Code

git.git
15 years agoMake usage of windows.h lean and mean
Marius Storm-Olsen [Wed, 16 Sep 2009 08:20:26 +0000 (10:20 +0200)]
Make usage of windows.h lean and mean

Centralize the include of windows.h in git-compat-util.h, turn on
WIN32_LEAN_AND_MEAN to avoid including plenty of other header files
which is not needed in Git. Also ensure we load winsock2.h first,
so we don't load the older winsock definitions at a later stage,
since they contain duplicate definitions.

When moving windows.h into git-compat-util.h, we need to protect
the definition of struct pollfd in mingw.h, since this file is used
by both MinGW and MSVC, and the latter defines this struct in
winsock2.h.

We need to keep the windows.h include in compat/win32.h, since its
shared by both MinGW and Cygwin, and we're not touching Cygwin in
this commit. The include in git-compat-util.h is protected with an
ifdef WIN32, which is not the case when compiling for Cygwin.

Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd platform files for porting to MSVC
Frank Li [Wed, 16 Sep 2009 08:20:25 +0000 (10:20 +0200)]
Add platform files for porting to MSVC

Add msvc.c and msvc.h to build git under MSVC.

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd MinGW header files to build git with MSVC
Frank Li [Wed, 16 Sep 2009 08:20:24 +0000 (10:20 +0200)]
Add MinGW header files to build git with MSVC

Added the header files dirent.h, unistd.h and utime.h
Add alloca.h, which simply includes malloc.h, which defines alloca().

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd empty header files for MSVC port
Marius Storm-Olsen [Wed, 16 Sep 2009 08:20:23 +0000 (10:20 +0200)]
Add empty header files for MSVC port

MSVC lacks many of the header files included by git-compat-util.h; add
blank header files for these instead of going ifdef crazy.

Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoTest for WIN32 instead of __MINGW32_
Frank Li [Wed, 16 Sep 2009 08:20:22 +0000 (10:20 +0200)]
Test for WIN32 instead of __MINGW32_

The code which is conditional on MinGW32 is actually conditional on Windows.
Use the WIN32 symbol, which is defined by the MINGW32 and MSVC environments,
but not by Cygwin.

Define SNPRINTF_SIZE_CORR=1 for MSVC too, as its vsnprintf function does
not add NUL at the end of the buffer if the result fits the buffer size
exactly.

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoFix __stdcall placement and function prototype
Frank Li [Wed, 16 Sep 2009 08:20:21 +0000 (10:20 +0200)]
Fix __stdcall placement and function prototype

MSVC requires __stdcall to be between the functions return value and the
function name, and that the function pointer type is in the form of

    return_type (WINAPI *function_name)(arguments...)

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoSet _O_BINARY as default fmode for both MinGW and MSVC
Marius Storm-Olsen [Wed, 16 Sep 2009 08:20:20 +0000 (10:20 +0200)]
Set _O_BINARY as default fmode for both MinGW and MSVC

MinGW set the _CRT_fmode to set both the default fmode and _O_BINARY on
stdin/stdout/stderr. Rather use the main() define in mingw.h to set this
for both MinGW and MSVC.

This will ensure that a MinGW and MSVC build will handle input and output
identically.

Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoChange regerror() declaration from K&R style to ANSI C (C89)
Frank Li [Wed, 16 Sep 2009 08:20:19 +0000 (10:20 +0200)]
Change regerror() declaration from K&R style to ANSI C (C89)

The MSVC headers typedef errcode as int, and thus confused the compiler in
the K&R style definition. ANSI style deconfuses it.

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd include guards to compat/win32.h
Marius Storm-Olsen [Wed, 16 Sep 2009 08:20:18 +0000 (10:20 +0200)]
Add include guards to compat/win32.h

Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAvoid declaration after statement
Frank Li [Wed, 16 Sep 2009 08:20:17 +0000 (10:20 +0200)]
Avoid declaration after statement

MSVC does not understand this C99 style.

Signed-off-by: Frank Li <lznuaa@gmail.com>
Signed-off-by: Marius Storm-Olsen <mstormo@gmail.com>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopack-objects: remove SP at the end of usage string
Thiago Farina [Fri, 18 Sep 2009 19:11:52 +0000 (12:11 -0700)]
pack-objects: remove SP at the end of usage string

These spaces immediately before the end of lines are unnecessary.

While at it, instead of using a single string literal with backslashes
at end of each line, split the lines into individual string literals
and tell the compiler to concatenate them.

Signed-off-by: Thiago Farina <tfransosi@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUpdate the usage bundle string.
Thiago Farina [Thu, 17 Sep 2009 17:20:53 +0000 (13:20 -0400)]
Update the usage bundle string.

"git bundle -h" gives a single long line that is hard to read.  Rewrite it
into a multi-line format similar to the one used by other commands, e.g
"git stash -h".

Signed-off-by: Thiago Farina <tfransosi@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agocvs: initialize empty password
Clemens Buchacher [Thu, 17 Sep 2009 07:21:02 +0000 (09:21 +0200)]
cvs: initialize empty password

If we do not read a password from the command line, and there are no
passwords stored in .cvspass, we have to initialize the password with
just "A".

This fixes a regression introduced by 3fb9d582 (Do not scramble
password read from .cvspass).

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge 1.6.4.4 in
Junio C Hamano [Wed, 16 Sep 2009 22:04:21 +0000 (15:04 -0700)]
Merge 1.6.4.4 in

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.4.4 v1.6.4.4
Junio C Hamano [Wed, 16 Sep 2009 21:53:26 +0000 (14:53 -0700)]
GIT 1.6.4.4

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'jc/maint-unpack-objects-strict' into maint
Junio C Hamano [Wed, 16 Sep 2009 21:45:18 +0000 (14:45 -0700)]
Merge branch 'jc/maint-unpack-objects-strict' into maint

* jc/maint-unpack-objects-strict:
  Fix "unpack-objects --strict"

Conflicts:
builtin-unpack-objects.c

15 years agoMerge branch 'tf/diff-whitespace-incomplete-line' into maint
Junio C Hamano [Wed, 16 Sep 2009 21:27:08 +0000 (14:27 -0700)]
Merge branch 'tf/diff-whitespace-incomplete-line' into maint

* tf/diff-whitespace-incomplete-line:
  xutils: Fix xdl_recmatch() on incomplete lines
  xutils: Fix hashing an incomplete line with whitespaces at the end

15 years agoMerge branch 'jk/maint-1.6.3-checkout-unborn' into maint
Junio C Hamano [Wed, 16 Sep 2009 21:26:56 +0000 (14:26 -0700)]
Merge branch 'jk/maint-1.6.3-checkout-unborn' into maint

* jk/maint-1.6.3-checkout-unborn:
  checkout: do not imply "-f" on unborn branches

15 years agoMerge branch 'jc/maint-checkout-index-to-prefix' into maint
Junio C Hamano [Wed, 16 Sep 2009 21:26:40 +0000 (14:26 -0700)]
Merge branch 'jc/maint-checkout-index-to-prefix' into maint

* jc/maint-checkout-index-to-prefix:
  check_path(): allow symlinked directories to checkout-index --prefix

15 years agoMerge branch 'maint'
Junio C Hamano [Mon, 14 Sep 2009 21:48:27 +0000 (14:48 -0700)]
Merge branch 'maint'

* maint:
  http.c: avoid freeing an uninitialized pointer

15 years agoMerge branch 'rc/maint-http-no-head-pack-check' into maint
Junio C Hamano [Mon, 14 Sep 2009 21:48:20 +0000 (14:48 -0700)]
Merge branch 'rc/maint-http-no-head-pack-check' into maint

* rc/maint-http-no-head-pack-check:
  http.c: avoid freeing an uninitialized pointer

15 years agohttp.c: avoid freeing an uninitialized pointer
Junio C Hamano [Mon, 14 Sep 2009 21:48:15 +0000 (14:48 -0700)]
http.c: avoid freeing an uninitialized pointer

An earlier 59b8d38 (http.c: remove verification of remote packs) left
the variable "url" uninitialized; "goto cleanup" codepath can free it
which is not very nice.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoweb--browse: fix Mac OS X GUI detection for 10.6
Heiko Voigt [Mon, 14 Sep 2009 08:49:01 +0000 (10:49 +0200)]
web--browse: fix Mac OS X GUI detection for 10.6

Since OS X 10.6 the variable $SECURITYSESSIONID does not exist anymore,
so lets look for the $TERM_PROGRAM variable as backup.

Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoremove logical typo in documentation of sample update hook
Heiko Voigt [Mon, 14 Sep 2009 08:47:06 +0000 (10:47 +0200)]
remove logical typo in documentation of sample update hook

Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoNicolas Pitre has a new email address
Nicolas Pitre [Mon, 14 Sep 2009 06:41:16 +0000 (02:41 -0400)]
Nicolas Pitre has a new email address

Due to problems at cam.org, my nico@cam.org email address is no longer
valid.  From now on, nico@fluxnic.net should be used instead.

Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-archive: infer output format from filename when unspecified
Dmitry Potapov [Sun, 13 Sep 2009 20:17:01 +0000 (00:17 +0400)]
git-archive: infer output format from filename when unspecified

A command line

    $ git archive -o my-v2.0.zip v2.0

almost certainly wants the output in zip format, even though it does not
specify any --format option.

When --format is not given, but output filename is, try to infer what
format is requested from the filename extension.  Currently this code only
knows about '.zip'.  When the format is unspecified and the filename does
not tell us, the output will be in 'tar' format as before.

Of course, an explicit --format will not trigger this guesswork.

Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agocompletion: Replace config --list with --get-regexp
Todd Zullinger [Fri, 11 Sep 2009 23:23:45 +0000 (19:23 -0400)]
completion: Replace config --list with --get-regexp

James Bardin noted that the completion spewed warnings when no git config
file is present.  This is likely a bug to be fixed in git config, but it's
also good to simplify the completion code by using the --get-regexp option
as Jeff King pointed out.

Signed-off-by: Todd Zullinger <tmz@pobox.com>
Trivially-acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-archive: add '-o' as a alias for '--output'
Dmitry Potapov [Sun, 13 Sep 2009 12:05:52 +0000 (16:05 +0400)]
git-archive: add '-o' as a alias for '--output'

The '-o' option is commonly used in many tools to specify the output file.
Typing '--output' every time is a bit too long to be a practical alternative
to redirecting output. But specifying the output name has the advantage of
making possible to guess the desired output format by filename extension.

Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agotest local clone by copying
Clemens Buchacher [Sun, 13 Sep 2009 10:49:45 +0000 (12:49 +0200)]
test local clone by copying

Test the effect of an earlier change by f7835a2 (preserve mtime of local
clone, 2009-09-12) to keep stale loose object files stale in the new
repository when a local clone is made by copying files in .git/
directory.

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-clone doc: typofix
Johannes Gilger [Sun, 13 Sep 2009 13:35:05 +0000 (15:35 +0200)]
git-clone doc: typofix

Signed-off-by: Johannes Gilger <heipei@hackvalue.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-push: Accept -n as a synonym for --dry-run.
Nelson Elhage [Sun, 13 Sep 2009 16:56:45 +0000 (12:56 -0400)]
git-push: Accept -n as a synonym for --dry-run.

git-push is not currently using -n for anything else, and it seems
unlikely we will want to use it to mean anything else in the future,
so add it as an alias for convenience.

Signed-off-by: Nelson Elhage <nelhage@mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-commit doc: remove duplicated --dry-run description
Junio C Hamano [Sun, 13 Sep 2009 18:23:34 +0000 (11:23 -0700)]
git-commit doc: remove duplicated --dry-run description

60c2993 (Documentation/git-commit.txt: describe --dry-run, 2009-08-15)
wanted to update the documentation to say that "git status" is not the
same as "git commit --dry-run" anymore, but it screwed up and also added
the description of --dry-run that was already present.

Noticed by Johannes Gilger.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.5-rc1 v1.6.5-rc1
Junio C Hamano [Sun, 13 Sep 2009 08:50:22 +0000 (01:50 -0700)]
GIT 1.6.5-rc1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoImprove --patch option documentation in git-add
Jari Aalto [Sun, 13 Sep 2009 06:43:10 +0000 (09:43 +0300)]
Improve --patch option documentation in git-add

Signed-off-by: Jari Aalto <jari.aalto@cante.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'pk/fast-import-tars'
Junio C Hamano [Sun, 13 Sep 2009 08:33:29 +0000 (01:33 -0700)]
Merge branch 'pk/fast-import-tars'

* pk/fast-import-tars:
  import-tars: Allow per-tar author and commit message.

15 years agoMerge branch 'pk/fast-import-dirs'
Junio C Hamano [Sun, 13 Sep 2009 08:33:26 +0000 (01:33 -0700)]
Merge branch 'pk/fast-import-dirs'

* pk/fast-import-dirs:
  Add script for importing bits-and-pieces to Git.

15 years agoMerge branch 'jt/pushinsteadof'
Junio C Hamano [Sun, 13 Sep 2009 08:33:20 +0000 (01:33 -0700)]
Merge branch 'jt/pushinsteadof'

* jt/pushinsteadof:
  Add url.<base>.pushInsteadOf: URL rewriting for push only
  Wrap rewrite globals in a struct in preparation for adding another set

15 years agoMerge branch 'jk/unwanted-advices'
Junio C Hamano [Sun, 13 Sep 2009 08:33:18 +0000 (01:33 -0700)]
Merge branch 'jk/unwanted-advices'

* jk/unwanted-advices:
  status: make "how to stage" messages optional
  push: make non-fast-forward help message configurable

15 years agoMerge branch 'jc/merge-saner-messages'
Junio C Hamano [Sun, 13 Sep 2009 08:33:15 +0000 (01:33 -0700)]
Merge branch 'jc/merge-saner-messages'

* jc/merge-saner-messages:
  merge-recursive: give less scary messages when merge did not start

15 years agoquiltimport documentation: --dry-run and -n are synonyms
Junio C Hamano [Sun, 13 Sep 2009 02:40:27 +0000 (19:40 -0700)]
quiltimport documentation: --dry-run and -n are synonyms

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopreserve mtime of local clone
Clemens Buchacher [Sat, 12 Sep 2009 09:03:48 +0000 (11:03 +0200)]
preserve mtime of local clone

A local clone without hardlinks copies all objects, including dangling
ones, to the new repository. Since the mtimes are renewed, those
dangling objects cannot be pruned by "git gc --prune", even if they
would have been old enough for pruning in the original repository.

Instead, preserve mtime during copy. "git gc --prune" will then work
in the clone just like it did in the original.

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agodon't dereference NULL upon fdopen failure
Jim Meyering [Sat, 12 Sep 2009 08:43:27 +0000 (10:43 +0200)]
don't dereference NULL upon fdopen failure

There were several unchecked use of fdopen(); replace them with xfdopen()
that checks and dies.

Signed-off-by: Jim Meyering <meyering@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agouse write_str_in_full helper to avoid literal string lengths
Jim Meyering [Sat, 12 Sep 2009 08:54:32 +0000 (10:54 +0200)]
use write_str_in_full helper to avoid literal string lengths

This is the same fix to use write_str_in_full() helper to write a constant
string out without counting the length of it ourselves.

Signed-off-by: Jim Meyering <meyering@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'db/vcs-helper'
Junio C Hamano [Sun, 13 Sep 2009 08:31:55 +0000 (01:31 -0700)]
Merge branch 'db/vcs-helper'

* db/vcs-helper:
  Makefile: remove remnant of separate http/https/ftp helpers
  Use a clearer style to issue commands to remote helpers
  Make the "traditionally-supported" URLs a special case
  Makefile: install hardlinks for git-remote-<scheme> supported by libcurl if possible
  Makefile: do not link three copies of git-remote-* programs
  Makefile: git-http-fetch does not need expat
  http-fetch: Fix Makefile dependancies
  Add transport native helper executables to .gitignore
  git-http-fetch: not a builtin
  Use an external program to implement fetching with curl
  Add support for external programs for handling native fetches

15 years agouse write_str_in_full helper to avoid literal string lengths
Jim Meyering [Sat, 12 Sep 2009 08:54:32 +0000 (10:54 +0200)]
use write_str_in_full helper to avoid literal string lengths

In 2d14d65 (Use a clearer style to issue commands to remote helpers,
2009-09-03) I happened to notice two changes like this:

- write_in_full(helper->in, "list\n", 5);
+
+ strbuf_addstr(&buf, "list\n");
+ write_in_full(helper->in, buf.buf, buf.len);
+ strbuf_reset(&buf);

IMHO, it would be better to define a new function,

    static inline ssize_t write_str_in_full(int fd, const char *str)
    {
           return write_in_full(fd, str, strlen(str));
    }

and then use it like this:

-       strbuf_addstr(&buf, "list\n");
-       write_in_full(helper->in, buf.buf, buf.len);
-       strbuf_reset(&buf);
+       write_str_in_full(helper->in, "list\n");

Thus not requiring the added allocation, and still avoiding
the maintenance risk of literal string lengths.
These days, compilers are good enough that strlen("literal")
imposes no run-time cost.

Transformed via this:

    perl -pi -e \
        's/write_in_full\((.*?), (".*?"), \d+\)/write_str_in_full($1, $2)/'\
      $(git grep -l 'write_in_full.*"')

Signed-off-by: Jim Meyering <meyering@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agowrap git's main usage string.
Matthieu Moy [Sat, 12 Sep 2009 10:39:30 +0000 (12:39 +0200)]
wrap git's main usage string.

It's now similar wrapped the same way as in Documentation/git.txt, and
fits in a 67 characters wide terminal.

Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Sun, 13 Sep 2009 08:30:53 +0000 (01:30 -0700)]
Merge branch 'maint'

* maint:
  GIT 1.6.4.3
  svn: properly escape arguments for authors-prog
  http.c: remove verification of remote packs
  grep: accept relative paths outside current working directory
  grep: fix exit status if external_grep() punts

Conflicts:
GIT-VERSION-GEN
RelNotes

15 years agoGIT 1.6.4.3 v1.6.4.3
Junio C Hamano [Sun, 13 Sep 2009 08:04:23 +0000 (01:04 -0700)]
GIT 1.6.4.3

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agosvn: properly escape arguments for authors-prog
Mark Lodato [Sun, 13 Sep 2009 00:33:23 +0000 (20:33 -0400)]
svn: properly escape arguments for authors-prog

Previously, the call to authors-prog was not properly escaped, so any
special characters in the Subversion username, such as spaces and
semi-colons, would be interpreted by the shell rather than being passed
in as the first argument.  Now all unsafe characters are escaped using
"git rev-parse --sq-quote"

[ew: switched from "\Q..\E" to "rev-parse --sq-quote"]

Signed-off-by: Mark Lodato <lodatom@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'cb/maint-1.6.3-grep-relative-up' into maint
Junio C Hamano [Sun, 13 Sep 2009 08:24:20 +0000 (01:24 -0700)]
Merge branch 'cb/maint-1.6.3-grep-relative-up' into maint

* cb/maint-1.6.3-grep-relative-up:
  grep: accept relative paths outside current working directory
  grep: fix exit status if external_grep() punts

Conflicts:
t/t7002-grep.sh

15 years agoMakefile: remove remnant of separate http/https/ftp helpers
Junio C Hamano [Sat, 12 Sep 2009 05:35:30 +0000 (22:35 -0700)]
Makefile: remove remnant of separate http/https/ftp helpers

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agorebase: use plumbing to show dirty state
Jeff King [Wed, 9 Sep 2009 14:59:37 +0000 (10:59 -0400)]
rebase: use plumbing to show dirty state

Commit 4cfbe06 introduced the use of "git diff" to show
dirty state in a format more familiar to users. However, it
should have used the plumbing "git diff-files" instead.

Not only is it good practice in general to use plumbing in
scripts, but in this case we really don't want the automatic
pager to kick in for an error message.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agostatus: make "how to stage" messages optional
Jeff King [Wed, 9 Sep 2009 11:43:03 +0000 (07:43 -0400)]
status: make "how to stage" messages optional

These messages are nice for new users, but experienced git
users know how to manipulate the index, and these messages
waste a lot of screen real estate.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopush: make non-fast-forward help message configurable
Jeff King [Wed, 9 Sep 2009 11:38:58 +0000 (07:38 -0400)]
push: make non-fast-forward help message configurable

This message is designed to help new users understand what
has happened when refs fail to push. However, it does not
help experienced users at all, and significantly clutters
the output, frequently dwarfing the regular status table and
making it harder to see.

This patch introduces a general configuration mechanism for
optional messages, with this push message as the first
example.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopager: set LESS=FRSX also on Windows
Johannes Sixt [Fri, 11 Sep 2009 17:45:07 +0000 (19:45 +0200)]
pager: set LESS=FRSX also on Windows

Previously, this environment variable was set in the pager_preexec
callback, which is conditionally-compiled only on Unix, because it is not,
and cannot be, called on Windows.

With this patch the env member of struct child_process is used to set
the environment variable, which also works on Windows.

Noticed by Alexey Borzenkov.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agostart_command: do not clobber cmd->env on Windows code path
Johannes Sixt [Fri, 11 Sep 2009 17:40:08 +0000 (19:40 +0200)]
start_command: do not clobber cmd->env on Windows code path

Previously, it would not be possible to call start_command twice for the
same struct child_process that has env set.

The fix is achieved by moving the loop that modifies the environment block
into a helper function. This also allows us to make two other helper
functions static.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agohttp.c: remove verification of remote packs
Tay Ray Chuan [Wed, 9 Sep 2009 12:33:50 +0000 (20:33 +0800)]
http.c: remove verification of remote packs

Make http.c::fetch_pack_index() no longer check for the remote pack
with a HEAD request before fetching the corresponding pack index file.

Not only does sending a HEAD request before we do a GET incur a
performance penalty, it does not offer any significant error-
prevention advantages (pack fetching in the *_http_pack_request()
methods is capable of handling any errors on its own).

This addresses an issue raised elsewhere:

  http://code.google.com/p/msysgit/issues/detail?id=323
  http://support.github.com/discussions/repos/957-cant-clone-over-http-or-git

Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoadd documentation for mailinfo.scissors and '--no-scissors'
Nicolas Sebrecht [Fri, 11 Sep 2009 00:29:58 +0000 (02:29 +0200)]
add documentation for mailinfo.scissors and '--no-scissors'

Signed-off-by: Nicolas Sebrecht <nicolas.s.dev@gmx.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomailinfo: add '--scissors' to usage message
Nicolas Sebrecht [Fri, 11 Sep 2009 00:09:20 +0000 (02:09 +0200)]
mailinfo: add '--scissors' to usage message

Signed-off-by: Nicolas Sebrecht <nicolas.s.dev@gmx.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoINSTALL: Describe dependency knobs from Makefile
Brian Gernhardt [Thu, 10 Sep 2009 20:28:19 +0000 (16:28 -0400)]
INSTALL: Describe dependency knobs from Makefile

We said that some of our dependencies were optional, but didn't say
how to turn them off.  Add information for that and mention where to
save the options close to the top of the file.

Also, standardize on both using quotes for the names of the dependencies
and tabs for indentation of the list.

Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoINSTALL: Reorder dependencies, split shell and Perl
Brian Gernhardt [Wed, 9 Sep 2009 01:51:00 +0000 (21:51 -0400)]
INSTALL: Reorder dependencies, split shell and Perl

The most important and non-optional dependencies should go first, so put
them there.  While we're moving them, the descriptions for shell and perl
were archaic, referring to "bare-bones Porcelainish scripts" that have
become powerful and essential.

Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit-p4: Avoid modules deprecated in Python 2.6.
Reilly Grant [Thu, 10 Sep 2009 07:02:38 +0000 (00:02 -0700)]
git-p4: Avoid modules deprecated in Python 2.6.

The popen2, sha and sets modules are deprecated in Python 2.6 (sha in
Python 2.5).  Both popen2 and sha are not actually used in git-p4.
Replace usage of sets.Set with the builtin set object.

The built-in set object was added in Python 2.4 and is already used in
other parts of this script, so this dependency is nothing new.

Signed-off-by: Reilly Grant <reillyeon@qotw.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMakefile: Add NEEDS_CRYPTO_WITH_SSL
Brian Gernhardt [Tue, 8 Sep 2009 13:54:38 +0000 (09:54 -0400)]
Makefile: Add NEEDS_CRYPTO_WITH_SSL

The Makefile comment for NEEDS_SSL_WITH_CRYPTO says to define it "if
you need -lcrypto with -lssl (Darwin)."  However, what it actually
does is add -lssl when you use -lcrypto and not the other way around.
However, libcrypto contains a majority of the ERR_* functions from
OpenSSL (at least on OS X) so we need it both ways.

So, add NEEDS_CRYPTO_WITH_SSL which adds -lcrypto to the OpenSSL link
flags and clarify the difference between it and NEEDS_SSL_WITH_CRYPTO.

Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogit.el: Use git-add-file for unmerged files, remove git-resolve-file
Martin Nordholts [Thu, 3 Sep 2009 20:27:24 +0000 (22:27 +0200)]
git.el: Use git-add-file for unmerged files, remove git-resolve-file

Use `git-add-file' to mark unmerged files as resolved in the
*git-status* buffer to be consistent with git's CLI instructions. Also
remove `git-resolve-file' to make it clear that that "R" is a now a
free keybinding.

Signed-off-by: Martin Nordholts <martinn@src.gnome.org>
Acked-by: Alexandre Julliard <julliard@winehq.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoINSTALL: Update description of our SHA-1 code
Brian Gernhardt [Wed, 9 Sep 2009 01:50:59 +0000 (21:50 -0400)]
INSTALL: Update description of our SHA-1 code

We haven't had Mozilla's code or an ARM optimized algorithm since
30ae47b.  Reword the paragraph to give credit but not authorship to
Mozilla.

Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomerge-recursive: give less scary messages when merge did not start
Junio C Hamano [Tue, 8 Sep 2009 05:43:11 +0000 (22:43 -0700)]
merge-recursive: give less scary messages when merge did not start

When unpack_trees() three-way merge logic is called from merge-recursive
and finds that local changes are going to be clobbered, its plumbing level
messages were given as errors first, and then the merge driver added even
more scary message "fatal: merging of trees <a long object name> and
<another long object name> failed".

This is most often encountered by new CVS/SVN migrants who are used to
start a merge from a dirty work tree.  The saddest part is that the merge
refused to run to prevent _any_ damage from being done to your work tree
when these messages are given, but the messages look a lot more scarier
than the conflicted case where the user needs to resolve them.

Replace the plumbing level messages so that they talk about what it is
protecting the user from, and end the messages with "Aborting." so that it
becomes clear that the command did not do any harm.

The final "merging of trees failed" message is superfluous, unless you are
interested in debugging the merge-recursive itself.  Squelch the current
die() message by default, but allow it to help people who debug git with
verbosity level 4 or greater.

Unless there is some bug, an inner merge that does not touch working tree
should not trigger any such error, so emit the current die() message when
we see an error return from it while running the inner merge, too.  It
would also help people who debug git.

We could later add instructions on how to recover (i.e. "stash changes
away or commit on a side branch and retry") instead of the silent
exit(128) I have in this patch, and then use Peff's advice.* mechanism to
squelch it (e.g. "advice.mergeindirtytree"), but they are separate topics.

Tested-by: Nanako Shiraishi <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd url.<base>.pushInsteadOf: URL rewriting for push only
Josh Triplett [Mon, 7 Sep 2009 08:56:33 +0000 (01:56 -0700)]
Add url.<base>.pushInsteadOf: URL rewriting for push only

This configuration option allows systematically rewriting fetch-only URLs
to push-capable URLs when used with push.  For instance:

[url "ssh://example.org/"]
    pushInsteadOf = "git://example.org/"

This will allow clones of "git://example.org/path/to/repo" to subsequently
push to "ssh://example.org/path/to/repo", without manually configuring
pushurl for that remote.

Includes documentation for the new option, bash completion updates, and
test cases (both that pushInsteadOf applies to push, that it does not
apply to fetch, and that it is ignored when pushURL is already defined).

Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoGIT 1.6.5-rc0 v1.6.5-rc0
Junio C Hamano [Tue, 8 Sep 2009 00:20:02 +0000 (17:20 -0700)]
GIT 1.6.5-rc0

Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Mon, 7 Sep 2009 22:45:48 +0000 (15:45 -0700)]
Merge branch 'maint'

* maint:
  git-pull: do not mention --quiet and --verbose twice
  githooks.txt: put hooks into subsections

15 years agogit-pull: do not mention --quiet and --verbose twice
Emmanuel Trillaud [Mon, 7 Sep 2009 12:34:35 +0000 (14:34 +0200)]
git-pull: do not mention --quiet and --verbose twice

git-pull.txt includes fetch-options.txt and merge-options.txt, both of
which document the --quiet and --verbose.

Supress the ones from fetch-options.txt.

Signed-off-by: Emmanuel Trillaud <etrillaud@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogithooks.txt: put hooks into subsections
Bert Wesarg [Sun, 6 Sep 2009 10:22:58 +0000 (12:22 +0200)]
githooks.txt: put hooks into subsections

All hooks are currently in its own section. Which may confuse users,
because the section name serves as the hook file name and sections are
all caps for man pages. Putting them into a new HOOKS section and each
hook into a subsection keeps the case to lower case.

Signed-off-by: Bert Wesarg <bert.wesarg@googlemail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'jc/mailinfo-scissors'
Junio C Hamano [Mon, 7 Sep 2009 22:25:37 +0000 (15:25 -0700)]
Merge branch 'jc/mailinfo-scissors'

* jc/mailinfo-scissors:
  mailinfo.scissors: new configuration
  am/mailinfo: Disable scissors processing by default
  Documentation: describe the scissors mark support of "git am"
  Teach mailinfo to ignore everything before -- >8 -- mark
  builtin-mailinfo.c: fix confusing internal API to mailinfo()

15 years agoMerge branch 'jk/clone-b'
Junio C Hamano [Mon, 7 Sep 2009 22:24:53 +0000 (15:24 -0700)]
Merge branch 'jk/clone-b'

* jk/clone-b:
  clone: add --branch option to select a different HEAD

15 years agoMerge branch 'jc/upload-pack-hook'
Junio C Hamano [Mon, 7 Sep 2009 22:24:47 +0000 (15:24 -0700)]
Merge branch 'jc/upload-pack-hook'

* jc/upload-pack-hook:
  upload-pack: feed "kind [clone|fetch]" to post-upload-pack hook
  upload-pack: add a trigger for post-upload-pack hook

15 years agoMerge branch 'tr/reset-checkout-patch'
Junio C Hamano [Mon, 7 Sep 2009 22:24:38 +0000 (15:24 -0700)]
Merge branch 'tr/reset-checkout-patch'

* tr/reset-checkout-patch:
  stash: simplify defaulting to "save" and reject unknown options
  Make test case number unique
  tests: disable interactive hunk selection tests if perl is not available
  DWIM 'git stash save -p' for 'git stash -p'
  Implement 'git stash save --patch'
  Implement 'git checkout --patch'
  Implement 'git reset --patch'
  builtin-add: refactor the meat of interactive_add()
  Add a small patch-mode testing library
  git-apply--interactive: Refactor patch mode code
  Make 'git stash -k' a short form for 'git stash save --keep-index'

15 years agoMerge branch 'np/maint-1.6.3-deepen'
Junio C Hamano [Mon, 7 Sep 2009 22:23:50 +0000 (15:23 -0700)]
Merge branch 'np/maint-1.6.3-deepen'

* np/maint-1.6.3-deepen:
  pack-objects: free preferred base memory after usage
  make shallow repository deepening more network efficient

15 years agogrep: accept relative paths outside current working directory
Clemens Buchacher [Sat, 5 Sep 2009 12:31:17 +0000 (14:31 +0200)]
grep: accept relative paths outside current working directory

"git grep" would barf at relative paths pointing outside the current
working directory (or subdirectories thereof). Use quote_path_relative(),
which can handle such cases just fine.

[jc: added tests.]

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agogrep: fix exit status if external_grep() punts
Clemens Buchacher [Mon, 7 Sep 2009 08:48:01 +0000 (10:48 +0200)]
grep: fix exit status if external_grep() punts

If external_grep() is called and punts, grep_cache() mistakenly reported a
hit, even if there were none.  The bug can be triggered by calling "git
grep --no-color" from a subdirectory.

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoWrap rewrite globals in a struct in preparation for adding another set
Josh Triplett [Mon, 7 Sep 2009 08:56:00 +0000 (01:56 -0700)]
Wrap rewrite globals in a struct in preparation for adding another set

remote.c has a global set of URL rewrites, accessed by alias_url and
make_rewrite.  Wrap them in a new "struct rewrites", passed to alias_url
and make_rewrite.  This allows adding other sets of rewrites.

Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Sun, 6 Sep 2009 07:39:32 +0000 (00:39 -0700)]
Merge branch 'maint'

* maint:
  push: re-flow non-fast-forward message
  push: fix english in non-fast-forward message

15 years agopush: re-flow non-fast-forward message
Jeff King [Sun, 6 Sep 2009 06:47:20 +0000 (02:47 -0400)]
push: re-flow non-fast-forward message

The extreme raggedness of the right edge make this jarring
to read. Let's re-flow the text to fill the lines in a more
even way.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopush: fix english in non-fast-forward message
Jeff King [Sun, 6 Sep 2009 06:46:25 +0000 (02:46 -0400)]
push: fix english in non-fast-forward message

We must use an article when referring to the section
because it is a non-proper noun, and it must be the definite
article because we are referring to a specific section.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoDo not scramble password read from .cvspass
Pascal Obry [Fri, 4 Sep 2009 11:58:32 +0000 (13:58 +0200)]
Do not scramble password read from .cvspass

Passwords stored in .cvspass are already scrambled, we do not
want to scramble them twice. Only passwords read from the
command line are scrambled.

This fixes a regression introduced by b2139db (git-cvsimport: add support
for cvs pserver password scrambling., 2009-08-14).

Signed-off-by: Pascal Obry <pascal@obry.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agopack-objects: free preferred base memory after usage
Nicolas Pitre [Fri, 4 Sep 2009 01:54:03 +0000 (21:54 -0400)]
pack-objects: free preferred base memory after usage

When adding objects for preferred delta base, the content from tree
objects leading to given paths is kept in a cache. This has the
potential to grow significantly, especially with large directories as
the whole tree object content is loaded in memory, even if in practice
the number of those objects is limited to the 256 cache entries plus the
$window root tree objects.  Still, that can't hurt freeing that up after
object enumeration is done, and before more memory is needed for delta
search.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agomake shallow repository deepening more network efficient
Nicolas Pitre [Thu, 3 Sep 2009 23:08:33 +0000 (19:08 -0400)]
make shallow repository deepening more network efficient

First of all, I can't find any reason why thin pack generation is
explicitly disabled when dealing with a shallow repository.  The
possible delta base objects are collected from the edge commits which
are always obtained through history walking with the same shallow refs
as the client, Therefore the client is always going to have those base
objects available. So let's remove that restriction.

Then we can make shallow repository deepening much more efficient by
using the remote's unshallowed commits as edge commits to get preferred
base objects for thin pack generation.  On git.git, this makes the data
transfer for the deepening of a shallow repository from depth 1 to depth 2
around 134 KB instead of 3.68 MB.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoUse a clearer style to issue commands to remote helpers
Daniel Barkalow [Fri, 4 Sep 2009 02:13:51 +0000 (22:13 -0400)]
Use a clearer style to issue commands to remote helpers

This style is overkill for some commands, but it's worthwhile to use
the same style to issue all commands, and it's useful to avoid
open-coding string lengths.

Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMake the "traditionally-supported" URLs a special case
Daniel Barkalow [Fri, 4 Sep 2009 02:13:49 +0000 (22:13 -0400)]
Make the "traditionally-supported" URLs a special case

Instead of trying to make http://, https://, and ftp:// URLs
indicative of some sort of pattern of transport helper usage, make
them a special case which runs the "curl" helper, and leave the
mechanism by which arbitrary helpers will be chosen entirely to future
work.

Signed-off-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoAdd script for importing bits-and-pieces to Git.
Peter Krefting [Thu, 3 Sep 2009 12:15:00 +0000 (13:15 +0100)]
Add script for importing bits-and-pieces to Git.

Allows the user to import version history that is stored in bits and
pieces in the file system, for instance snapshots of old development
trees, or day-by-day backups. A configuration file is used to
describe the relationship between the different files and allow
describing branches and merges, as well as authorship and commit
messages.

Output is created in a format compatible with git-fast-import.

Full documentation is provided inline in perldoc format.

Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoimport-tars: Allow per-tar author and commit message.
Peter Krefting [Thu, 3 Sep 2009 12:15:00 +0000 (13:15 +0100)]
import-tars: Allow per-tar author and commit message.

If the "--metainfo=<ext>" option is given on the command line, a file
called "<filename.tar>.<ext>" will be used to create the commit message
for "<filename.tar>", instead of using "Imported from filename.tar".

The author and committer of the tar ball can also be overridden by
embedding an "Author:" or "Committer:" header in the metainfo file.

Signed-off-by: Peter Krefting <peter@softwolves.pp.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Thu, 3 Sep 2009 16:43:08 +0000 (09:43 -0700)]
Merge branch 'maint'

* maint:
  git-clone: add missing comma in --reference documentation

15 years agoMerge branch 'maint-1.6.3' into maint
Junio C Hamano [Thu, 3 Sep 2009 16:42:56 +0000 (09:42 -0700)]
Merge branch 'maint-1.6.3' into maint

* maint-1.6.3:
  git-clone: add missing comma in --reference documentation
  git-cvsserver: no longer use deprecated 'git-subcommand' commands
  clone: disconnect transport after fetching

15 years agoMerge branch 'maint-1.6.2' into maint-1.6.3
Junio C Hamano [Thu, 3 Sep 2009 16:42:38 +0000 (09:42 -0700)]
Merge branch 'maint-1.6.2' into maint-1.6.3

* maint-1.6.2:
  git-clone: add missing comma in --reference documentation
  clone: disconnect transport after fetching

15 years agogit-clone: add missing comma in --reference documentation
Miklos Vajna [Thu, 3 Sep 2009 11:24:16 +0000 (13:24 +0200)]
git-clone: add missing comma in --reference documentation

Signed-off-by: Miklos Vajna <vmiklos@frugalware.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint'
Junio C Hamano [Thu, 3 Sep 2009 02:52:18 +0000 (19:52 -0700)]
Merge branch 'maint'

* maint:
  git-cvsserver: no longer use deprecated 'git-subcommand' commands
  clone: disconnect transport after fetching

15 years agoMerge branch 'maint-1.6.3' into maint
Junio C Hamano [Thu, 3 Sep 2009 02:51:55 +0000 (19:51 -0700)]
Merge branch 'maint-1.6.3' into maint

* maint-1.6.3:
  git-cvsserver: no longer use deprecated 'git-subcommand' commands
  clone: disconnect transport after fetching

15 years agopush: teach --quiet to suppress "Everything up-to-date"
Jeff King [Mon, 31 Aug 2009 19:28:34 +0000 (15:28 -0400)]
push: teach --quiet to suppress "Everything up-to-date"

This should have been part of 481c7a6, whose goal was to
make "git push -q" silent unless there is an error.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoMerge branch 'maint-1.6.2' into maint-1.6.3
Junio C Hamano [Thu, 3 Sep 2009 01:45:44 +0000 (18:45 -0700)]
Merge branch 'maint-1.6.2' into maint-1.6.3

* maint-1.6.2:
  clone: disconnect transport after fetching

15 years agogit-cvsserver: no longer use deprecated 'git-subcommand' commands
Gerrit Pape [Wed, 2 Sep 2009 09:23:10 +0000 (09:23 +0000)]
git-cvsserver: no longer use deprecated 'git-subcommand' commands

git-cvsserver still references git commands like 'git-config', which
is depcrecated.  This commit changes git-cvsserver to use the
'git subcommand' form.

Sylvain Beucler reported the problem through
 http://bugs.debian.org/536067

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agoclone: disconnect transport after fetching
Jeff King [Wed, 2 Sep 2009 06:36:47 +0000 (02:36 -0400)]
clone: disconnect transport after fetching

The current code just leaves the transport in whatever state
it was in after performing the fetch.  For a non-empty clone
over the git protocol, the transport code already
disconnects at the end of the fetch.

But for an empty clone, we leave the connection hanging, and
eventually close the socket when clone exits. This causes
the remote upload-pack to complain "the remote end hung up
unexpectedly". While this message is harmless to the clone
itself, it is unnecessarily scary for a user to see and may
pollute git-daemon logs.

This patch just explicitly calls disconnect after we are
done with the remote end, which sends a flush packet to
upload-pack and cleanly disconnects, avoiding the error
message.

Other transports are unaffected or slightly improved:

 - for a non-empty repo over the git protocol, the second
   disconnect is a no-op (since we are no longer connected)

 - for "walker" transports (like HTTP or FTP), we actually
   free some used memory (which previously just sat until
   the clone process exits)

 - for "rsync", disconnect is always a no-op anyway

Signed-off-by: Jeff King <peff@peff.net>
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
15 years agostatus: list unmerged files much later
Johannes Sixt [Tue, 1 Sep 2009 20:13:53 +0000 (22:13 +0200)]
status: list unmerged files much later

When resolving a conflicted merge, two lists in the status output need
more attention from the user than other parts.

 - the list of updated paths is useful to review the amount of changes the
   merge brings in (the user cannot do much about them other than
   reviewing, though); and

 - the list of unmerged paths needs the most attention from the user; the
   user needs to resolve them in order to proceed.

Since the output of git status does not by default go through the pager,
the early parts of the output can scroll away at the top. It is better to
put the more important information near the bottom.  During a merge, local
changes that are not in the index are minimum, and you should keep the
untracked list small in any case, so moving the unmerged list from the top
of the output to immediately after the list of updated paths would give us
the optimum layout.

Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>