Code

git.git
18 years agoMerge branch 'js/lsfix'
Junio C Hamano [Thu, 22 Jun 2006 17:35:13 +0000 (10:35 -0700)]
Merge branch 'js/lsfix'

* js/lsfix:
  Initialize lock_file struct to all zero.
  Make git-update-ref a builtin
  Make git-update-index a builtin
  Make git-stripspace a builtin
  Make git-mailinfo a builtin
  Make git-mailsplit a builtin
  Make git-write-tree a builtin

18 years agoMerge branch 'ew/rebase'
Junio C Hamano [Thu, 22 Jun 2006 17:34:02 +0000 (10:34 -0700)]
Merge branch 'ew/rebase'

* ew/rebase:
  rebase --merge: fix for rebasing more than 7 commits.
  rebase: error out for NO_PYTHON if they use recursive merge
  Add renaming-rebase test.
  rebase: Allow merge strategies to be used when rebasing

18 years agoMerge branch 'jn/web'
Junio C Hamano [Thu, 22 Jun 2006 17:33:34 +0000 (10:33 -0700)]
Merge branch 'jn/web'

* jn/web:
  gitweb: whitespace cleanup around '='
  gitweb: Use $hash_base as $search_hash if possible
  gitweb: Make use of $PATH_INFO for project parameter
  Move $gitbin earlier in gitweb.cgi
  Add git version to gitweb output
  gitweb: whitespace cleanup
  gitweb: style done with stylesheet
  gitweb: A couple of page title tweaking
  Fix: Support for the standard mime.types map in gitweb
  gitweb: add type="text/css" to stylesheet link
  Make CSS file gitweb/gitweb.css more readable
  Fix gitweb stylesheet
  Support for the standard mime.types map in gitweb
  gitweb: text files for 'blob_plain' action without charset by default
  gitweb: safely output binary files for 'blob_plain' action
  Move gitweb style to gitweb.css

18 years agoMerge early parts of branch 'js/diff'
Junio C Hamano [Thu, 22 Jun 2006 17:32:58 +0000 (10:32 -0700)]
Merge early parts of branch 'js/diff'

18 years agoMerge branch 'jc/upload-corrupt'
Junio C Hamano [Thu, 22 Jun 2006 17:25:51 +0000 (10:25 -0700)]
Merge branch 'jc/upload-corrupt'

* jc/upload-corrupt:
  daemon: send stderr to /dev/null instead of closing.
  upload-pack/fetch-pack: support side-band communication
  Retire git-clone-pack
  upload-pack: prepare for sideband message support.
  upload-pack: avoid sending an incomplete pack upon failure

18 years agogitweb: whitespace cleanup around '='
Jakub Narebski [Thu, 22 Jun 2006 06:52:57 +0000 (08:52 +0200)]
gitweb: whitespace cleanup around '='

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Use $hash_base as $search_hash if possible
Timo Hirvonen [Tue, 20 Jun 2006 13:41:05 +0000 (16:41 +0300)]
gitweb: Use $hash_base as $search_hash if possible

$hash (h parameter) does not always point to a commit. Use $hash_base as
$search_hash when it is defined.

Signed-off-by: Timo Hirvonen <tihirvon@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'jc/waitpid'
Junio C Hamano [Thu, 22 Jun 2006 09:19:08 +0000 (02:19 -0700)]
Merge branch 'jc/waitpid'

* jc/waitpid:
  Restore SIGCHLD to SIG_DFL where we care about waitpid().

18 years agoMerge branch 'ff/c99'
Junio C Hamano [Thu, 22 Jun 2006 09:18:51 +0000 (02:18 -0700)]
Merge branch 'ff/c99'

* ff/c99:
  Remove all void-pointer arithmetic.
  Change types used in bitfields to be `int's.
  Don't use empty structure initializers.
  Cast pointers to `void *' when used in a format.
  Don't instantiate structures with FAMs.
  Initialize FAMs using `FLEX_ARRAY'.
  Remove ranges from switch statements.

18 years agoMerge branch 'pb/config'
Junio C Hamano [Thu, 22 Jun 2006 09:15:45 +0000 (02:15 -0700)]
Merge branch 'pb/config'

* pb/config:
  git_config: access() returns 0 on success, not > 0
  repo-config: Fix late-night bug
  Read configuration also from $HOME/.gitconfig
  Fix setting config variables with an alternative GIT_CONFIG
  Support for extracting configuration from different files

18 years agoMerge branch 'lt/objlist'
Junio C Hamano [Thu, 22 Jun 2006 09:15:22 +0000 (02:15 -0700)]
Merge branch 'lt/objlist'

* lt/objlist:
  Add "named object array" concept

18 years agogit-svn: fix commit --edit flag when using SVN:: libraries
Eric Wong [Thu, 22 Jun 2006 08:22:46 +0000 (01:22 -0700)]
git-svn: fix commit --edit flag when using SVN:: libraries

Trying to open an interactive editor in the console while stdout is
being piped to the parent process doesn't work out very well.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMakefile: do not force unneeded recompilation upon GIT_VERSION changes
Junio C Hamano [Thu, 22 Jun 2006 09:04:27 +0000 (02:04 -0700)]
Makefile: do not force unneeded recompilation upon GIT_VERSION changes

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorebase --merge: fix for rebasing more than 7 commits.
Junio C Hamano [Thu, 22 Jun 2006 08:44:54 +0000 (01:44 -0700)]
rebase --merge: fix for rebasing more than 7 commits.

Instead of using 4-digit numbers to name commits being rebased,
just use "cmt.$msgnum" string, with $msgnum as a decimal number
without leading zero padding.  This makes it possible to rebase
more than 9999 commits, but of more practical importance is that
the earlier code used "printf" to format already formatted
$msgnum and barfed when it counted up to 0008.  In other words,
the old code was incapable of rebasing more than 7 commits, and
this fixes that problem.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoCheck and document the options to prevent mistakes.
Eric W. Biederman [Wed, 21 Jun 2006 13:17:31 +0000 (07:17 -0600)]
Check and document the options to prevent mistakes.

When multiple recipients are given to git-send-email on the same
--cc line the code does not properly handle it.

Full and proper parsing of the email addresses so I can detect
which commas mean a new email address is more than I care to implement.

In particular this email address: "bibo,mao" <bibo.mao@intel.com>
must not be treated as two email addresses.

So this patch simply treats all commas in recipient lists as
an error and fails if one is given.

At the same time it documents that git-send-email wants multiple
instances of --cc specified on the command line if you want to
cc multiple recipients.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoPass -DDEFAULT_GIT_TEMPLATE_DIR only where actually used.
Yakov Lerner [Thu, 22 Jun 2006 01:47:00 +0000 (04:47 +0300)]
Pass -DDEFAULT_GIT_TEMPLATE_DIR only where actually used.

Before this patch, -DDEFAULT_GIT_TEMPLATE_DIR was passed on compilation
command line to all and every .c file compiled. In fact the macro
is used by only one .c file, and unused by all other .c files.
Remove -DDEFAULT_GIT_TEMPLATE_DIR where unused. Follow the example of
exec_cmd.o. Pass -DDEFAULT_GIT_TEMPLATE_DIR only where actually used.

Signed-off-by: Yakov Lerner <iler.ml@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agodaemon: send stderr to /dev/null instead of closing.
Junio C Hamano [Wed, 21 Jun 2006 23:37:48 +0000 (16:37 -0700)]
daemon: send stderr to /dev/null instead of closing.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Make use of $PATH_INFO for project parameter
Jakub Narebski [Wed, 21 Jun 2006 15:06:39 +0000 (17:06 +0200)]
gitweb: Make use of $PATH_INFO for project parameter

Allow to have project name in the path part of URL, just after the name of
script. For example instead of gitweb.cgi?p=git.git you can write
gitweb.cgi/git.git or gitweb.cgi/git.git/

Not used in URLs inside gitweb; it means that the above alternate syntax
must be generated by hand, at least for now.

Side effect: project name parameter is now stripped of leading and
trailing slash before validation.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix grow_refs_hash()
Linus Torvalds [Wed, 21 Jun 2006 18:01:12 +0000 (11:01 -0700)]
Fix grow_refs_hash()

Earlier commit 3e4339e6f96e8c4f38a9c6607b98d3e96a2ed783 had a
thinko that did not check for collisions while repopulating the
objects in the new hash table.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorebase: error out for NO_PYTHON if they use recursive merge
Eric Wong [Wed, 21 Jun 2006 10:04:42 +0000 (03:04 -0700)]
rebase: error out for NO_PYTHON if they use recursive merge

recursive merge relies on Python, and we can't perform
rename-aware merges without the recursive merge.  So bail out
before trying it.

The test won't work w/o recursive merge, either, so skip that,
too.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd renaming-rebase test.
Junio C Hamano [Sun, 18 Jun 2006 10:33:12 +0000 (03:33 -0700)]
Add renaming-rebase test.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorebase: Allow merge strategies to be used when rebasing
Eric Wong [Wed, 21 Jun 2006 10:04:41 +0000 (03:04 -0700)]
rebase: Allow merge strategies to be used when rebasing

This solves the problem of rebasing local commits against an
upstream that has renamed files.

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoobject-refs: avoid division by zero
Andre Noll [Wed, 21 Jun 2006 02:46:05 +0000 (04:46 +0200)]
object-refs: avoid division by zero

Currently, we don't check refs_hash_size size and happily call
lookup_object_refs() even if refs_hash_size is zero which leads to
a division by zero in hash_obj().

Signed-off-by: Andre Noll <maan@systemlinux.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupload-pack/fetch-pack: support side-band communication
Junio C Hamano [Wed, 21 Jun 2006 07:30:21 +0000 (00:30 -0700)]
upload-pack/fetch-pack: support side-band communication

This implements a protocol extension between fetch-pack and
upload-pack to allow stderr stream from upload-pack (primarily
used for the progress bar display) to be passed back.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRetire git-clone-pack
Junio C Hamano [Wed, 21 Jun 2006 06:54:26 +0000 (23:54 -0700)]
Retire git-clone-pack

The program is not used by git-clone since git-fetch-pack was extended
to allow its caller do what git-clone-pack alone did, and git-clone was
updated to use it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupload-pack: prepare for sideband message support.
Junio C Hamano [Wed, 21 Jun 2006 05:48:23 +0000 (22:48 -0700)]
upload-pack: prepare for sideband message support.

This does not implement sideband for propagating the status to
the downloader yet, but add code to capture the standard error
output from the pack-objects process in preparation for sending
it off to the client when the protocol extension allows us to do
so.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoupload-pack: avoid sending an incomplete pack upon failure
Junio C Hamano [Wed, 21 Jun 2006 01:26:34 +0000 (18:26 -0700)]
upload-pack: avoid sending an incomplete pack upon failure

When the repository on the remote side is corrupted, rev-list
spawned from upload-pack would die with error, but pack-objects
that reads from the rev-list happily created a packfile that can
be unpacked by the downloader.  When this happens, the resulting
packfile is not corrupted and unpacks cleanly, but the list of
the objects contained in it is not what the protocol exchange
computed.

This update makes upload-pack to monitor its subprocesses, and
when either of them dies with error, sends an incomplete pack
data to the downloader to cause it to fail.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix possible out-of-bounds array access
Uwe Zeisberger [Wed, 21 Jun 2006 09:04:12 +0000 (11:04 +0200)]
Fix possible out-of-bounds array access

If match is "", match[-1] is accessed.  Let pathspec_matches return 1 in that
case indicating that "" matches everything.

Incidently this fixes git-grep'ing in ".".

Signed-off-by: Uwe Zeisberger <Uwe_Zeisberger@digi.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMove $gitbin earlier in gitweb.cgi
Jakub Narebski [Wed, 21 Jun 2006 07:48:04 +0000 (09:48 +0200)]
Move $gitbin earlier in gitweb.cgi

(cherry picked from 9dca843086356b964f27d8fabe1e3c48074a9f02 commit)

Signed-off-by: Jakub Narebski <jnareb.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd git version to gitweb output
Jakub Narebski [Wed, 21 Jun 2006 07:48:03 +0000 (09:48 +0200)]
Add git version to gitweb output

Add git-core binaries used version as the comment at the beginning of HTML
output, just below the comment with version of git web interface version.

Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: whitespace cleanup
Jakub Narebski [Wed, 21 Jun 2006 07:48:02 +0000 (09:48 +0200)]
gitweb: whitespace cleanup

Do not use tabs to align variable initialization (actually use
tabs only at the beginning of line, for code indent).  Remove trailing
whitespace.  Make whitespace usage more consistent.

Signed-off-by: Jakub Narebski <jnareb.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agocheckout -f: do not leave untracked working tree files.
Junio C Hamano [Tue, 20 Jun 2006 09:47:14 +0000 (02:47 -0700)]
checkout -f: do not leave untracked working tree files.

Earlier we did not consider untracked working tree files
"precious", but we have always considered them fair game to
clobber.  These days, branch switching by read-tree is more
careful and tries to protect untracked working tree files.  This
caused the following workflow to stop working:

git checkout one-branch-with-file-F
git checkout -f another-without-file-F
git pull . one-branch-with-file-F

Because the second checkout leaves F from the previous state as
untracked file in the working tree, the merge would fail, trying
to protect F from being clobbered.

This changes "git checkout -f" to remove working tree files that
are known to git in the switched-from state but do not exist in
the switched-to state, borrowing the same logic from "reset --hard".

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoLog peer address when git-daemon called from inetd
David Woodhouse [Tue, 20 Jun 2006 14:38:13 +0000 (15:38 +0100)]
Log peer address when git-daemon called from inetd

When we run git-daemon from inetd, even with the --verbose option, it
doesn't log the peer address. That logic was only in the standalone
daemon code -- move it to the execute() function instead. Tested with
both IPv6 and Legacy IP clients, in both inetd and daemon mode.

Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Acked-by: Jon Loeliger <jdl@jdl.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: style done with stylesheet
Jakub Narebski [Tue, 20 Jun 2006 14:58:12 +0000 (14:58 +0000)]
gitweb: style done with stylesheet

Replace (almost) all 'style' attributes with 'class' attribute
and adding rule to CSS file. Some tables use CSS for styling
instead of legacy styling attributes.

[jc: too many rejects -- hand fixed and reindented]

18 years agogitweb: A couple of page title tweaking
Jakub Narebski [Tue, 20 Jun 2006 06:17:03 +0000 (06:17 +0000)]
gitweb: A couple of page title tweaking

[jc: the e-mailed patch did not apply, so I had to guess but I think
 I got the result right.]

18 years agoFix: Support for the standard mime.types map in gitweb
Jakub Narebski [Tue, 20 Jun 2006 06:19:31 +0000 (06:19 +0000)]
Fix: Support for the standard mime.types map in gitweb

Temporary fix: commented out offending line in mimetype_guess.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRemove all void-pointer arithmetic.
Florian Forster [Sun, 18 Jun 2006 15:18:09 +0000 (17:18 +0200)]
Remove all void-pointer arithmetic.

ANSI C99 doesn't allow void-pointer arithmetic. This patch fixes this in
various ways. Usually the strategy that required the least changes was used.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-svn: fix --rmdir when using SVN:: libraries
Eric Wong [Tue, 20 Jun 2006 00:59:35 +0000 (17:59 -0700)]
git-svn: fix --rmdir when using SVN:: libraries

When tracking directories with nearly all of its files at
the most nested levels, --rmdir would accidentally go too
far when deleting.

Of course, we'll add a test for this condition, too.

Makefile: automatically run new tests as they appear in t/

Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit_config: access() returns 0 on success, not > 0
Johannes Schindelin [Tue, 20 Jun 2006 07:51:09 +0000 (09:51 +0200)]
git_config: access() returns 0 on success, not > 0

Another late-night bug. Sorry again.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agorepo-config: Fix late-night bug
Johannes Schindelin [Tue, 20 Jun 2006 07:45:53 +0000 (09:45 +0200)]
repo-config: Fix late-night bug

This bug was hidden by the "future-proofing" of the test. Sigh.

When neither GIT_CONFIG nor GIT_CONFIG_LOCAL is set, do not use NULL,
but $GIT_DIR/config. Instead of using $GIT_DIR/config when only
GIT_CONFIG_LOCAL is set. Sorry.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: add type="text/css" to stylesheet link
Jakub Narebski [Tue, 20 Jun 2006 04:11:32 +0000 (04:11 +0000)]
gitweb: add type="text/css" to stylesheet link

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd "named object array" concept
Linus Torvalds [Tue, 20 Jun 2006 00:42:35 +0000 (17:42 -0700)]
Add "named object array" concept

We've had this notion of a "object_list" for a long time, which eventually
grew a "name" member because some users (notably git-rev-list) wanted to
name each object as it is generated.

That object_list is great for some things, but it isn't all that wonderful
for others, and the "name" member is generally not used by everybody.

This patch splits the users of the object_list array up into two: the
traditional list users, who want the list-like format, and who don't
actually use or want the name. And another class of users that really used
the list as an extensible array, and generally wanted to name the objects.

The patch is fairly straightforward, but it's also biggish. Most of it
really just cleans things up: switching the revision parsing and listing
over to the array makes things like the builtin-diff usage much simpler
(we now see exactly how many members the array has, and we don't get the
objects reversed from the order they were on the command line).

One of the main reasons for doing this at all is that the malloc overhead
of the simple object list was actually pretty high, and the array is just
a lot denser. So this patch brings down memory usage by git-rev-list by
just under 3% (on top of all the other memory use optimizations) on the
mozilla archive.

It does add more lines than it removes, and more importantly, it adds a
whole new infrastructure for maintaining lists of objects, but on the
other hand, the new dynamic array code is pretty obvious. The change to
builtin-diff-tree.c shows a fairly good example of why an array interface
is sometimes more natural, and just much simpler for everybody.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRestore SIGCHLD to SIG_DFL where we care about waitpid().
Junio C Hamano [Tue, 20 Jun 2006 01:25:21 +0000 (18:25 -0700)]
Restore SIGCHLD to SIG_DFL where we care about waitpid().

It was reported that under one implementation of socks client
"git clone" fails with "error: waitpid failed (No child processes)",
because "git" is spawned after setting SIGCHLD to SIG_IGN.

Arguably it may be a broken setting, but we should protect
ourselves so that we can get reliable results from waitpid() for
the children we care about.

This patch resets SIGCHLD to SIG_DFL in three places:

 - connect.c::git_connect() - initiators of git native
   protocol transfer are covered with this.

 - daemon.c::main() - obviously.

 - merge-index.c::main() - obviously.

There are other programs that do fork() but do not waitpid():
http-push, imap-send.  upload-pack does not either, but in the
case of that program, each of the forked halves runs exec()
another program, so this change would not have much effect
there.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoxdiff: minor changes to match libxdiff-0.21
Junio C Hamano [Tue, 20 Jun 2006 00:01:35 +0000 (17:01 -0700)]
xdiff: minor changes to match libxdiff-0.21

This reformats the change 621c53cc082299eaf69e9f2dc0274547c7d87fb0
introduced to match what upstream author implemented in libxdiff-0.21
without changing any logic (hopefully ;-).  This is to help keep
us in sync with the upstream.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agofix rfc2047 formatter.
Junio C Hamano [Mon, 19 Jun 2006 22:00:17 +0000 (15:00 -0700)]
fix rfc2047 formatter.

Running git-format-patch on patches from Lukas destroyed
the From: line.  This fixes it.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix t8001-annotate and t8002-blame for ActiveState Perl
Dennis Stosberg [Mon, 19 Jun 2006 19:40:37 +0000 (21:40 +0200)]
Fix t8001-annotate and t8002-blame for ActiveState Perl

There seems to be at least one implementation of Perl which requires the
user to specify an extension for backup files.

Reported by Alex Riesen.

Signed-off-by: Dennis Stosberg <dennis@stosberg.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoAdd specialized object allocator
Linus Torvalds [Mon, 19 Jun 2006 17:44:15 +0000 (10:44 -0700)]
Add specialized object allocator

This creates a simple specialized object allocator for basic
objects.

This avoids wasting space with malloc overhead (metadata and
extra alignment), since the specialized allocator knows the
alignment, and that objects, once allocated, are never freed.

It also allows us to track some basic statistics about object
allocations. For example, for the mozilla import, it shows
object usage as follows:

     blobs:   627629 (14710 kB)
     trees:  1119035 (34969 kB)
   commits:   196423  (8440 kB)
      tags:     1336    (46 kB)

and the simpler allocator shaves off about 2.5% off the memory
footprint off a "git-rev-list --all --objects", and is a bit
faster too.

[ Side note: this concludes the series of "save memory in object storage".
  The thing is, there simply isn't much more to be saved on the objects.

  Doing "git-rev-list --all --objects" on the mozilla archive has a final
  total RSS of 131498 pages for me: that's about 513MB. Of that, the
  object overhead is now just 56MB, the rest is going somewhere else (put
  another way: the fact that this patch shaves off 2.5% of the total
  memory overhead, considering that objects are now not much more than 10%
  of the total shows how big the wasted space really was: this makes
  object allocations much more memory- and time-efficient).

  I haven't looked at where the rest is, but I suspect the bulk of it is
  just the pack-file loading. It may be that we should pack the tree
  objects separately from the blob objects: for git-rev-list --objects, we
  don't actually ever need to even look at the blobs, but since trees and
  blobs are interspersed in the pack-file, we end up not being dense in
  the tree accesses, so we end up looking at more pages than we strictly
  need to.

  So with a 535MB pack-file, it's entirely possible - even likely - that
  most of the remaining RSS is just the mmap of the pack-file itself. We
  don't need to map in _all_ of it, but we do end up mapping a fair
  amount. ]

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRead configuration also from $HOME/.gitconfig
Johannes Schindelin [Mon, 19 Jun 2006 23:48:03 +0000 (01:48 +0200)]
Read configuration also from $HOME/.gitconfig

This patch is based on Pasky's, with three notable differences:

- I did not yet update the documentation
- I named it .gitconfig, not .gitrc
- git-repo-config does not barf when a unique key is overridden locally

The last means that if you have something like

[alias]
l = log --stat -M

in ~/.gitconfig, and

[alias]
l = log --stat -M next..

in $GIT_DIR/config, then

git-repo-config alias.l

returns only one value, namely the value from $GIT_DIR/config.

If you set the environment variable GIT_CONFIG, $HOME/.gitconfig is not
read, and neither $GIT_DIR/config, but $GIT_CONFIG instead.

If you set GIT_CONFIG_LOCAL instead, it is interpreted instead of
$GIT_DIR/config, but $HOME/.gitconfig is still read.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix setting config variables with an alternative GIT_CONFIG
Johannes Schindelin [Mon, 19 Jun 2006 22:51:58 +0000 (00:51 +0200)]
Fix setting config variables with an alternative GIT_CONFIG

When setting a config variable, git_config_set() ignored the variables
GIT_CONFIG and GIT_CONFIG_LOCAL. Now, when GIT_CONFIG_LOCAL is set, it
will write to that file. If not, GIT_CONFIG is checked, and only as a
fallback, the change is written to $GIT_DIR/config.

Add a test for it, and also future-proof the test for the upcoming
$HOME/.gitconfig support.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoInitialize lock_file struct to all zero.
Johannes Schindelin [Mon, 19 Jun 2006 22:55:20 +0000 (00:55 +0200)]
Initialize lock_file struct to all zero.

hold_lock_file_for_update() relies on that.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake CSS file gitweb/gitweb.css more readable
Jakub Narebski [Mon, 19 Jun 2006 18:27:59 +0000 (20:27 +0200)]
Make CSS file gitweb/gitweb.css more readable

Taken from git://git.xmms.se/xmms2/gitweb-xmms2.git
commit  561262030d58a6325f500b36d836dbe02a5abc68
"Make CSS readable" by Daniel Svensson, with extra
parts removed and consistent whitespace usage.

[jc: tabified the results to cleaning things up, and removed an
 added item that was commented out. ]

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-update-ref a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:22:00 +0000 (22:22 +0200)]
Make git-update-ref a builtin

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-update-index a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:21:57 +0000 (22:21 +0200)]
Make git-update-index a builtin

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-stripspace a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:21:53 +0000 (22:21 +0200)]
Make git-stripspace a builtin

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-mailinfo a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:21:50 +0000 (22:21 +0200)]
Make git-mailinfo a builtin

[jc: with a bit of constness tightening]

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-mailsplit a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:21:46 +0000 (22:21 +0200)]
Make git-mailsplit a builtin

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake git-write-tree a builtin
Lukas Sandström [Tue, 13 Jun 2006 20:21:42 +0000 (22:21 +0200)]
Make git-write-tree a builtin

Signed-off-by: Lukas Sandström <lukass@etek.chalmers.se>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix gitweb stylesheet
Jakub Narebski [Sun, 18 Jun 2006 06:45:03 +0000 (23:45 -0700)]
Fix gitweb stylesheet

An earlier commit forgot to move some piece from the CGI script
to the external stylesheet.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoSupport for the standard mime.types map in gitweb
Petr Baudis [Sat, 17 Jun 2006 22:01:06 +0000 (00:01 +0200)]
Support for the standard mime.types map in gitweb

gitweb will try to look up the filename mimetype in /etc/mime.types
and optionally a user-configured mime.types map as well.

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: text files for 'blob_plain' action without charset by default
Jakub Narebski [Sat, 17 Jun 2006 16:07:24 +0000 (18:07 +0200)]
gitweb: text files for 'blob_plain' action without charset by default

$default_text_plain_charset is undefined (no specified charset) by
default. Additionally ':raw' layer for binmode is used for outputting file
content.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: safely output binary files for 'blob_plain' action
Jakub Narebski [Sat, 17 Jun 2006 11:32:15 +0000 (13:32 +0200)]
gitweb: safely output binary files for 'blob_plain' action

gitweb tries now to output correct Content-Type header for
'blob_plain' action; for now text/plain for text files,
appropriate image MIME type for *.png, *.gif and *.jpg/*.jpeg files,
and application/octet-stream for other binary files.

Introduced new configuration variables: $default_blob_plain_mimetype
and $default_text_plain_charset (only 'utf-8' is guaranteed to work
for the latter).

binmode changed to ':raw' in git_blob_plain for output of non-text files.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMove gitweb style to gitweb.css
Jakub Narebski [Sat, 17 Jun 2006 09:23:56 +0000 (11:23 +0200)]
Move gitweb style to gitweb.css

Move gitweb style from embedded <style> element in gitweb/gitweb.cgi
to external CSS file gitweb/gitweb.css.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoChange types used in bitfields to be `int's.
Florian Forster [Sun, 18 Jun 2006 15:18:08 +0000 (17:18 +0200)]
Change types used in bitfields to be `int's.

According to ANSI C99 bitfields are only defined for `signed int' and `unsigned
int'. This patch corrects the bitfield in the `msg_data_t' type from
`imap-send.c'.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDon't use empty structure initializers.
Florian Forster [Sun, 18 Jun 2006 15:18:07 +0000 (17:18 +0200)]
Don't use empty structure initializers.

Empty initializers for structures are not allowed in ANSI C99. This patch
removes such an initializer from `builtin-read-tree.c'. Since the struct was
static (and is therefore implicitely initialized to zero anyway) it wasn't
actually needed.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoCast pointers to `void *' when used in a format.
Florian Forster [Sun, 18 Jun 2006 15:18:06 +0000 (17:18 +0200)]
Cast pointers to `void *' when used in a format.

ANSI C99 requires void-pointers when using the `%p' format. This patch adds the
neccessary cast in `blame.c'.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDon't instantiate structures with FAMs.
Florian Forster [Sun, 18 Jun 2006 15:18:05 +0000 (17:18 +0200)]
Don't instantiate structures with FAMs.

Since structures with `flexible array members' are an incomplete datatype ANSI
C99 forbids creating instances of them. This patch removes such an instance
from `diff-lib.c' and replaces it with a pointer to a `struct
combine_diff_path'. Since all neccessary memory is allocated at once the number
of calls to `xmalloc' is not increased.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoInitialize FAMs using `FLEX_ARRAY'.
Florian Forster [Sun, 18 Jun 2006 15:18:04 +0000 (17:18 +0200)]
Initialize FAMs using `FLEX_ARRAY'.

When initializing a `flexible array member' the macro `FLEX_ARRAY' should be
used. This was forgotten in `diff-delta.c'.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRemove ranges from switch statements.
Florian Forster [Sun, 18 Jun 2006 15:18:03 +0000 (17:18 +0200)]
Remove ranges from switch statements.

Though very nice and readable, the "case 'a'...'z':" construct is not ANSI C99
compliant. This patch unfolds the range in `quote.c' and substitutes the
switch-statement with an if-statement in `http-fetch.c' and `http-push.c'.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoSupport for extracting configuration from different files
Petr Baudis [Sat, 17 Jun 2006 23:23:58 +0000 (01:23 +0200)]
Support for extracting configuration from different files

Add $GIT_CONFIG environment variable whose content is used instead
of .git/config if set. Also add $GIT_CONFIG_LOCAL as a
forward-compatibility cue for whenever we will finally come to support]
global configuration files (properly).

Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge branch 'yl/build'
Junio C Hamano [Mon, 19 Jun 2006 04:18:49 +0000 (21:18 -0700)]
Merge branch 'yl/build'

* yl/build:
  auto-detect changed prefix and/or changed build flags

18 years agoMerge branch 'jc/shared'
Junio C Hamano [Mon, 19 Jun 2006 03:19:09 +0000 (20:19 -0700)]
Merge branch 'jc/shared'

* jc/shared:
  shared repository: optionally allow reading to "others".

18 years agoMerge branch 'eb/mail'
Junio C Hamano [Mon, 19 Jun 2006 03:18:21 +0000 (20:18 -0700)]
Merge branch 'eb/mail'

* eb/mail:
  Fix git-format-patch -s

18 years agoFix PPC SHA1 routine for large input buffers
Paul Mackerras [Sun, 18 Jun 2006 23:25:16 +0000 (09:25 +1000)]
Fix PPC SHA1 routine for large input buffers

The PPC SHA1 routine had an overflow which meant that it gave
incorrect results for input buffers >= 512MB.  This fixes it by
ensuring that the update of the total length in bits is done using
64-bit arithmetic.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake t8001-annotate and t8002-blame more portable
Dennis Stosberg [Sun, 18 Jun 2006 20:33:21 +0000 (22:33 +0200)]
Make t8001-annotate and t8002-blame more portable

These two tests assume that "sed" will not modify the final line of a
stream if it does not end with a newline character.  The assumption is
not true at least for FreeBSD and Solaris 9.  FreeBSD's "sed" appends
a newline character; "sed" in Solaris 9 even removes the incomplete
final line.  This patch makes the test use perl instead.

Signed-off-by: Dennis Stosberg <dennis@stosberg.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoRemove "refs" field from "struct object"
Linus Torvalds [Sun, 18 Jun 2006 18:45:02 +0000 (11:45 -0700)]
Remove "refs" field from "struct object"

This shrinks "struct object" to the absolutely minimal size possible.
It now contains /only/ the object flags and the SHA1 hash name of the
object.

The "refs" field, which is really needed only for fsck, is maintained in
a separate hashed lookup-table, allowing all normal users to totally
ignore it.

This helps memory usage, although not as much as I hoped: it looks like
the allocation overhead of malloc (and the alignment constraints in
particular) means that while the structure size shrinks, the actual
allocation overhead mostly does not.

[ That said: memory usage is actually down, but not as much as it should
  be: I suspect just one of the object types actually ended up shrinking
  its effective allocation size.

  To get to the next level, we probably need specialized allocators that
  don't pad the allocation more than necessary. ]

The separation makes for some code cleanup, though, and makes the ref
tracking that fsck wants a clearly separate thing.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake release tarballs friendlier to older tar versions
Rene Scharfe [Sun, 18 Jun 2006 13:25:33 +0000 (15:25 +0200)]
Make release tarballs friendlier to older tar versions

git-tar-tree adds an extended pax header to archives if its first
parameter points to a commit.  It confuses older tars and isn't
very useful in the case of git anyway, so stop doing it.

Idea: Junio, implementation: Junio.  I just wrote it up. :-)

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-tar-tree: no more void pointer arithmetic
Rene Scharfe [Sun, 18 Jun 2006 10:58:01 +0000 (12:58 +0200)]
git-tar-tree: no more void pointer arithmetic

Noticed by Florian Forster: Use a char pointer when adding offsets,
because void pointer arithmetic is a GNU extension.   Const'ify the
function arguments while we're at it.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-tar-tree: documentation update
Rene Scharfe [Sun, 18 Jun 2006 10:57:59 +0000 (12:57 +0200)]
git-tar-tree: documentation update

 * add example on how to avoid adding a global extended pax header
 * don't mention linux anymore, use git itself as an example instead
 * update to v1.4.0 ;-)
 * append missing :: to the examples

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-tar-tree: Simplify write_trailer()
Rene Scharfe [Sun, 18 Jun 2006 10:57:56 +0000 (12:57 +0200)]
git-tar-tree: Simplify write_trailer()

We can write the trailer in one or at most two steps; it will always
fit within two blocks.  With the last caller of get_record() gone we
can get rid of it.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoauto-detect changed prefix and/or changed build flags
Yakov Lerner [Wed, 14 Jun 2006 22:36:00 +0000 (01:36 +0300)]
auto-detect changed prefix and/or changed build flags

Detect changed prefix and/or changed build flags in the middle
of the build (or between 'make' and 'make install'), and if change
is detected, make sure all objects are compiled with same build
flags and same prefix, thus avoiding inconsistent/broken build.

[jc: removed otherwise unnecessary Makefile target to test the
 change this patch introduces. ]

Signed-off-by: Yakov Lerner <iler.ml@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoFix git-format-patch -s
Eric W. Biederman [Mon, 12 Jun 2006 19:31:38 +0000 (13:31 -0600)]
Fix git-format-patch -s

When git-format-patch was converted to a builtin an appropriate call
to setup_ident was missed and thus git-format-patch -s fails because
it doesn't look up anything in the password file.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoSome more memory leak avoidance
Linus Torvalds [Sun, 18 Jun 2006 01:47:58 +0000 (18:47 -0700)]
Some more memory leak avoidance

This is really the dregs of my effort to not waste memory in git-rev-list,
and makes barely one percent of a difference in the memory footprint, but
hey, it's also a pretty small patch.

It discards the parent lists and the commit buffer after the commit has
been shown by git-rev-list (and "git log" - which already did the commit
buffer part), and frees the commit list entry that was used by the
revision walker.

The big win would be to get rid of the "refs" pointer in the object
structure (another 5%), because it's only used by fsck. That would require
some pretty major surgery to fsck, though, so I'm timid and did the less
interesting but much easier part instead.

This (percentually) makes a bigger difference to "git log" and friends,
since those are walking _just_ commits, and thus the list entries tend to
be a bigger percentage of the memory use. But the "list all objects" case
does improve too.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMove "void *util" from "struct object" into "struct commit"
Linus Torvalds [Sun, 18 Jun 2006 01:26:18 +0000 (18:26 -0700)]
Move "void *util" from "struct object" into "struct commit"

Every single user actually wanted this only for commit objects, and we
have no reason to waste space on it for other object types. So just move
the structure member from the low-level "struct object" into the "struct
commit".

This leaves the commit object the same size, and removes one unnecessary
pointer from all other object allocations.

This shrinks memory usage (still at a fairly hefty half-gig, admittedly)
of "git-rev-list --all --objects" on the mozilla repo by another 5% in my
tests.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoShrink "struct object" a bit
Linus Torvalds [Wed, 14 Jun 2006 23:45:13 +0000 (16:45 -0700)]
Shrink "struct object" a bit

This shrinks "struct object" by a small amount, by getting rid of the
"struct type *" pointer and replacing it with a 3-bit bitfield instead.

In addition, we merge the bitfields and the "flags" field, which
incidentally should also remove a useless 4-byte padding from the object
when in 64-bit mode.

Now, our "struct object" is still too damn large, but it's now less
obviously bloated, and of the remaining fields, only the "util" (which is
not used by most things) is clearly something that should be eventually
discarded.

This shrinks the "git-rev-list --all" memory use by about 2.5% on the
kernel archive (and, perhaps more importantly, on the larger mozilla
archive). That may not sound like much, but I suspect it's more on a
64-bit platform.

There are other remaining inefficiencies (the parent lists, for example,
probably have horrible malloc overhead), but this was pretty obvious.

Most of the patch is just changing the comparison of the "type" pointer
from one of the constant string pointers to the appropriate new TYPE_xxx
small integer constant.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMerge early part of branch 'jc/fetchupload'
Junio C Hamano [Sun, 18 Jun 2006 00:57:33 +0000 (17:57 -0700)]
Merge early part of branch 'jc/fetchupload'

18 years agoMerge branch 'jc/rw-prefix'
Junio C Hamano [Sun, 18 Jun 2006 00:56:52 +0000 (17:56 -0700)]
Merge branch 'jc/rw-prefix'

* jc/rw-prefix:
  read-tree: reorganize bind_merge code.
  write-tree: --prefix=<path>
  read-tree: --prefix=<path>/ option.

18 years agoMerge branch 'pe/date'
Junio C Hamano [Sun, 18 Jun 2006 00:49:28 +0000 (17:49 -0700)]
Merge branch 'pe/date'

* pe/date:
  date.c: improve guess between timezone offset and year.

18 years agodiff options: add --color
Johannes Schindelin [Tue, 13 Jun 2006 16:45:44 +0000 (18:45 +0200)]
diff options: add --color

This patch is a slightly adjusted version of Junio's patch:
http://www.gelato.unsw.edu.au/archives/git/0604/19354.html

However, instead of using a config variable, this patch makes it available
as a diff option.

Signed-off-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agomailinfo: ignore blanks after in-body headers.
Junio C Hamano [Sat, 17 Jun 2006 23:58:51 +0000 (16:58 -0700)]
mailinfo: ignore blanks after in-body headers.

[jc: this is based on Eric's patch but also fixes up the parsed
 subject headers].

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoDon't parse any headers in the real body of an email message.
Eric W. Biederman [Mon, 12 Jun 2006 19:48:35 +0000 (13:48 -0600)]
Don't parse any headers in the real body of an email message.

It was pointed out that the current behaviour might mispart a patch comment
so remove this behaviour for now.

[jc: this fixes "From: line in the middle" check in t5100 test.]

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agot5100: mailinfo and mailsplit tests.
Junio C Hamano [Sat, 17 Jun 2006 22:20:36 +0000 (15:20 -0700)]
t5100: mailinfo and mailsplit tests.

Currently the test passes with 1.3.3 but not with the tip of
"master".  This is to verify the fixes from Eric W Biedermann.

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoMake t4101-apply-nonl bring along its patches
Dennis Stosberg [Sat, 17 Jun 2006 15:02:45 +0000 (17:02 +0200)]
Make t4101-apply-nonl bring along its patches

Some versions of "diff" (e.g. on FreeBSD and older Linux systems) do
not support the "\ No newline at end of file" remark and are not
able to generate the patches needed for this test.  This lets the
test fail, although git-apply is working perfectly.  This patch adds
the pre-generated patches to t/t4100/ and makes the test use them.

Signed-off-by: Dennis Stosberg <dennis@stosberg.net>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoUpdate gitweb README: gitweb is now included with git
Jakub Narebski [Sat, 17 Jun 2006 06:53:45 +0000 (08:53 +0200)]
Update gitweb README: gitweb is now included with git

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogit-cvsexportcommit.perl: fix typo
Sven Verdoolaege [Sat, 17 Jun 2006 17:46:40 +0000 (19:46 +0200)]
git-cvsexportcommit.perl: fix typo

Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb.cgi history not shown
Linus Torvalds [Sun, 11 Jun 2006 17:57:35 +0000 (10:57 -0700)]
gitweb.cgi history not shown

This does:

 - add a "rev.simplify_history" flag which defaults to on
 - it turns it off for "git whatchanged" (which thus now has real
   semantics outside of "git log")
 - it adds a command line flag ("--full-history") to turn it off for
   others (ie you can make "git log" and "gitk" etc get the semantics if
   you want to.

Now, just as an example of _why_ you really really really want to simplify
history by default, apply this patch, install it, and try these two
command lines:

gitk --full-history -- git.c
gitk -- git.c

and compare the output.

So with this, you can also now do

git whatchanged -p -- gitweb.cgi
git log -p --full-history -- gitweb.cgi

and it will show the old history of gitweb.cgi, even though it's not
relevant to the _current_ state of the name "gitweb.cgi"

NOTE NOTE NOTE! It will still actually simplify away merges that didn't
change anything at all into either child. That creates these bogus strange
discontinuities if you look at it with "gitk" (look at the --full-history
gitk output for git.c, and you'll see a few strange cases).

So the whole "--parent" thing ends up somewhat bogus with --full-history
because of this, but I'm not sure it's worth even worrying about. I don't
think you'd ever want to really use "--full-history" with the graphical
representation, I just give it as an example exactly to show _why_ doing
so would be insane.

I think this is trivial enough and useful enough to be worth merging into
the stable branch.

Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agoImplement safe_strncpy() as strlcpy() and use it more.
Peter Eriksen [Sun, 11 Jun 2006 12:03:28 +0000 (14:03 +0200)]
Implement safe_strncpy() as strlcpy() and use it more.

Signed-off-by: Peter Eriksen <s022018@student.dtu.dk>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Make the `blame' interface in gitweb optional.
Florian Forster [Mon, 12 Jun 2006 08:31:57 +0000 (10:31 +0200)]
gitweb: Make the `blame' interface in gitweb optional.

Since `git-annotate' is an expensive operation to run it may be
desirable to deactivate this functionality. This patch introduces
the `gitweb.blame' option to git-repo-config and disables the blame
support by default.

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agogitweb: Adding a `blame' interface.
Florian Forster [Sun, 11 Jun 2006 15:45:19 +0000 (17:45 +0200)]
gitweb: Adding a `blame' interface.

This patch adds an interface for `git-blame' to `gitweb.cgi'.
Links to it are placed in `git_blob'.

Internally the code uses `git-annotate' because `git-blame's output
differs for files that have been renamed in the past. However, I like
the term `blame' better.

[jc: blame can be told to produce the compatible format btw...]

Signed-off-by: Florian Forster <octo@verplant.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agocvsimport: keep one index per branch during import
Martin Langhoff [Mon, 12 Jun 2006 11:50:49 +0000 (23:50 +1200)]
cvsimport: keep one index per branch during import

With this patch we have a speedup and much lower IO when
importing trees with many branches. Instead of forcing
index re-population for each branch switch, we keep
many index files around, one per branch.

Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
18 years agocvsimport: complete the cvsps run before starting the import
Martin Langhoff [Sun, 11 Jun 2006 08:12:20 +0000 (20:12 +1200)]
cvsimport: complete the cvsps run before starting the import

We now capture the output of cvsps to a tempfile, and then read it in.
cvsps 2.1 works quite a bit "in memory", and only prints its patchset
info once it has finished talking with cvs, but apparently retaining
all that memory allocation. With this patch, cvsps is finished and
reaped before cvsimport start working (and growing). So the footprint
of the whole process is much lower.

Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
Signed-off-by: Junio C Hamano <junkio@cox.net>