching "git blame" to also use the working tree files
>    and/or index.
>  - "git remote add -t -f -m" ($gmane/38470).

These four are in ("git clone" rewrite is not).

>  - We might want to allow git-push to use the wildcard refspec,
>    like git-fetch does, for symmetry.

I've done this which was not too painful -- a patch in a
separate message to follow.

I also applied a handful patches from early this year that was
left unapplied.  After reviewing the message I am responding to,
and an earlier "a few remaining issues..." ($gmane/35993), I
think I am pretty much done with 1.5.0 preparation.

After pulling gfi from Shawn and applying the outstanding gitk
patches (either via Paul or myself), I'll tag -rc4 and give it a
week or so before the final.

I should remember to remove git-merge-recur before -rc4 ;-).

[PATCH] git-push: allow globbing wildcard refspec.

[Junio C Hamano]

This allows you to set up mothership-satellite configuration
more symmetrically and naturally by allowing the globbing
wildcard refspec for git-push.  On your satellite machine:

    [remote "mothership"]
        url = mothership:project.git
        pull = refs/heads/*:refs/remotes/mothership/*
        push = refs/heads/*:refs/remotes/satellite/*

You would say "git fetch mothership" to update your tracking
branches under mothership/ to keep track of the progress on the
mothership side, and when you are done working on the satellite
machine, you would "git push mothership" to update their
tracking branches under satellite/.  Corresponding configuration
on the mothership machine can be used to make "git fetch satellite"
update its tracking branch under satellite/. on the mothership.

Signed-off-by: Junio C Hamano <junkio@cox.net>
---
 builtin-push.c |  114 ++++++++++++++++++++++++++++++++++++++++----------------
 1 files changed, 82 insertions(+), 32 deletions(-)

diff --git a/builtin-push.c b/builtin-push.c
index 5f4d7d3..c45649e 100644
--- a/builtin-push.c
+++ b/builtin-push.c
@@ -54,38 +54,84 @@ static void expand_refspecs(void)
 	for_each_ref(expand_one_ref, NULL);
 }
 
+struct wildcard_cb {
+	const char *from_prefix;
+	int from_prefix_len;
+	const char *to_prefix;
+	int to_prefix_len;
+	int force;
+};
+
+static int expand_wildcard_ref(const char *ref, const unsigned char *sha1, int flag, void *cb_data)
+{
+	struct wildcard_cb *cb = cb_data;
+	int len = strlen(ref);
+	char *expanded, *newref;
+
+	if (len < cb->from_prefix_len ||
+	    memcmp(cb->from_prefix, ref, cb->from_prefix_len))
+		return 0;
+	expanded = xmalloc(len * 2 + cb->force +
+			   (cb->to_prefix_len - cb->from_prefix_len) + 2);
+	newref = expanded + cb->force;
+	if (cb->force)
+		expanded[0] = '+';
+	memcpy(newref, ref, len);
+	newref[len] = ':';
+	memcpy(newref + len + 1, cb->to_prefix, cb->to_prefix_len);
+	strcpy(newref + len + 1 + cb->to_prefix_len,
+	       ref + cb->from_prefix_len);
+	add_refspec(expanded);
+	return 0;
+}
+
+static int wildcard_ref(const char *ref)
+{
+	int len;
+	const char *colon;
+	struct wildcard_cb cb;
+
+	memset(&cb, 0, sizeof(cb));
+	if (ref[0] == '+') {
+		cb.force = 1;
+		ref++;
+	}
+	len = strlen(ref);
+	colon = strchr(ref, ':');
+	if (! (colon && ref < colon &&
+	       colon[-2] == '/' && colon[-1] == '*' &&
+	       /* "<mine>/<asterisk>:<yours>/<asterisk>" is at least 7 bytes */
+	       7 <= len &&
+	       ref[len-2] == '/' && ref[len-1] == '*') )
+		return 0 ;
+	cb.from_prefix = ref;
+	cb.from_prefix_len = colon - ref - 1;
+	cb.to_prefix = colon + 1;
+	cb.to_prefix_len = len - (colon - ref) - 2;
+	for_each_ref(expand_wildcard_ref, &cb);
+	return 1;
+}
+
 static void set_refspecs(const char **refs, int nr)
 {
 	if (nr) {
-		int pass;
-		for (pass = 0; pass < 2; pass++) {
-			/* pass 0 counts and allocates, pass 1 fills */
-			int i, cnt;
-			for (i = cnt = 0; i < nr; i++) {
-				if (!strcmp("tag", refs[i])) {
-					int len;
-					char *tag;
-					if (nr <= ++i)
-						die("tag <tag> shorthand without <tag>");
-					if (pass) {
-						len = strlen(refs[i]) + 11;
-						tag = xmalloc(len);
-						strcpy(tag, "refs/tags/");
-						strcat(tag, refs[i]);
-						refspec[cnt] = tag;
-					}
-					cnt++;
-					continue;
-				}
-				if (pass)
-					refspec[cnt] = refs[i];
-				cnt++;
-			}
-			if (!pass) {
-				size_t bytes = cnt * sizeof(char *);
-				refspec_nr = cnt;
-				refspec = xrealloc(refspec, bytes);
+		int i;
+		for (i = 0; i < nr; i++) {
+			const char *ref = refs[i];
+			if (!strcmp("tag", ref)) {
+				char *tag;
+				int len;
+				if (nr <= ++i)
+					die("tag shorthand without <tag>");
+				len = strlen(refs[i]) + 11;
+				tag = xmalloc(len);
+				strcpy(tag, "refs/tags/");
+				strcat(tag, refs[i]);
+				ref = tag;
 			}
+			else if (wildcard_ref(ref))
+				continue;
+			add_refspec(ref);
 		}
 	}
 	expand_refspecs();
@@ -129,8 +175,10 @@ static int get_remotes_uri(const char *repo, const char *uri[MAX_URI])
 			else
 				error("more than %d URL's specified, ignoring the rest", MAX_URI);
 		}
-		else if (is_refspec && !has_explicit_refspec)
-			add_refspec(xstrdup(s));
+		else if (is_refspec && !has_explicit_refspec) {
+			if (!wildcard_ref(s))
+				add_refspec(xstrdup(s));
+		}
 	}
 	fclose(f);
 	if (!n)
@@ -156,8 +204,10 @@ static int get_remote_config(const char* key, const char* value)
 				error("more than %d URL's specified, ignoring the rest", MAX_URI);
 		}
 		else if (config_get_refspecs &&
-			 !strcmp(key + 7 + config_repo_len, ".push"))
-			add_refspec(xstrdup(value));
+			 !strcmp(key + 7 + config_repo_len, ".push")) {
+			if (!wildcard_ref(value))
+				add_refspec(xstrdup(value));
+		}
 		else if (config_get_receivepack &&
 			 !strcmp(key + 7 + config_repo_len, ".receivepack")) {
 			if (!receivepack) {
-- 
1.5.0.rc3.83.g88293c-dirty

Re: [PATCH] git-push: allow globbing wildcard refspec.

[Shawn O. Pearce]

Junio C Hamano <junkio@cox.net> wrote:
>     [remote "mothership"]
>         url = mothership:project.git
>         pull = refs/heads/*:refs/remotes/mothership/*

Don't you mean `fetch = ...` here?

-- 
Shawn.