Code

git.git
17 years agoMerge branch 'jk/diff'
Junio C Hamano [Mon, 18 Sep 2006 01:14:03 +0000 (18:14 -0700)]
Merge branch 'jk/diff'

* jk/diff:
  wt-status: remove extraneous newline from 'deleted:' output
  git-status: document colorization config options
  Teach runstatus about --untracked
  git-commit.sh: convert run_status to a C builtin
  Move color option parsing out of diff.c and into color.[ch]
  diff: support custom callbacks for output

17 years agogitweb fix validating pg (page) parameter
Matthias Lederhofer [Sun, 17 Sep 2006 11:52:45 +0000 (13:52 +0200)]
gitweb fix validating pg (page) parameter

Currently it is possible to give any string ending with a number as
page.  -1 for example is quite bad (error log shows probably 100
warnings).

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-repack(1): document --window and --depth
Jonas Fonseca [Sun, 17 Sep 2006 11:02:59 +0000 (13:02 +0200)]
git-repack(1): document --window and --depth

Copy and pasted from git-pack-objects(1).

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-apply(1): document --unidiff-zero
Jonas Fonseca [Sun, 17 Sep 2006 10:55:38 +0000 (12:55 +0200)]
git-apply(1): document --unidiff-zero

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: fix warnings in PATH_INFO code and add export_ok/strict_export
Matthias Lederhofer [Sun, 17 Sep 2006 13:29:48 +0000 (15:29 +0200)]
gitweb: fix warnings in PATH_INFO code and add export_ok/strict_export

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoupload-archive: monitor child communication even more carefully.
Franck Bui-Huu [Sun, 17 Sep 2006 14:09:18 +0000 (16:09 +0200)]
upload-archive: monitor child communication even more carefully.

The current code works like this: if others flags than POLLIN is
raised we assume that (a) something bad happened and the child died or
(b) the child has closed the pipe because it had no more data to send.

For the latter case, we assume wrongly that one call to
process_input() will empty the pipe. Indeed it reads only 16Ko of data
by call and the the pipe capacity can be larger than that (on current
Linux kernel, it is 65536 bytes). Therefore the child can write 32ko
of data, for example, and close the pipe. After that poll will return
POLLIN _and_ POLLHUP and the parent will read only 16ko of data.

This patch forces the parent to empty the pipe as soon as POLLIN is
raised and even if POLLHUP or something else is raised too.

Moreover, some implementations of poll might return POLLRDNORM flag
even if it is non standard.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/archive'
Junio C Hamano [Sun, 17 Sep 2006 09:46:00 +0000 (02:46 -0700)]
Merge branch 'jc/archive'

* jc/archive:
  git-tar-tree: devolve git-tar-tree into a wrapper for git-archive
  git-archive: inline default_parse_extra()
  builtin-archive.c: rename remote_request() to extract_remote_arg()
  upload-archive: monitor child communication more carefully.
  Add sideband status report to git-archive protocol
  Prepare larger packet buffer for upload-pack protocol.
  Teach --exec to git-archive --remote
  Add --verbose to git-archive
  archive: force line buffered output to stderr
  Use xstrdup instead of strdup in builtin-{tar,zip}-tree.c
  Move sideband server side support into reusable form.
  Move sideband client side support into reusable form.
  archive: allow remote to have more formats than we understand.
  git-archive: make compression level of ZIP archives configurable
  Add git-upload-archive
  git-archive: wire up ZIP format.
  git-archive: wire up TAR format.
  Add git-archive

17 years agogitweb: export options
Matthias Lederhofer [Sat, 16 Sep 2006 22:31:01 +0000 (00:31 +0200)]
gitweb: export options

$export_ok: If this variable evaluates to true it is checked
if a file with this name exists in the repository.  If it
does not exist the repository cannot be viewed from gitweb.
(Similar to git-daemon-export-ok for git-daemon).

$strict_export: If this variable evaluates to true only
repositories listed on the project-list-page of gitweb can
be accessed.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: do not use 'No such directory' error message
Matthias Lederhofer [Sat, 16 Sep 2006 22:30:27 +0000 (00:30 +0200)]
gitweb: do not use 'No such directory' error message

undef $project; to prevent a file named description to be read.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-tar-tree: devolve git-tar-tree into a wrapper for git-archive
Rene Scharfe [Sat, 16 Sep 2006 19:20:36 +0000 (21:20 +0200)]
git-tar-tree: devolve git-tar-tree into a wrapper for git-archive

This patch removes the custom tree walker tree_traverse(), and makes
generate_tar() use write_tar_archive() and the infrastructure provided
by git-archive instead.

As a kind of side effect, make write_tar_archive() able to handle NULL
as base directory, as this is what the new and simple generate_tar()
uses to indicate the absence of a base directory.  This was simpler
and cleaner than playing tricks with empty strings.

The behaviour of git-tar-tree should be unchanged (quick tests didn't
indicate otherwise) except for the text of some error messages.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: fix uninitialized variable warning.
Martin Waitz [Sat, 16 Sep 2006 21:09:02 +0000 (23:09 +0200)]
gitweb: fix uninitialized variable warning.

Perl spit out a varning when "blob" or "blob_plain" actions were
used without a $hash parameter.

Signed-off-by: Martin Waitz <tali@admingilde.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: more support for PATH_INFO based URLs
Martin Waitz [Sat, 16 Sep 2006 21:08:32 +0000 (23:08 +0200)]
gitweb: more support for PATH_INFO based URLs

Now three types of path based URLs are supported:
gitweb.cgi/project.git
gitweb.cgi/project.git/branch
gitweb.cgi/project.git/branch/filename

The first one (show project summary) was already supported for a long time
now.  The other two are new: they show the shortlog of a branch or
the plain file contents of some file contained in the repository.

This is especially useful to support project web pages for small
projects: just create an html branch and then use an URL like
gitweb.cgi/project.git/html/index.html.

Signed-off-by: Martin Waitz <tali@admingilde.org>
Acked-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoapply --unidiff-zero: loosen sanity checks for --unidiff=0 patches
Junio C Hamano [Sun, 17 Sep 2006 08:04:24 +0000 (01:04 -0700)]
apply --unidiff-zero: loosen sanity checks for --unidiff=0 patches

In "git-apply", we have a few sanity checks and heuristics that
expects that the patch fed to us is a unified diff with at least
one line of context.

 * When there is no leading context line in a hunk, the hunk
   must apply at the beginning of the preimage.  Similarly, no
   trailing context means that the hunk is anchored at the end.

 * We learn a patch deletes the file from a hunk that has no
   resulting line (i.e. all lines are prefixed with '-') if it
   has not otherwise been known if the patch deletes the file.
   Similarly, no old line means the file is being created.

And we declare an error condition when the file created by a
creation patch already exists, and/or when a deletion patch
still leaves content in the file.

These sanity checks are good safety measures, but breaks down
when people feed a diff generated with --unified=0.  This was
recently noticed first by Matthew Wilcox and Gerrit Pape.

This adds a new flag, --unified-zero, to allow bypassing these
checks.  If you are in control of the patch generation process,
you should not use --unified=0 patch and fix it up with this
flag; rather you should try work with a patch with context.  But
if all you have to work with is a patch without context, this
flag may come handy as the last resort.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agot1400: make test debuggable.
Junio C Hamano [Sat, 16 Sep 2006 20:41:14 +0000 (13:41 -0700)]
t1400: make test debuggable.

I had a hard time figuring out why this test was failing with
the packed-refs update without running it under "sh -x".  This
makes output from "sh t1400-update-ref.sh -v" more descriptive.

Updating other tests would be a good janitorial task.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAllow multiple "git_path()" uses
Linus Torvalds [Mon, 11 Sep 2006 19:03:15 +0000 (12:03 -0700)]
Allow multiple "git_path()" uses

This allows you to maintain a few filesystem pathnames concurrently, by
simply replacing the single static "pathname" buffer with a LRU of four
buffers.

We did exactly the same thing with sha1_to_hex(), for pretty much exactly
the same reason. Sometimes you want to use two pathnames, and while it's
easy enough to xstrdup() them, why not just do the LU buffer thing.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Add link to "project_index" view to "project_list" page
Jakub Narebski [Fri, 15 Sep 2006 09:11:33 +0000 (11:11 +0200)]
gitweb: Add link to "project_index" view to "project_list" page

Add link to "project_index" view as [TXT] beside link to "opml" view,
(which is marked by [OPML]) to "project_list" page.

While at it add alternate links for "opml" and "project_list" to HTML
header for "project_list" view.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Allow for href() to be used for links without project param
Jakub Narebski [Fri, 15 Sep 2006 17:30:34 +0000 (19:30 +0200)]
gitweb: Allow for href() to be used for links without project param

Make it possible to use href() subroutine to generate link with
query string which does not include project ('p') parameter.
href() used to add project=$project to its parameters, if it
was not set (to be more exact if $params{'project'} was false).
Now you can pass "project => undef" if you don't want for href()
to add project parameter to query string in the generated link.

Links to "project_list", "project_index" and "opml" (all related
to list of all projects/all git repositories) doesn't need project
parameter. Moreover "project_list" is default view (action) if
project ('p') parameter is not set, just like "summary" is default
view (action) if project is set; project list served as a kind
of "home" page for gitweb instalation, and links to "project_list"
view were done without specyfying it as an action.

Convert remaining links (except $home_link and anchor links)
to use href(); this required adding 'order => "o"' to @mapping
in href(). This finishes consolidation of URL generation.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDefine fallback PATH_MAX on systems that do not define one in <limits.h>
Junio C Hamano [Sat, 16 Sep 2006 05:47:21 +0000 (22:47 -0700)]
Define fallback PATH_MAX on systems that do not define one in <limits.h>

Notably on GNU/Hurd, as reported by Gerrit Pape.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDocument git-grep -[Hh]
Junio C Hamano [Sat, 16 Sep 2006 01:37:01 +0000 (18:37 -0700)]
Document git-grep -[Hh]

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix duplicate xmalloc in builtin-add
Liu Yubao [Fri, 15 Sep 2006 20:46:07 +0000 (13:46 -0700)]
Fix duplicate xmalloc in builtin-add

[jc: patch came without sign-off but it was too obvious and trivial.]

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Add git_project_index for generating index.aux
Jakub Narebski [Fri, 15 Sep 2006 02:56:03 +0000 (04:56 +0200)]
gitweb: Add git_project_index for generating index.aux

Add git_project_index, which generates index.aux file that can be used
as a source of projects list, instead of generating projects list from
a directory.  Using file as a source of projects list allows for some
projects to be not present in gitweb main (project_list) page, and/or
correct project owner info. And is probably faster.

Additionally it can be used to get the list of all available repositories
for scripts (in easily parseable form).

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Do not parse refs by hand, use git-peek-remote instead
Jakub Narebski [Fri, 15 Sep 2006 01:43:28 +0000 (03:43 +0200)]
gitweb: Do not parse refs by hand, use git-peek-remote instead

This is in response to Linus's work on packed refs. Additionally it
makes gitweb work with symrefs, too.

Do not parse refs by hand, using File::Find and reading individual
heads to get hash of reference, but use git-peek-remote output
instead.  Assume that the hash for deref (with ^{}) always follows hash
for ref, and that we have derefs only for tag objects; this removes
call to git_get_type (and git-cat-file -t invocation) for tags, which
speeds "summary" and "tags" views generation, but might slow generation
of "heads" view a bit.  For now, we do not save and use the deref hash.

Remove git_get_hash_by_ref while at it, as git_get_refs_list was the
only place it was used.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Use File::Find::find in git_get_projects_list
Jakub Narebski [Thu, 14 Sep 2006 20:18:59 +0000 (22:18 +0200)]
gitweb: Use File::Find::find in git_get_projects_list

Earlier code to get list of projects when $projects_list is a
directory (e.g. when it is equal to $projectroot) had a hardcoded flat
(one level) list of directories.  Allow for projects to be in
subdirectories also for $projects_list being a directory by using
File::Find.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoHandle invalid argc gently
Dmitry V. Levin [Thu, 14 Sep 2006 01:04:09 +0000 (05:04 +0400)]
Handle invalid argc gently

describe, git: Handle argc==0 case the same way as argc==1.
merge-tree: Refuse excessive arguments.

Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd "-h/-H" parsing to "git grep"
Linus Torvalds [Thu, 14 Sep 2006 17:45:12 +0000 (10:45 -0700)]
Add "-h/-H" parsing to "git grep"

It turns out that I actually wanted to avoid the filenames (because I
didn't care - I just wanted to see the context in which something was
used) when doing a grep. But since "git grep" didn't take the "-h"
parameter, I ended up having to do "grep -5 -h *.c" instead.

So here's a trivial patch that adds "-h" (and thus has to enable -H too)
to "git grep" parsing.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTrivial support for cloning and fetching via ftp://.
Sasha Khapyorsky [Thu, 14 Sep 2006 02:24:04 +0000 (05:24 +0300)]
Trivial support for cloning and fetching via ftp://.

This adds trivial support for cloning and fetching via ftp://.

Signed-off-by: Sasha Khapyorsky <sashak@voltaire.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoDocumentation: Fix broken links
Dmitry V. Levin [Thu, 14 Sep 2006 01:04:33 +0000 (05:04 +0400)]
Documentation: Fix broken links

core-tutorial.txt, cvs-migration.txt, tutorial-2.txt:
Fix broken links.

Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMake count-objects, describe and merge-tree work in subdirectory
Dmitry V. Levin [Thu, 14 Sep 2006 01:03:59 +0000 (05:03 +0400)]
Make count-objects, describe and merge-tree work in subdirectory

Call setup_git_directory() to make these commands work in subdirectory.

Signed-off-by: Dmitry V. Levin <ldv@altlinux.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agowt-status: remove extraneous newline from 'deleted:' output
Jeff King [Wed, 13 Sep 2006 22:37:14 +0000 (18:37 -0400)]
wt-status: remove extraneous newline from 'deleted:' output

This was accidentally introduced during the fixes to avoid putting newlines
inside of colorized output.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-archive: inline default_parse_extra()
Rene Scharfe [Wed, 13 Sep 2006 20:55:04 +0000 (22:55 +0200)]
git-archive: inline default_parse_extra()

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branches 'maint' and 'jc/http'
Junio C Hamano [Thu, 14 Sep 2006 06:36:03 +0000 (23:36 -0700)]
Merge branches 'maint' and 'jc/http'

* maint:
  http-fetch: fix alternates handling.

* jc/http:
  http-fetch: fix alternates handling.

17 years agoMerge branch 'jc/unpack'
Junio C Hamano [Wed, 13 Sep 2006 20:00:02 +0000 (13:00 -0700)]
Merge branch 'jc/unpack'

* jc/unpack:
  unpack-objects -r: call it "recover".
  unpack-objects desperately salvages objects from a corrupt pack

17 years agounpack-objects -r: call it "recover".
Junio C Hamano [Wed, 13 Sep 2006 19:59:20 +0000 (12:59 -0700)]
unpack-objects -r: call it "recover".

The code called this operation "desperate" but the option flag is -r
and the word "recover" describes what it does better.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/binary'
Junio C Hamano [Wed, 13 Sep 2006 19:39:02 +0000 (12:39 -0700)]
Merge branch 'jc/binary'

* jc/binary:
  diff --binary generates full index on binary files.
  Make apply --binary a no-op.

17 years agoMerge branch 'aw/send-pack'
Junio C Hamano [Wed, 13 Sep 2006 19:30:20 +0000 (12:30 -0700)]
Merge branch 'aw/send-pack'

* aw/send-pack:
  send-pack: switch to using git-rev-list --stdin

17 years agoTest return value of finish_connect()
Franck Bui-Huu [Wed, 13 Sep 2006 08:26:47 +0000 (10:26 +0200)]
Test return value of finish_connect()

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agohttp-fetch: fix alternates handling. v1.4.2.1
Junio C Hamano [Wed, 13 Sep 2006 06:53:27 +0000 (23:53 -0700)]
http-fetch: fix alternates handling.

Fetch over http from a repository that uses alternates to borrow
from neighbouring repositories were quite broken, apparently for
some time now.

We parse input and count bytes to allocate the new buffer, and
when we copy into that buffer we know exactly how many bytes we
want to copy from where.  Using strlcpy for it was simply
stupid, and the code forgot to take it into account that strlcpy
terminated the string with NUL.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agohttp-fetch: fix alternates handling.
Junio C Hamano [Wed, 13 Sep 2006 06:53:27 +0000 (23:53 -0700)]
http-fetch: fix alternates handling.

Fetch over http from a repository that uses alternates to borrow
from neighbouring repositories were quite broken, apparently for
some time now.

We parse input and count bytes to allocate the new buffer, and
when we copy into that buffer we know exactly how many bytes we
want to copy from where.  Using strlcpy for it was simply
stupid, and the code forgot to take it into account that strlcpy
terminated the string with NUL.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agocontrib/vim: add syntax highlighting file for commits
Jeff King [Mon, 11 Sep 2006 23:22:49 +0000 (19:22 -0400)]
contrib/vim: add syntax highlighting file for commits

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-status: document colorization config options
Jeff King [Mon, 11 Sep 2006 23:21:17 +0000 (19:21 -0400)]
git-status: document colorization config options

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach runstatus about --untracked
Johannes Schindelin [Tue, 12 Sep 2006 20:45:12 +0000 (22:45 +0200)]
Teach runstatus about --untracked

Actually, teach runstatus what to do if it is not passed; it should not list
the contents of completely untracked directories, but only the name of that
directory (plus a trailing '/').

[jc: with comments by Jeff King to match hide-empty-directories
 behaviour of the original.]

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agobuiltin-archive.c: rename remote_request() to extract_remote_arg()
Junio C Hamano [Wed, 13 Sep 2006 05:42:31 +0000 (22:42 -0700)]
builtin-archive.c: rename remote_request() to extract_remote_arg()

Suggested by Franck, and I think it makes sense.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoupload-archive: monitor child communication more carefully.
Junio C Hamano [Tue, 12 Sep 2006 07:26:57 +0000 (00:26 -0700)]
upload-archive: monitor child communication more carefully.

Franck noticed that the code around polling and relaying messages
from the child process was quite bogus.  Here is an attempt to
clean it up a bit, based on his patch:

 - When POLLHUP is set, it goes ahead and reads the file
   descriptor.  Worse yet, it does not check the return value of
   read() for errors when it does.

 - When we processed one POLLIN, we should just go back and see
   if any more data is available.  We can check if the child is
   still there when poll gave control back at us but without any
   actual input.

[jc: with simplification suggested by Franck. ]

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix space in string " false" problem in "trace.c".
Christian Couder [Tue, 12 Sep 2006 04:43:08 +0000 (06:43 +0200)]
Fix space in string " false" problem in "trace.c".

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoconnect.c: finish_connect(): allow null pid parameter
Franck Bui-Huu [Tue, 12 Sep 2006 09:00:13 +0000 (11:00 +0200)]
connect.c: finish_connect(): allow null pid parameter

git_connect() can return 0 if we use git protocol for example.
Users of this function don't know and don't care if a process
had been created or not, and to avoid them to check it before
calling finish_connect() this patch allows finish_connect() to
take a null pid. And in that case return 0.

[jc: updated function signature of git_connect() with a comment on
 its return value. ]

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix a memory leak in "connect.c" and die if command too long.
Christian Couder [Mon, 11 Sep 2006 05:04:50 +0000 (07:04 +0200)]
Fix a memory leak in "connect.c" and die if command too long.

Use "add_to_string" instead of "sq_quote" and "snprintf", so
that there is no memory allocation and no memory leak.
Also check if the command is too long to fit into the buffer
and die if this is the case, instead of truncating it to the
buffer size.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMove add_to_string to "quote.c" and make it extern.
Christian Couder [Mon, 11 Sep 2006 04:59:22 +0000 (06:59 +0200)]
Move add_to_string to "quote.c" and make it extern.

So that this function may be used in places other than "rsh.c".

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Paginate history output
Jakub Narebski [Sun, 10 Sep 2006 22:36:04 +0000 (00:36 +0200)]
gitweb: Paginate history output

git_history output is now divided into pages, like git_shortlog,
git_tags and git_heads output. As whole git-rev-list output is now
read into array before writing anything, it allows for better
signaling of errors.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Make pickaxe search a feature
Jakub Narebski [Sun, 10 Sep 2006 22:29:27 +0000 (00:29 +0200)]
gitweb: Make pickaxe search a feature

As pickaxe search (selected using undocumented 'pickaxe:' operator in
search query) is resource consuming, allow to turn it on/off using
feature meachanism.  Turned on by default, for historical reasons.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd sideband status report to git-archive protocol
Junio C Hamano [Sun, 10 Sep 2006 10:33:34 +0000 (03:33 -0700)]
Add sideband status report to git-archive protocol

Using the refactored sideband code from existing upload-pack protocol,
this lets the error condition and status output sent from the remote
process to be shown locally.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/sideband' into jc/archive
Junio C Hamano [Mon, 11 Sep 2006 00:58:45 +0000 (17:58 -0700)]
Merge branch 'jc/sideband' into jc/archive

* jc/sideband:
  Prepare larger packet buffer for upload-pack protocol.
  Move sideband server side support into reusable form.
  Move sideband client side support into reusable form.
  get_sha1_hex() micro-optimization

17 years agoPrepare larger packet buffer for upload-pack protocol.
Junio C Hamano [Sun, 10 Sep 2006 23:27:08 +0000 (16:27 -0700)]
Prepare larger packet buffer for upload-pack protocol.

The original side-band support added to the upload-pack protocol used the
default 1000-byte packet length.  The pkt-line format allows up to 64k, so
prepare the receiver for the maximum size, and have the uploader and
downloader negotiate if larger packet length is allowed.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach --exec to git-archive --remote
Junio C Hamano [Sun, 10 Sep 2006 11:02:57 +0000 (04:02 -0700)]
Teach --exec to git-archive --remote

Some people needed --exec to specify the location of the upload-pack
executable, because their default SSH log-in does not include the
directory they have their own private copy of git on the $PATH.
These people need to be able to say --exec to git-archive --remote
for the same reason.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd --verbose to git-archive
Junio C Hamano [Sun, 10 Sep 2006 05:42:02 +0000 (22:42 -0700)]
Add --verbose to git-archive

And teach backends about it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
(cherry picked from 9e2c44a2893ae90944a0b7c9f40a9d22b759b5c0 commit)

17 years agoarchive: force line buffered output to stderr
Junio C Hamano [Sun, 10 Sep 2006 11:16:39 +0000 (04:16 -0700)]
archive: force line buffered output to stderr

Otherwise the remote notification that comes with -v option can get
clumped together.

Signed-off-by: Junio C Hamano <junkio@cox.net>
(cherry picked from a675cda60ead41f439b04bc69e0f19ace04e59d3 commit)

17 years agoUse xstrdup instead of strdup in builtin-{tar,zip}-tree.c
Rene Scharfe [Sun, 10 Sep 2006 16:10:01 +0000 (18:10 +0200)]
Use xstrdup instead of strdup in builtin-{tar,zip}-tree.c

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
(cherry picked from 5d2aea4cb383a43e40d47ab69d8ad7a495df6ea2 commit)

17 years agoMove sideband server side support into reusable form.
Junio C Hamano [Sun, 10 Sep 2006 10:20:24 +0000 (03:20 -0700)]
Move sideband server side support into reusable form.

The server side support; this is just the very low level, and the
caller needs to know which band it wants to send things out.

Signed-off-by: Junio C Hamano <junkio@cox.net>
(cherry picked from b786552b67878c7780c50def4c069d46dc54efbe commit)

17 years agoMove sideband client side support into reusable form.
Junio C Hamano [Sun, 10 Sep 2006 08:06:33 +0000 (01:06 -0700)]
Move sideband client side support into reusable form.

This moves the receiver side of the sideband support from
fetch-clone.c to sideband.c and its header file, so that
archiver protocol can use it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoarchive: allow remote to have more formats than we understand.
Junio C Hamano [Sun, 10 Sep 2006 06:48:03 +0000 (23:48 -0700)]
archive: allow remote to have more formats than we understand.

This fixes git-archive --remote not to parse archiver arguments;
otherwise if the remote end implements formats other than the
one known locally we will not be able to access that format.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoget_sha1_hex() micro-optimization
Junio C Hamano [Sun, 10 Sep 2006 05:21:27 +0000 (22:21 -0700)]
get_sha1_hex() micro-optimization

The function appeared high on a gprof output for a rev-list run of
a non-trivial size, and it was an obvious low-hanging fruit.

The code is from Linus.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-archive: make compression level of ZIP archives configurable
Rene Scharfe [Sat, 9 Sep 2006 15:02:38 +0000 (17:02 +0200)]
git-archive: make compression level of ZIP archives configurable

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd git-upload-archive
Franck Bui-Huu [Thu, 7 Sep 2006 13:12:05 +0000 (15:12 +0200)]
Add git-upload-archive

This command implements the git archive protocol on the server
side. This command is not intended to be used by the end user.
Underlying git-archive command line options are sent over the
protocol from "git-archive --remote=...", just like upload-tar
currently does with "git-tar-tree=...".

As for "git-archive" command implementation, this new command
does not execute any existing "git-{tar,zip}-tree" but rely
on the archive API defined by "git-archive" patch. Hence we
get 2 good points:

 - "git-archive" and "git-upload-archive" share all option
   parsing code.

 - All kind of git-upload-{tar,zip} can be deprecated.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-archive: wire up ZIP format.
Franck Bui-Huu [Thu, 7 Sep 2006 13:12:04 +0000 (15:12 +0200)]
git-archive: wire up ZIP format.

Again, this is based on Rene Scharfe's earlier patch, but uses
the archiver support introduced by the previous patch.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Acked-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-archive: wire up TAR format.
Franck Bui-Huu [Thu, 7 Sep 2006 13:12:03 +0000 (15:12 +0200)]
git-archive: wire up TAR format.

This is based on Rene Scharfe's earlier patch, but uses the
archiver support introduced by the previous patch.

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Acked-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoAdd git-archive
Franck Bui-Huu [Thu, 7 Sep 2006 13:12:02 +0000 (15:12 +0200)]
Add git-archive

git-archive is a command to make TAR and ZIP archives of a git tree.
It helps prevent a proliferation of git-{format}-tree commands.

Instead of directly calling git-{tar,zip}-tree command, it defines
a very simple API, that archiver should implement and register in
"git-archive.c". This API is made up by 2 functions whose prototype
is defined in "archive.h" file.

 - The first one is used to parse 'extra' parameters which have
   signification only for the specific archiver. That would allow
   different archive backends to have different kind of options.

 - The second one is used to ask to an archive backend to build
   the archive given some already resolved parameters.

The main reason for making this API is to avoid using
git-{tar,zip}-tree commands, hence making them useless. Maybe it's
time for them to die ?

It also implements remote operations by defining a very simple
protocol: it first sends the name of the specific uploader followed
the repository name (git-upload-tar git://example.org/repo.git).
Then it sends options. It's done by sending a sequence of one
argument per packet, with prefix "argument ", followed by a flush.

The remote protocol is implemented in "git-archive.c" for client
side and is triggered by "--remote=<repo>" option. For example,
to fetch a TAR archive in a remote repo, you can issue:

$ git archive --format=tar --remote=git://xxx/yyy/zzz.git HEAD

We choose to not make a new command "git-fetch-archive" for example,
avoind one more GIT command which should be nice for users (less
commands to remember, keeps existing --remote option).

Signed-off-by: Franck Bui-Huu <vagabon.xyz@gmail.com>
Acked-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogit-commit.sh: convert run_status to a C builtin
Jeff King [Fri, 8 Sep 2006 08:05:34 +0000 (04:05 -0400)]
git-commit.sh: convert run_status to a C builtin

This creates a new git-runstatus which should do roughly the same thing
as the run_status function from git-commit.sh. Except for color support,
the main focus has been to keep the output identical, so that it can be
verified as correct and then used as a C platform for other improvements to
the status printing code.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMove color option parsing out of diff.c and into color.[ch]
Jeff King [Fri, 8 Sep 2006 08:03:18 +0000 (04:03 -0400)]
Move color option parsing out of diff.c and into color.[ch]

The intent is to lib-ify colorizing code so it can be reused.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agodiff: support custom callbacks for output
Jeff King [Thu, 7 Sep 2006 06:35:42 +0000 (02:35 -0400)]
diff: support custom callbacks for output

Users can request the DIFF_FORMAT_CALLBACK output format to get a callback
consisting of the whole diff_queue_struct.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Add config.cache to .gitignore
Jakub Narebski [Thu, 7 Sep 2006 12:30:06 +0000 (14:30 +0200)]
autoconf: Add config.cache to .gitignore

Add generated file config.cache (default cache file, when running
./configure with -C, --config-cache option) to the list of ignored
files.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Add support for setting NO_ICONV and ICONVDIR
Jakub Narebski [Thu, 7 Sep 2006 11:48:49 +0000 (13:48 +0200)]
autoconf: Add support for setting NO_ICONV and ICONVDIR

Add support for ./configure options --without-iconv (if neither libc
nor libiconv properly support iconv), and for --with-iconv=PATH (to
set prefix to libiconv library and headers, used only when
NEED_LIBICONV is set).  While at it, make ./configure set or unset
NO_ICONV always (it is not autodetected in Makefile).

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Set NO_ICONV if iconv is found neither in libc, nor in libiconv
Jakub Narebski [Thu, 7 Sep 2006 11:48:08 +0000 (13:48 +0200)]
autoconf: Set NO_ICONV if iconv is found neither in libc, nor in libiconv

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agodiff --binary generates full index on binary files.
Junio C Hamano [Thu, 7 Sep 2006 07:54:22 +0000 (00:54 -0700)]
diff --binary generates full index on binary files.

... without --full-index.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMake apply --binary a no-op.
Junio C Hamano [Thu, 7 Sep 2006 05:45:21 +0000 (22:45 -0700)]
Make apply --binary a no-op.

Historically we did not allow binary patch applied without an
explicit permission from the user, and this flag was the way to
do so.  This makes the flag a no-op by always allowing binary
patch application.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agosend-pack: switch to using git-rev-list --stdin
Andy Whitcroft [Tue, 5 Sep 2006 21:52:12 +0000 (22:52 +0100)]
send-pack: switch to using git-rev-list --stdin

When we are generating packs to update remote repositories we
want to supply as much information as possible about the revisions
that already exist to rev-list in order optimise the pack as much
as possible.  We need to pass two revisions for each branch we are
updating in the remote repository and one for each additional branch.
Where the remote repository has numerous branches we can run out
of command line space to pass them.

Utilise the git-rev-list --stdin mode to allow unlimited numbers
of revision constraints.  This allows us to move back to the much
simpler unordered revision selection code.

[jc: added some comments in the code to describe the pipe flow
 a bit.]

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/daemon'
Junio C Hamano [Thu, 7 Sep 2006 09:37:18 +0000 (02:37 -0700)]
Merge branch 'jc/daemon'

* jc/daemon:
  Revert "daemon: add upload-tar service."
  multi-service daemon: documentation
  daemon: add upload-tar service.

17 years agoRevert "daemon: add upload-tar service."
Junio C Hamano [Thu, 7 Sep 2006 08:40:04 +0000 (01:40 -0700)]
Revert "daemon: add upload-tar service."

This reverts parts of commit 74c0cc2 and part of commit 355f541.

Franck and Rene are working on a unified upload-archive which
would supersede this when done, so better not to get in their
way.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'pm/diff'
Junio C Hamano [Thu, 7 Sep 2006 08:33:45 +0000 (01:33 -0700)]
Merge branch 'pm/diff'

* pm/diff:
  diff-index --cc shows a 3-way diff between HEAD, index and working tree.

17 years agogit-repack: create new packs inside $GIT_DIR, not cwd
Martin Langhoff [Mon, 4 Sep 2006 05:42:32 +0000 (17:42 +1200)]
git-repack: create new packs inside $GIT_DIR, not cwd

Avoid failing when cwd is !writable by writing the
packfiles in $GIT_DIR, which is more in line with other commands.

Without this, git-repack was failing when run from crontab
by non-root user accounts. For large repositories, this
also makes the mv operation a lot cheaper, and avoids leaving
temp packfiles around the fs upon failure.

Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTeach rev-list an option to read revs from the standard input.
Junio C Hamano [Wed, 6 Sep 2006 04:39:02 +0000 (21:39 -0700)]
Teach rev-list an option to read revs from the standard input.

When --stdin option is given, in addition to the <rev>s listed
on the command line, the command can read one rev parameter per
line from the standard input.  The list of revs ends at the
first empty line or EOF.

Note that you still have to give all the flags from the command
line; only rev arguments (including A..B, A...B, and A^@ notations)
can be give from the standard input.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agorevision.c: allow injecting revision parameters after setup_revisions().
Junio C Hamano [Wed, 6 Sep 2006 04:28:36 +0000 (21:28 -0700)]
revision.c: allow injecting revision parameters after setup_revisions().

setup_revisions() wants to get all the parameters at once and
then postprocesses the resulting revs structure after it is done
with them.  This code structure is a bit cumbersome to deal with
efficiently when we want to inject revision parameters from the
side (e.g. read from standard input).

Fortunately, the nature of this postprocessing is not affected by
revision parameters; they are affected only by flags.  So it is
Ok to do add_object() after the it returns.

This splits out the code that deals with the revision parameter
out of the main loop of setup_revisions(), so that we can later
call it from elsewhere after it returns.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix memory leak in prepend_to_path (git.c).
Christian Couder [Tue, 5 Sep 2006 06:22:16 +0000 (08:22 +0200)]
Fix memory leak in prepend_to_path (git.c).

Some memory was allocated for a new path but not freed
after the path was used.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agosend-pack: remove remote reference limit
Andy Whitcroft [Tue, 5 Sep 2006 19:00:17 +0000 (20:00 +0100)]
send-pack: remove remote reference limit

When build a pack for a push we query the remote copy for existant
heads.  These are used to prune unnecessary objects from the pack.
As we receive the remote references in get_remote_heads() we validate
the reference names via check_ref() which includes a length check;
rejecting those >45 characters in size.

This is a miss converted change, it was originally designed to reject
messages which were less than 45 characters in length (a 40 character
sha1 and refs/) to prevent comparing unitialised memory.  check_ref()
now gets the raw length so check for at least 5 characters.

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Fix copy'n'paste error
Jakub Narebski [Tue, 5 Sep 2006 20:03:48 +0000 (22:03 +0200)]
autoconf: Fix copy'n'paste error

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agodiff-index --cc shows a 3-way diff between HEAD, index and working tree.
Paul Mackerras [Mon, 4 Sep 2006 11:38:40 +0000 (21:38 +1000)]
diff-index --cc shows a 3-way diff between HEAD, index and working tree.

This implements a 3-way diff between the HEAD commit, the state in the
index, and the working directory.  This is like the n-way diff for a
merge, and uses much of the same code.  It is invoked with the -c flag
to git-diff-index, which it already accepted and did nothing with.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'jc/pack'
Junio C Hamano [Tue, 5 Sep 2006 01:57:35 +0000 (18:57 -0700)]
Merge branch 'jc/pack'

* jc/pack:
  more lightweight revalidation while reusing deflated stream in packing
  pack-objects: fix thinko in revalidate code
  pack-objects: re-validate data we copy from elsewhere.

17 years agoautoconf: Quote AC_CACHE_CHECK arguments
Jakub Narebski [Mon, 4 Sep 2006 22:58:25 +0000 (00:58 +0200)]
autoconf: Quote AC_CACHE_CHECK arguments

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Check for subprocess.py
Jakub Narebski [Mon, 4 Sep 2006 22:57:45 +0000 (00:57 +0200)]
autoconf: Check for subprocess.py

Add custom test for checking if Python comes with subprocess.py, or
should we use our own subprocess.py by defining WITH_OWN_SUBPROCESS_PY.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoautoconf: Add -liconv to LIBS when NEEDS_LIBICONV
Jakub Narebski [Mon, 4 Sep 2006 22:55:52 +0000 (00:55 +0200)]
autoconf: Add -liconv to LIBS when NEEDS_LIBICONV

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Divide page path into directories -- path's "breadcrumbs"
Jakub Narebski [Mon, 4 Sep 2006 16:17:58 +0000 (18:17 +0200)]
gitweb: Divide page path into directories -- path's "breadcrumbs"

Divide page path into directories, so that each part of path links to
the "tree" view of the $hash_base (or HEAD, if $hash_base is not set)
version of the directory.

If the entity is blob, final part (basename) links to $hash_base or
HEAD revision of the "raw" blob ("blob_plain" view).  If the entity is
tree, link to the "tree" view.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Correct typo: '==' instead of 'eq' in git_difftree_body
Jakub Narebski [Mon, 4 Sep 2006 16:19:58 +0000 (18:19 +0200)]
gitweb: Correct typo: '==' instead of 'eq' in git_difftree_body

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Add GIT favicon, assuming image/png type
Jakub Narebski [Mon, 4 Sep 2006 18:32:13 +0000 (20:32 +0200)]
gitweb: Add GIT favicon, assuming image/png type

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoFix git-fsck-objects SIGSEGV/divide-by-zero
Linus Torvalds [Mon, 4 Sep 2006 15:34:12 +0000 (08:34 -0700)]
Fix git-fsck-objects SIGSEGV/divide-by-zero

If you try to fsck a repository that isn't entirely empty, but that has no
inter-object references (ie all the objects are blobs, and don't refer to
anything else), git-fsck-objects currently fails.

This probably cannot happen in practice, but can be tested with something
like

git init-db
touch dummy
git add dummy
git fsck-objects

where the fsck will die by a divide-by-zero when it tries to look up the
references from the one object it found (hash_obj() will do a modulus by
refs_hash_size).

On some other archiectures (ppc, sparc) the divide-by-zero will go
unnoticed, and we'll instead SIGSEGV when we hit the "refs_hash[j]"
access.

So move the test that should protect against this from mark_reachable()
into lookup_object_refs(), which incidentally in the process also fixes
mark_reachable() itself (it used to not mark the one object that _was_
reachable, because it decided that it had no refs too early).

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agounpack-objects desperately salvages objects from a corrupt pack
Junio C Hamano [Mon, 4 Sep 2006 05:55:54 +0000 (22:55 -0700)]
unpack-objects desperately salvages objects from a corrupt pack

The command unpack-objects dies upon the first error.  This is
probably considered a feature -- if a pack is corrupt, instead
of trying to extract from it and possibly risking to contaminate
a good repository with objects whose validity is dubious, we
should seek a good copy of the pack and retry.  However, we may
not have any good copy anywhere.  This implements the last
resort effort to extract what are salvageable from such a
corrupt pack.

This flag might have helped Sergio when recovering from a
corrupt pack.  In my test, it managed to salvage 247 objects out
of a pack that had 251 objects but without it the command
stopped after extracting 73 objects.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agomore lightweight revalidation while reusing deflated stream in packing
Junio C Hamano [Mon, 4 Sep 2006 04:09:18 +0000 (21:09 -0700)]
more lightweight revalidation while reusing deflated stream in packing

When copying from an existing pack and when copying from a loose
object with new style header, the code makes sure that the piece
we are going to copy out inflates well and inflate() consumes
the data in full while doing so.

The check to see if the xdelta really apply is quite expensive
as you described, because you would need to have the image of
the base object which can be represented as a delta against
something else.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agogitweb: Change the name of diff to parent link in "commit" view to "diff
Jakub Narebski [Sun, 3 Sep 2006 21:43:03 +0000 (23:43 +0200)]
gitweb: Change the name of diff to parent link in "commit" view to "diff

Change the name of diff to parent (current commit to one of parents)
link in "commit" view (git_commit subroutine) from "commitdiff" to
"diff".  Let's leave "commitdiff" for equivalent of git-show, or
git-diff-tree with one revision, i.e. diff for a given commit to its
parent (parents).

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agopack-objects: fix thinko in revalidate code
Junio C Hamano [Sun, 3 Sep 2006 21:44:46 +0000 (14:44 -0700)]
pack-objects: fix thinko in revalidate code

When revalidating an entry from an existing pack entry->size and
entry->type are not necessarily the size of the final object
when the entry is deltified, but for base objects they must
match.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoUpdate GIT_TRACE documentation.
Christian Couder [Sun, 3 Sep 2006 15:32:24 +0000 (17:32 +0200)]
Update GIT_TRACE documentation.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoMerge branch 'master' into cc/trace
Junio C Hamano [Sun, 3 Sep 2006 00:10:16 +0000 (17:10 -0700)]
Merge branch 'master' into cc/trace

* master:
  Trace into a file or an open fd and refactor tracing code.
  Replace uses of strdup with xstrdup.
  consolidate two copies of new style object header parsing code.
  Documentation: Fix howto/revert-branch-rebase.html generation
  fmt-merge-msg: fix off-by-one bug
  git-rev-list(1): group options; reformat; document more options
  Constness tightening for move/link_temp_to_file()
  gitweb: Fix git_blame
  Include config.mak.autogen in the doc Makefile
  Use xmalloc instead of malloc
  git(7): move gitk(1) to the list of porcelain commands
  gitk: Fix some bugs in the new cherry-picking code
  gitk: Improve responsiveness while reading and layout out the graph
  gitk: Update preceding/following tag info when creating a tag
  gitk: Add a menu item for cherry-picking commits
  gitk: Fix a couple of buglets in the branch head menu items
  gitk: Add a context menu for heads
  gitk: Add a row context-menu item for creating a new branch
  gitk: Recompute ancestor/descendent heads/tags when rereading refs
  gitk: Minor cleanups

17 years agopack-objects: re-validate data we copy from elsewhere.
Junio C Hamano [Fri, 1 Sep 2006 22:05:12 +0000 (15:05 -0700)]
pack-objects: re-validate data we copy from elsewhere.

When reusing data from an existing pack and from a new style
loose objects, we used to just copy it staight into the
resulting pack.  Instead make sure they are not corrupt, but
do so only when we are not streaming to stdout, in which case
the receiving end will do the validation either by unpacking
the stream or by constructing the .idx file.

Signed-off-by: Junio C Hamano <junkio@cox.net>
17 years agoTrace into a file or an open fd and refactor tracing code.
Christian Couder [Sat, 2 Sep 2006 16:23:48 +0000 (18:23 +0200)]
Trace into a file or an open fd and refactor tracing code.

If GIT_TRACE is set to an absolute path (starting with a
'/' character), we interpret this as a file path and we
trace into it.

Also if GIT_TRACE is set to an integer value greater than
1 and lower than 10, we interpret this as an open fd value
and we trace into it.

Note that this behavior is not compatible with the
previous one.

We also trace whole messages using one write(2) call to
make sure messages from processes do net get mixed up in
the middle.

This patch makes it possible to get trace information when
running "make test".

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>