git/git.git
13 years agoMerge branch 'js/rebase-detached'
Junio C Hamano [Wed, 14 Nov 2007 22:04:06 +0000 (14:04 -0800)]
Merge branch 'js/rebase-detached'

* js/rebase-detached:
  rebase: fix "rebase --continue" breakage
  rebase: operate on a detached HEAD

13 years agoMerge branch 'rs/pretty'
Junio C Hamano [Wed, 14 Nov 2007 22:03:50 +0000 (14:03 -0800)]
Merge branch 'rs/pretty'

* rs/pretty:
  Fix preprocessor logic that determines the availablity of strchrnul().
  Simplify strchrnul() compat code
  --format=pretty: avoid calculating expensive expansions twice
  add strbuf_adddup()
  --pretty=format: parse commit message only once
  --pretty=format: on-demand format expansion
  Add strchrnul()

13 years agoMerge branch 'rr/cvsexportcommit-w'
Junio C Hamano [Wed, 14 Nov 2007 22:03:40 +0000 (14:03 -0800)]
Merge branch 'rr/cvsexportcommit-w'

* rr/cvsexportcommit-w:
  cvsexportcommit: Add switch to specify CVS workdir

13 years agoMerge branch 'gh/cvsimport-user'
Junio C Hamano [Wed, 14 Nov 2007 22:03:27 +0000 (14:03 -0800)]
Merge branch 'gh/cvsimport-user'

* gh/cvsimport-user:
  git-cvsimport: fix handling of user name when it is not set in CVSROOT

13 years agouser-manual: minor rewording for clarity.
Sergei Organov [Wed, 14 Nov 2007 20:08:15 +0000 (12:08 -0800)]
user-manual: minor rewording for clarity.

Junio screwed up when applying the previous round of the patch;
rewording from "previous" to "old" does make the description
clearer.

Also revert the rewording from head to branch.  The description
is talking about the branch's tip commit and using the word head
is clearer.

Based on input from Sergei and Bruce.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Wed, 14 Nov 2007 11:37:18 +0000 (03:37 -0800)]
Merge branch 'maint'

* maint:
  git-clean: honor core.excludesfile
  Documentation: Fix man page breakage with DocBook XSL v1.72
  git-remote.txt: fix typo
  core-tutorial.txt: Fix argument mistake in an example.
  replace reference to git-rm with git-reset in git-commit doc
  Grammar fixes for gitattributes documentation
  Don't allow fast-import tree delta chains to exceed maximum depth
  revert/cherry-pick: allow starting from dirty work tree.
  t/t3404: fix test for a bogus todo file.

Conflicts:

fast-import.c

13 years agoFix dependencies of parse-options test program
Alex Riesen [Tue, 13 Nov 2007 23:16:36 +0000 (00:16 +0100)]
Fix dependencies of parse-options test program

A stale test-parse-options can break t0040 otherwise.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agouser-manual.txt: fix a few mistakes
Sergei Organov [Tue, 13 Nov 2007 18:19:39 +0000 (21:19 +0300)]
user-manual.txt: fix a few mistakes

Signed-off-by: Sergei Organov <osv@javad.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-clean: honor core.excludesfile
Junio C Hamano [Wed, 14 Nov 2007 09:54:43 +0000 (01:54 -0800)]
git-clean: honor core.excludesfile

git-clean did not honor core.excludesfile configuration
variable, although some other commands such as git-add and
git-status did.  Fix this inconsistency.

Original report and patch from Shun'ichi Fuji.  Rewritten by me
and bugs and tests are mine.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDocumentation: Fix man page breakage with DocBook XSL v1.72
Jonas Fonseca [Wed, 14 Nov 2007 09:38:46 +0000 (10:38 +0100)]
Documentation: Fix man page breakage with DocBook XSL v1.72

From version 1.72 it will replace all dots in roff requests with U+2302
("house" character), and add escaping in output for all instances of dot
that are not in roff requests. This caused the ".ft" hack forcing
monospace font in listingblocks to end up as "\&.ft" and being visible
in the resulting man page.

The fix adds a DOCBOOK_XSL_172 build variable that will disable the
hack. To allow this variable to be defined in config.mak it also moves
build variable handling below the inclusion of config.mak.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoHandle broken vsnprintf implementations in strbuf
Shawn O. Pearce [Tue, 13 Nov 2007 08:22:44 +0000 (03:22 -0500)]
Handle broken vsnprintf implementations in strbuf

Solaris 9's vsnprintf implementation returns -1 if we pass it a
buffer of length 0.  The only way to get it to give us the actual
length necessary for the formatted string is to grow the buffer
out to have at least 1 byte available in the strbuf and then ask
it to compute the length.

If the available space is 0 I'm growing it out by 64 to ensure
we will get an accurate length estimate from all implementations.
Some callers may need to grow the strbuf again but 64 should be a
reasonable enough initial growth.

We also no longer silently fail to append to the string when we are
faced with a broken vsnprintf implementation.  On Solaris 9 this
silent failure caused me to no longer be able to execute "git clone"
as we tried to exec the empty string rather than "git-clone".

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-remote.txt: fix typo
Sergei Organov [Tue, 13 Nov 2007 18:17:47 +0000 (21:17 +0300)]
git-remote.txt: fix typo

Signed-off-by: Sergei Organov <osv@javad.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocore-tutorial.txt: Fix argument mistake in an example.
Sergei Organov [Sat, 10 Nov 2007 13:17:33 +0000 (16:17 +0300)]
core-tutorial.txt: Fix argument mistake in an example.

One of examples has wrong output given the arguments provided.
Fix arguments to match the output.

Fix a minor syntax mistake in another place.

Signed-off-by: Sergei Organov <osv@javad.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoreplace reference to git-rm with git-reset in git-commit doc
Jing Xue [Mon, 12 Nov 2007 04:43:00 +0000 (23:43 -0500)]
replace reference to git-rm with git-reset in git-commit doc

The message in git-commit suggesting to use 'git rm --cached'
to unstage is just plain wrong. It really should mention 'git reset'.

Suggested by Jan Hudec.

Signed-off-by: Jing Xue <jingxue@digizenstudio.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoGrammar fixes for gitattributes documentation
Wincent Colaiuta [Wed, 14 Nov 2007 07:51:41 +0000 (08:51 +0100)]
Grammar fixes for gitattributes documentation

Tweak the "filter" section of the gitattributes documentation to add
some
missing articles and improve some word choices without changing the
semantics of the section.

Signed-off-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDon't allow fast-import tree delta chains to exceed maximum depth
Shawn O. Pearce [Wed, 14 Nov 2007 04:48:42 +0000 (23:48 -0500)]
Don't allow fast-import tree delta chains to exceed maximum depth

Brian Downing noticed fast-import can produce tree depths of up
to 6,035 objects and even deeper.  Long delta chains can create
very small packfiles but cause problems during repacking as git
needs to unpack each tree to count the reachable blobs.

What's happening here is the active branch cache isn't big enough.
We're swapping out the branch and thus recycling the tree information
(struct tree_content) back into the free pool.  When we later reload
the tree we set the delta_depth to 0 but we kept the tree we just
reloaded as a delta base.

So if the tree we reloaded was already at the maximum depth we
wouldn't know it and make the new tree a delta.  Multiply the
number of times the branch cache has to swap out the tree times
max_depth (10) and you get the maximum delta depth of a tree created
by fast-import.  In Brian's case above the active branch cache had
to swap the branch out 603/604 times during this import to produce
a tree with a delta depth of 6035.

Acked-by: Brian Downing <bdowning@lavos.net>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoResurrect git-revert.sh example and add comment to builtin-revert.c
Junio C Hamano [Tue, 13 Nov 2007 21:45:11 +0000 (13:45 -0800)]
Resurrect git-revert.sh example and add comment to builtin-revert.c

I had to scratch my head for quite some time figuring out why we
cannot optimize out write_tree() we do when --no-commit option
is given, whose purpose seem to be only to check if the index is
unmerged, with a simple loop over the active_cache[].

So add a comment to describe why the write_tree() is there, and
resurrect the last scripted version as a reference material in
contrib/example directory with others.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorevert/cherry-pick: allow starting from dirty work tree.
Junio C Hamano [Tue, 13 Nov 2007 20:28:53 +0000 (12:28 -0800)]
revert/cherry-pick: allow starting from dirty work tree.

There is no reason to forbid a dirty work tree when reverting or
cherry-picking a change, as long as the index is clean.

The scripted version used to allow it:

    case "$no_commit" in
    t)
     # We do not intend to commit immediately.  We just want to
     # merge the differences in.
     head=$(git-write-tree) ||
     die "Your index file is unmerged."
     ;;
    *)
     head=$(git-rev-parse --verify HEAD) ||
     die "You do not have a valid HEAD"
     files=$(git-diff-index --cached --name-only $head) || exit
     if [ "$files" ]; then
     die "Dirty index: cannot $me (dirty: $files)"
     fi
     ;;
    esac

but C rewrite tightened the check, probably by mistake.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agot/t3404: fix test for a bogus todo file.
Junio C Hamano [Tue, 13 Nov 2007 21:05:50 +0000 (13:05 -0800)]
t/t3404: fix test for a bogus todo file.

The test wants to see if there are still remaining tasks, but checked
a wrong file.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoFix preprocessor logic that determines the availablity of strchrnul().
Johannes Sixt [Mon, 12 Nov 2007 10:09:05 +0000 (11:09 +0100)]
Fix preprocessor logic that determines the availablity of strchrnul().

Apart from the error in the condition (&& should actually be ||), the
construct

    #if !defined(A) || !A

leads to a syntax error in the C preprocessor if A is indeed not defined.

Tested-by: David Symonds <dsymonds@gmail.com>
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-clean: Fix error message if clean.requireForce is not set.
Johannes Sixt [Mon, 12 Nov 2007 08:27:35 +0000 (09:27 +0100)]
git-clean: Fix error message if clean.requireForce is not set.

It was distracting to see this error message:

     clean.requireForce set and -n or -f not given; refusing to clean

even though clean.requireForce was not set at all. This patch distinguishes
the cases and gives a different message depending on whether the
configuration variable is not set or set to true.

While we are here, we also divert the error messages to stderr.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorebase: fix "rebase --continue" breakage
Johannes Schindelin [Mon, 12 Nov 2007 13:11:46 +0000 (13:11 +0000)]
rebase: fix "rebase --continue" breakage

The --skip case was handled properly when rebasing without --merge,
but the --continue case was not.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-svn: support for funky branch and project names over HTTP(S)
Eric Wong [Mon, 12 Nov 2007 07:37:42 +0000 (23:37 -0800)]
git-svn: support for funky branch and project names over HTTP(S)

SVN requires that paths be URI-escaped for HTTP(S) repositories.
file:// and svn:// repositories do not need these rules.

Additionally, accessing individual paths inside repositories
(check_path() and get_log() do NOT require escapes to function
and in fact it breaks things).

Noticed-by: Michael J. Cohen <mjc@cruiseplanners.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDocumentation: Fix references to deprecated commands
Jonas Fonseca [Mon, 12 Nov 2007 00:32:51 +0000 (01:32 +0100)]
Documentation: Fix references to deprecated commands

... by changing git-tar-tree reference to git-archive and removing
seemingly unrelevant footnote about git-ssh-{fetch,upload}.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Mon, 12 Nov 2007 08:14:15 +0000 (00:14 -0800)]
Merge branch 'maint'

* maint:
  for-each-ref: fix off by one read.
  git-branch: remove mention of non-existent '-b' option
  git-svn: prevent dcommitting if the index is dirty.
  Fix memory leak in traverse_commit_list

13 years agoMerge branch 'sp/maint-plug-traverse-commit-list-leak' into maint
Junio C Hamano [Mon, 12 Nov 2007 08:00:00 +0000 (00:00 -0800)]
Merge branch 'sp/maint-plug-traverse-commit-list-leak' into maint

* sp/maint-plug-traverse-commit-list-leak:
  Fix memory leak in traverse_commit_list

13 years agofor-each-ref: fix off by one read.
Christian Couder [Mon, 12 Nov 2007 04:37:25 +0000 (05:37 +0100)]
for-each-ref: fix off by one read.

Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-branch: remove mention of non-existent '-b' option
Jeff King [Mon, 12 Nov 2007 04:07:05 +0000 (23:07 -0500)]
git-branch: remove mention of non-existent '-b' option

This looks like a cut and paste error from the git-checkout
explanation of --no-track.

Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-svn: prevent dcommitting if the index is dirty.
Benoit Sigoure [Sun, 11 Nov 2007 18:41:41 +0000 (19:41 +0100)]
git-svn: prevent dcommitting if the index is dirty.

dcommit uses rebase to sync the history with what has just been pushed to
SVN.  Trying to dcommit with a dirty index is troublesome for rebase, so now
the user will get an error message if he attempts to dcommit with a dirty
index.

Signed-off-by: Benoit Sigoure <tsuna@lrde.epita.fr>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'js/upload-pack'
Junio C Hamano [Sun, 11 Nov 2007 23:19:57 +0000 (15:19 -0800)]
Merge branch 'js/upload-pack'

* js/upload-pack:
  upload-pack: Use finish_{command,async}() instead of waitpid().

13 years agoMerge branch 'js/reset'
Junio C Hamano [Sun, 11 Nov 2007 23:19:24 +0000 (15:19 -0800)]
Merge branch 'js/reset'

* js/reset:
  builtin-reset: avoid forking "update-index --refresh"
  builtin-reset: do not call "ls-files --unmerged"

13 years agoMerge branch 'js/parseopt-abbrev-fix'
Junio C Hamano [Sun, 11 Nov 2007 23:12:06 +0000 (15:12 -0800)]
Merge branch 'js/parseopt-abbrev-fix'

* js/parseopt-abbrev-fix:
  parse-options: abbreviation engine fix.

13 years agoMerge branch 'maint'
Junio C Hamano [Sun, 11 Nov 2007 23:00:05 +0000 (15:00 -0800)]
Merge branch 'maint'

* maint:
  fix index-pack with packs >4GB containing deltas on 32-bit machines
  git-hash-object should honor config variables
  gitweb: correct month in date display for atom feeds

13 years agopush: teach push to pass --verbose option to transport layer
Steffen Prohaska [Sun, 11 Nov 2007 14:01:44 +0000 (15:01 +0100)]
push: teach push to pass --verbose option to transport layer

A --verbose option to push should also be passed to the
transport layer, i.e. git-send-pack, git-http-push.

git push is modified to do so.

Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agopush: mention --verbose option in documentation
Steffen Prohaska [Sun, 11 Nov 2007 14:01:43 +0000 (15:01 +0100)]
push: mention --verbose option in documentation

Before this commit, only '-v' was documented.

Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoSimplify strchrnul() compat code
Andreas Ericsson [Sat, 10 Nov 2007 11:55:48 +0000 (12:55 +0100)]
Simplify strchrnul() compat code

strchrnul() was introduced in glibc in April 1999 and included in
glibc-2.1. Checking for that version means the majority of all git
users would get to use the optimized version in glibc. Of the
remaining few some might get to use a slightly slower version
than necessary but probably not slower than what we have today.

Unfortunately, __GLIBC_PREREQ() macro was not available in glibc 2.1.1
which was short lived but already supported strchrnul().  Odd minority
users of that library needs to live with our compatibility inline version.

Rediffed-against-next-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agofix index-pack with packs >4GB containing deltas on 32-bit machines
Nicolas Pitre [Sun, 11 Nov 2007 04:29:10 +0000 (23:29 -0500)]
fix index-pack with packs >4GB containing deltas on 32-bit machines

This probably hasn't been properly tested before.  Here's a script to
create a 8GB repo with the necessary characteristics (copy the
test-genrandom executable from the Git build tree to /tmp first):

-----
#!/bin/bash

git init
git config core.compression 0

# create big objects with no deltas
for i in $(seq -w 1 2 63)
do
echo $i
/tmp/test-genrandom $i 268435456 > file_$i
git add file_$i
rm file_$i
echo "file_$i -delta" >> .gitattributes
done

# create "deltifiable" objects in between big objects
for i in $(seq -w 2 2 64)
do
echo "$i $i $i" >> grow
cp grow file_$i
git add file_$i
rm file_$i
done
rm grow

# create a pack with them
git commit -q -m "commit of big objects interlaced with small deltas"
git repack -a -d
-----

Then clone this repo over the Git protocol.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-hash-object should honor config variables
Nicolas Pitre [Sat, 10 Nov 2007 20:00:33 +0000 (15:00 -0500)]
git-hash-object should honor config variables

... such as core.compression.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years ago--format=pretty: avoid calculating expensive expansions twice
René Scharfe [Sat, 10 Nov 2007 11:18:26 +0000 (12:18 +0100)]
--format=pretty: avoid calculating expensive expansions twice

As Jeff King remarked, format strings with duplicate placeholders can
be slow to expand, because each instance is calculated anew.

This patch makes use of the fact that format_commit_message() and its
helper functions only ever add stuff to the end of the strbuf.  For
certain expensive placeholders, store the offset and length of their
expansion with the strbuf at the first occurrence.  Later they
expansion result can simply be copied from there -- no malloc() or
strdup() required.

These certain placeholders are the abbreviated commit, tree and
parent hashes, as the search for a unique abbreviated hash is quite
costly.  Here are the times for next (best of three runs):

$ time git log --pretty=format:%h >/dev/null

real    0m0.611s
user    0m0.404s
sys     0m0.204s

$ time git log --pretty=format:%h%h%h%h >/dev/null

real    0m1.206s
user    0m0.744s
sys     0m0.452s

And here those with this patch (and the previous two); the speedup
of the single placeholder case is just noise:

$ time git log --pretty=format:%h >/dev/null

real    0m0.608s
user    0m0.416s
sys     0m0.192s

$ time git log --pretty=format:%h%h%h%h >/dev/null

real    0m0.639s
user    0m0.488s
sys     0m0.140s

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoadd strbuf_adddup()
René Scharfe [Sat, 10 Nov 2007 11:16:05 +0000 (12:16 +0100)]
add strbuf_adddup()

Add a new function, strbuf_adddup(), that appends a duplicate of a
part of a struct strbuf to end of the latter.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years ago--pretty=format: parse commit message only once
René Scharfe [Sat, 10 Nov 2007 11:14:20 +0000 (12:14 +0100)]
--pretty=format: parse commit message only once

As Jeff King pointed out, some placeholder expansions are related to
each other: the steps to calculate one go most of the way towards
calculating the other, too.

This patch makes format_commit_message() parse the commit message
only once, remembering the position of each item.  This speeds up
handling of format strings containing multiple placeholders from the
set %s, %a*, %c*, %e, %b.

Here are the timings for the git version in next.  The first one is
to estimate the overhead of the caching, the second one is taken
from http://svn.tue.mpg.de/tentakel/trunk/tentakel/Makefile as an
example of a format string found in the wild.  The times are the
fastest of three consecutive runs in each case:

$ time git log --pretty=format:%e >/dev/null

real    0m0.381s
user    0m0.340s
sys     0m0.024s

$ time git log --pretty=format:"* %cd %cn%n%n%s%n%b" >/dev/null

real    0m0.623s
user    0m0.556s
sys     0m0.052s

And here the times with this patch:

$ time git log --pretty=format:%e >/dev/null

real    0m0.385s
user    0m0.332s
sys     0m0.040s

$ time git log --pretty=format:"* %cd %cn%n%n%s%n%b" >/dev/null

real    0m0.563s
user    0m0.504s
sys     0m0.048s

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogitweb: correct month in date display for atom feeds
Vincent Zanotti [Sat, 10 Nov 2007 18:55:27 +0000 (19:55 +0100)]
gitweb: correct month in date display for atom feeds

Signed-off-by: Vincent Zanotti <vincent.zanotti@m4x.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agotest-lib.sh: move error line after error() declaration
Michele Ballabio [Sat, 10 Nov 2007 14:17:25 +0000 (15:17 +0100)]
test-lib.sh: move error line after error() declaration

This patch removes a spurious "command not found" error
and actually makes the "Test script did not set test_description."
string follow the command line option "--no-color".

Signed-off-by: Michele Ballabio <barra_cuda@katamail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agofor-each-ref: fix setup of option-parsing for --sort
Lars Hjemli [Sat, 10 Nov 2007 16:47:54 +0000 (17:47 +0100)]
for-each-ref: fix setup of option-parsing for --sort

The option value for --sort is already a pointer to a pointer to struct
ref_sort, so just use it.

Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Sat, 10 Nov 2007 10:06:30 +0000 (02:06 -0800)]
Merge branch 'maint'

* maint:
  print warning/error/fatal messages in one shot

13 years agocore-tutorial.txt: Fix git-show-branch example and its description
Sergei Organov [Thu, 8 Nov 2007 15:10:28 +0000 (18:10 +0300)]
core-tutorial.txt: Fix git-show-branch example and its description

Signed-off-by: Sergei Organov <osv@javad.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoFix memory leak in traverse_commit_list
Shawn O. Pearce [Fri, 9 Nov 2007 11:06:10 +0000 (06:06 -0500)]
Fix memory leak in traverse_commit_list

If we were listing objects too then the objects were buffered in an
array only reachable from a stack allocated structure.  When this
function returns that array would be leaked as nobody would have
a reference to it anymore.

Historically this hasn't been a problem as the primary user of
traverse_commit_list() (the noble git-rev-list) would terminate
as soon as the function was finished, thus allowing the operating
system to cleanup memory.  However we have been leaking this data
in git-pack-objects ever since that program learned how to run the
revision listing internally, rather than relying on reading object
names from git-rev-list.

To better facilitate reuse of traverse_commit_list during other
builtin tools (such as git-fetch) we shouldn't leak temporary memory
like this and instead we need to clean up properly after ourselves.

Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-commit: a bit more tests
Junio C Hamano [Sat, 10 Nov 2007 09:49:34 +0000 (01:49 -0800)]
git-commit: a bit more tests

Add tests for -s (sign-off) and multiple -m options

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoprint warning/error/fatal messages in one shot
Nicolas Pitre [Fri, 9 Nov 2007 19:33:54 +0000 (14:33 -0500)]
print warning/error/fatal messages in one shot

Not doing so is likely to create a messed up display when sent over the
sideband protocol.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agobuiltin-blame: set up the work_tree before the first file access
Johannes Schindelin [Fri, 9 Nov 2007 11:34:07 +0000 (11:34 +0000)]
builtin-blame: set up the work_tree before the first file access

We check in cmd_blame() if the specified path is there, but we
failed to set up the working tree before that.

While at it, make setup_work_tree() just return if it was run
before.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoadd a howto document about corrupted blob recovery
Nicolas Pitre [Fri, 9 Nov 2007 17:28:19 +0000 (12:28 -0500)]
add a howto document about corrupted blob recovery

Extracted from a post by Linus on the mailing list.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDocumentation: remove documentation for removed tools.
Junio C Hamano [Fri, 9 Nov 2007 01:22:44 +0000 (17:22 -0800)]
Documentation: remove documentation for removed tools.

Old commit walkers other than http/curl transport have been removed
for some time now.  Remove their documents.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMake check-docs target detect removed commands
Junio C Hamano [Fri, 9 Nov 2007 02:38:27 +0000 (18:38 -0800)]
Make check-docs target detect removed commands

The maintainer should remember running "make check-docs" from
time to time.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDocumentation: lost-found is now deprecated.
Junio C Hamano [Fri, 9 Nov 2007 01:21:48 +0000 (17:21 -0800)]
Documentation: lost-found is now deprecated.

This makes it possible to mark commands that are deprecated in the
command list of the primary manual page git(7), and uses it to
mark "git lost-found" and "git tar-tree" as deprecated.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorebase: operate on a detached HEAD
Johannes Schindelin [Thu, 8 Nov 2007 18:19:08 +0000 (18:19 +0000)]
rebase: operate on a detached HEAD

The interactive version of rebase does all the operations on a detached
HEAD, so that after a successful rebase, <branch>@{1} is the pre-rebase
state.  The reflogs of "HEAD" still show all the actions in detail.

This teaches the non-interactive version to do the same.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years ago--pretty=format: on-demand format expansion
René Scharfe [Fri, 9 Nov 2007 00:49:42 +0000 (01:49 +0100)]
--pretty=format: on-demand format expansion

Some of the --pretty=format placeholders expansions are expensive to
calculate.  This is made worse by the current code's use of
interpolate(), which requires _all_ placeholders are to be prepared
up front.

One way to speed this up is to check which placeholders are present
in the format string and to prepare only the expansions that are
needed.  That still leaves the allocation overhead of interpolate().

Another way is to use a callback based approach together with the
strbuf library to keep allocations to a minimum and avoid string
copies.  That's what this patch does.  It introduces a new strbuf
function, strbuf_expand().

The function takes a format string, list of placeholder strings,
a user supplied function 'fn', and an opaque pointer 'context'
to tell 'fn' what thingy to operate on.

The function 'fn' is expected to accept a strbuf, a parsed
placeholder string and the 'context' pointer, and append the
interpolated value for the 'context' thingy, according to the
format specified by the placeholder.

Thanks to Pierre Habouzit for his suggestion to use strchrnul() and
the code surrounding its callsite.  And thanks to Junio for most of
this commit message. :)

Here my measurements of most of Paul Mackerras' test cases that
highlighted the performance problem (best of three runs):

(master)
$ time git log --pretty=oneline >/dev/null

real    0m0.390s
user    0m0.340s
sys     0m0.040s

(master)
$ time git log --pretty=raw >/dev/null

real    0m0.434s
user    0m0.408s
sys     0m0.016s

(master)
$ time git log --pretty="format:%H {%P} %ct" >/dev/null

real    0m1.347s
user    0m0.080s
sys     0m1.256s

(interp_find_active -- Dscho)
$ time ./git log --pretty="format:%H {%P} %ct" >/dev/null

real    0m0.694s
user    0m0.020s
sys     0m0.672s

(strbuf_expand -- this patch)
$ time ./git log --pretty="format:%H {%P} %ct" >/dev/null

real    0m0.395s
user    0m0.352s
sys     0m0.028s

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoAdd strchrnul()
René Scharfe [Fri, 9 Nov 2007 00:49:36 +0000 (01:49 +0100)]
Add strchrnul()

As suggested by Pierre Habouzit, add strchrnul().  It's a useful GNU
extension and can simplify string parser code.  There are several
places in git that can be converted to strchrnul(); as a trivial
example, this patch introduces its usage to builtin-fetch--tool.c.

Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-cvsimport: fix handling of user name when it is not set in CVSROOT
Gordon Hopper [Thu, 8 Nov 2007 20:15:20 +0000 (13:15 -0700)]
git-cvsimport: fix handling of user name when it is not set in CVSROOT

The cvs programs do not default to "anonymous" as the user name, but use the
currently logged in user.  This patch more closely matches the cvs behavior.

Signed-off-by: Gordon Hopper <g.hopper@computer.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoUpdate draft release notes for 1.5.4
Junio C Hamano [Fri, 9 Nov 2007 08:32:38 +0000 (00:32 -0800)]
Update draft release notes for 1.5.4

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'maint'
Junio C Hamano [Fri, 9 Nov 2007 08:21:44 +0000 (00:21 -0800)]
Merge branch 'maint'

* maint:
  Start preparing for 1.5.3.6
  git-send-email: Change the prompt for the subject of the initial message.
  SubmittingPatches: improve the 'Patch:' section of the checklist
  instaweb: Minor cleanups and fixes for potential problems
  stop t1400 hiding errors in tests
  Makefile: add missing dependency on wt-status.h
  refresh_index_quietly(): express "optional" nature of index writing better
  Fix sed string regex escaping in module_name.
  Avoid a few unportable, needlessly nested "...`...".
  git-mailsplit: with maildirs not only process cur/, but also new/

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-bisect.sh: Fix sed script to work with AIX and BSD sed.
Ralf Wildenhues [Thu, 8 Nov 2007 21:48:24 +0000 (22:48 +0100)]
git-bisect.sh: Fix sed script to work with AIX and BSD sed.

\n is not portable in a s/// replacement string, only
in the regex part.  backslash-newline helps.

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoStart preparing for 1.5.3.6
Junio C Hamano [Fri, 9 Nov 2007 08:17:26 +0000 (00:17 -0800)]
Start preparing for 1.5.3.6

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-send-email: Change the prompt for the subject of the initial message.
Benoit Sigoure [Thu, 8 Nov 2007 18:56:28 +0000 (19:56 +0100)]
git-send-email: Change the prompt for the subject of the initial message.

I never understood what this prompt was asking for until I read the actual
source code.  I think this wording is much more understandable.

Signed-off-by: Benoit Sigoure <tsuna@lrde.epita.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoSubmittingPatches: improve the 'Patch:' section of the checklist
Sergei Organov [Thu, 8 Nov 2007 16:40:25 +0000 (19:40 +0300)]
SubmittingPatches: improve the 'Patch:' section of the checklist

There were 2 items "send patch to..." but having different set of
addresses to send patch to. Merge them together and move the resulting
item to the end of checklist.

Signed-off-by: Sergei Organov <osv@javad.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoinstaweb: Minor cleanups and fixes for potential problems
Jonas Fonseca [Thu, 8 Nov 2007 23:21:42 +0000 (00:21 +0100)]
instaweb: Minor cleanups and fixes for potential problems

Fix path quoting and test of empty values that some shells do not like.
Remove duplicate check and setting of $browser.

Signed-off-by: Jonas Fonseca <fonseca@diku.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agostop t1400 hiding errors in tests
Alex Riesen [Thu, 8 Nov 2007 23:41:39 +0000 (00:41 +0100)]
stop t1400 hiding errors in tests

The last rm in the test was lacking an "&&" before it,
which caused the errors in the commands be silently hidden.

Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMakefile: add missing dependency on wt-status.h
Junio C Hamano [Fri, 9 Nov 2007 00:41:56 +0000 (16:41 -0800)]
Makefile: add missing dependency on wt-status.h

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorefresh_index_quietly(): express "optional" nature of index writing better
Junio C Hamano [Fri, 9 Nov 2007 00:24:00 +0000 (16:24 -0800)]
refresh_index_quietly(): express "optional" nature of index writing better

The point of the part of the code this patch touches is that if
we modified the active_cache, we try to write it out and make it
the index file for later users to use by calling
"commit_locked_index", but we do not really care about the
failure from this sequence because it is done purely as an
optimization.

The original code called three functions primarily for their
side effects but as condition of an if statement, which is
admittedly a bad style.

Incidentally, it squelches an "empty if body" warning from gcc.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoFix sed string regex escaping in module_name.
Ralf Wildenhues [Thu, 8 Nov 2007 21:48:49 +0000 (22:48 +0100)]
Fix sed string regex escaping in module_name.

When escaping a string to be used as a sed regex, it is important
to only escape active characters.  Escaping other characters is
undefined according to POSIX, and in practice leads to issues with
extensions such as GNU sed's \+.

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoAvoid a few unportable, needlessly nested "...`...".
Ralf Wildenhues [Thu, 8 Nov 2007 21:47:36 +0000 (22:47 +0100)]
Avoid a few unportable, needlessly nested "...`...".

Signed-off-by: Ralf Wildenhues <Ralf.Wildenhues@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoStyle: place opening brace of a function definition at column 1
Junio C Hamano [Thu, 8 Nov 2007 23:35:32 +0000 (15:35 -0800)]
Style: place opening brace of a function definition at column 1

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocontrib/hooks/post-receive-email: remove cruft, $committer is not used
Gerrit Pape [Thu, 8 Nov 2007 12:11:57 +0000 (12:11 +0000)]
contrib/hooks/post-receive-email: remove cruft, $committer is not used

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agohooks--update: decline deleting tags or branches by default, add config options
Gerrit Pape [Thu, 8 Nov 2007 09:47:39 +0000 (09:47 +0000)]
hooks--update: decline deleting tags or branches by default, add config options

Decline deleting tags or branches through git push <remote> :<ref> by
default, support config options hooks.allowdeletetag, hooks.allowdeletebranch
to override this per repository.

Before this patch the update hook interpreted deleting a tag, no matter if
annotated or not, through git push <remote> :<tag> as unannotated tag, and
declined it by default, but with an unappropriate error message:

 $ git push origin :atag
 deleting 'refs/tags/atag'
 *** The un-annotated tag, atag, is not allowed in this repository
 *** Use 'git tag [ -a | -s ]' for tags you want to propagate.
 ng refs/tags/atag hook declined
 error: hooks/update exited with error code 1
 error: hook declined to update refs/tags/atag
 error: failed to push to 'monolith:/git/qm/test-repo'

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agohooks--update: fix test for properly set up project description file
Gerrit Pape [Thu, 8 Nov 2007 14:02:00 +0000 (14:02 +0000)]
hooks--update: fix test for properly set up project description file

The update hook template intends to abort if the project description file
hasn't been adjusted or is empty.  This patch fixes the check for 'being
adjusted'.

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-mailsplit: with maildirs not only process cur/, but also new/
Gerrit Pape [Tue, 6 Nov 2007 08:54:18 +0000 (08:54 +0000)]
git-mailsplit: with maildirs not only process cur/, but also new/

When saving patches to a maildir with e.g. mutt, the files are put into
the new/ subdirectory of the maildir, not cur/.  This makes git-am state
"Nothing to do.".  This patch lets git-mailsplit additional check new/
after reading cur/.

This was reported by Joey Hess through
 http://bugs.debian.org/447396

Signed-off-by: Gerrit Pape <pape@smarden.org>
Acked-by: Jeff King <peff@peff.net>
Acked-by: Alex Riesen <raa.lkml@gmail.com>
Acked-by: Fernando J. Pereda <ferdy@gentoo.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMerge branch 'jc/clean-config'
Junio C Hamano [Thu, 8 Nov 2007 02:19:38 +0000 (18:19 -0800)]
Merge branch 'jc/clean-config'

* jc/clean-config:
  clean: require -f to do damage by default

13 years agoMerge branch 'gp/reset-q'
Junio C Hamano [Thu, 8 Nov 2007 02:18:55 +0000 (18:18 -0800)]
Merge branch 'gp/reset-q'

* gp/reset-q:
  git-reset: add -q option to operate quietly

13 years agoMerge branch 'ds/maint-deflatebound'
Junio C Hamano [Thu, 8 Nov 2007 02:17:20 +0000 (18:17 -0800)]
Merge branch 'ds/maint-deflatebound'

* ds/maint-deflatebound:
  Improve accuracy of check for presence of deflateBound.

13 years agoMerge branch 'cp/p4'
Junio C Hamano [Thu, 8 Nov 2007 02:16:18 +0000 (18:16 -0800)]
Merge branch 'cp/p4'

* cp/p4:
  git-p4: Detect changes to executable bit and include them in p4 submit.
  git-p4: Add a helper function to parse the full git diff-tree output.

13 years agoMerge branch 'maint'
Junio C Hamano [Thu, 8 Nov 2007 02:07:07 +0000 (18:07 -0800)]
Merge branch 'maint'

* maint:
  Add Documentation/CodingGuidelines
  When exec() fails include the failing command in the error message
  RelNotes-1.5.3.5: fix another typo

13 years agoMerge branch 'mh/work-tree'
Junio C Hamano [Thu, 8 Nov 2007 01:37:00 +0000 (17:37 -0800)]
Merge branch 'mh/work-tree'

* mh/work-tree:
  Make git-blame fail when working tree is needed and we're not in one
  Don't always require working tree for git-rm
  Use setup_work_tree() in builtin-ls-files.c
  Refactor working tree setup

13 years agoSmall code readability improvement in show_reference() in builtin-tag.c
Mike Hommey [Sat, 3 Nov 2007 13:08:05 +0000 (14:08 +0100)]
Small code readability improvement in show_reference() in builtin-tag.c

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agosend-email: apply --suppress-from to S-o-b and cc-cmd
Uwe Kleine-König [Wed, 7 Nov 2007 07:34:12 +0000 (08:34 +0100)]
send-email: apply --suppress-from to S-o-b and cc-cmd

Signed-off-by: Uwe Kleine-König <Uwe.Kleine-Koenig@digi.com>
Cc: Ryan Anderson <ryan@michonline.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoFix minor nits in configure.ac
Ralf Wildenhues [Tue, 6 Nov 2007 20:12:45 +0000 (21:12 +0100)]
Fix minor nits in configure.ac

Avoid "test -o" as it is only XSI not POSIX, and not portable.
Avoid exit(3) in test programs in favor of return, to accommodate
for newer Autoconf not providing a declaration for exit.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDeprecate git-lost-found
Johannes Schindelin [Thu, 8 Nov 2007 00:41:22 +0000 (00:41 +0000)]
Deprecate git-lost-found

"git fsck" learnt the option "--lost-found" in v1.5.3-rc0~5, to make
"git lost-found" obsolete.  It is time to deprecate "git lost-found".

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoImprove accuracy of check for presence of deflateBound.
David Symonds [Wed, 7 Nov 2007 03:24:28 +0000 (14:24 +1100)]
Improve accuracy of check for presence of deflateBound.

ZLIB_VERNUM isn't defined in some zlib versions, so this patch does a proper
linking test in autoconf to see whether deflateBound exists in zlib. Also,
setting NO_DEFLATE_BOUND will also work for folk not using autoconf.

Signed-off-by: David Symonds <dsymonds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoAdd Documentation/CodingGuidelines
Johannes Schindelin [Thu, 8 Nov 2007 00:33:19 +0000 (00:33 +0000)]
Add Documentation/CodingGuidelines

Even if our code is quite a good documentation for our coding style,
some people seem to prefer a document describing it.

The part about the shell scripts is clearly just copied from one of
Junio's helpful mails, and some parts were added from comments by
Junio, Andreas Ericsson and Robin Rosenberg.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocontrib/hooks/post-receive-email: make subject prefix configurable
Gerrit Pape [Tue, 6 Nov 2007 13:49:30 +0000 (13:49 +0000)]
contrib/hooks/post-receive-email: make subject prefix configurable

Email subjects are prefixed with "[SCM] " by default, make this optionally
configurable through the hooks.emailprefix config option.

Suggested by martin f krafft through
 http://bugs.debian.org/428418

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocontrib/hooks/post-receive-email: reformat to wrap comments at 76 chars
Gerrit Pape [Tue, 6 Nov 2007 13:48:34 +0000 (13:48 +0000)]
contrib/hooks/post-receive-email: reformat to wrap comments at 76 chars

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agocontrib/hooks/post-receive-email: fix typo
Gerrit Pape [Tue, 6 Nov 2007 13:48:07 +0000 (13:48 +0000)]
contrib/hooks/post-receive-email: fix typo

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agorestore fetching with thin-pack capability
Nicolas Pitre [Wed, 7 Nov 2007 22:20:22 +0000 (17:20 -0500)]
restore fetching with thin-pack capability

Broken since commit fa74052922cf39e5a39ad7178d1b13c2da9b4519.

Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoWhen exec() fails include the failing command in the error message
Ask Bjørn Hansen [Tue, 6 Nov 2007 10:54:01 +0000 (02:54 -0800)]
When exec() fails include the failing command in the error message

git-svn occasionally fails with no details as to what went wrong - this should help debug those situations.

Signed-off-by: Ask Bjørn Hansen <ask@develooper.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoRelNotes-1.5.3.5: fix another typo
David D Kilzer [Sun, 4 Nov 2007 12:45:22 +0000 (04:45 -0800)]
RelNotes-1.5.3.5: fix another typo

Signed-off-by: David D Kilzer <ddkilzer@kilzer.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agogit-reset: add -q option to operate quietly
Gerrit Pape [Sun, 4 Nov 2007 09:37:20 +0000 (09:37 +0000)]
git-reset: add -q option to operate quietly

Many git commands have a -q option to suppress output to stdout, let's
have it for git-reset too.

This was asked for by Joey Hess through
 http://bugs.debian.org/444933

Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoMake git-blame fail when working tree is needed and we're not in one
Mike Hommey [Sat, 3 Nov 2007 12:22:55 +0000 (13:22 +0100)]
Make git-blame fail when working tree is needed and we're not in one

Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoDon't always require working tree for git-rm
Mike Hommey [Sat, 3 Nov 2007 11:23:13 +0000 (12:23 +0100)]
Don't always require working tree for git-rm

This allows to do git rm --cached -r directory, instead of
git ls-files -z directory | git update-index --remove -z --stdin.
This can be particularly useful for git-filter-branch users.

Signed-off-by: Mike Hommey <mh@glandium.org>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoUse setup_work_tree() in builtin-ls-files.c
Mike Hommey [Sat, 3 Nov 2007 11:23:12 +0000 (12:23 +0100)]
Use setup_work_tree() in builtin-ls-files.c

Signed-off-by: Mike Hommey <mh@glandium.org>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoRefactor working tree setup
Mike Hommey [Sat, 3 Nov 2007 11:23:11 +0000 (12:23 +0100)]
Refactor working tree setup

Create a setup_work_tree() that can be used from any command requiring
a working tree conditionally.

Signed-off-by: Mike Hommey <mh@glandium.org>
Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agoupload-pack: Use finish_{command,async}() instead of waitpid().
Johannes Sixt [Sun, 4 Nov 2007 19:46:48 +0000 (20:46 +0100)]
upload-pack: Use finish_{command,async}() instead of waitpid().

upload-pack spawns two processes, rev-list and pack-objects, and carefully
monitors their status so that it can report failure to the remote end.
This change removes the complicated procedures on the grounds of the
following observations:

- If everything is OK, rev-list closes its output pipe end, upon which
  pack-objects (which reads from the pipe) sees EOF and terminates itself,
  closing its output (and error) pipes. upload-pack reads from both until
  it sees EOF in both. It collects the exit codes of the child processes
  (which indicate success) and terminates successfully.

- If rev-list sees an error, it closes its output and terminates with
  failure. pack-objects sees EOF in its input and terminates successfully.
  Again upload-pack reads its inputs until EOF. When it now collects
  the exit codes of its child processes, it notices the failure of rev-list
  and signals failure to the remote end.

- If pack-objects sees an error, it terminates with failure. Since this
  breaks the pipe to rev-list, rev-list is killed with SIGPIPE.
  upload-pack reads its input until EOF, then collects the exit codes of
  the child processes, notices their failures, and signals failure to the
  remote end.

- If upload-pack itself dies unexpectedly, pack-objects is killed with
  SIGPIPE, and subsequently also rev-list.

The upshot of this is that precise monitoring of child processes is not
required because both terminate if either one of them dies unexpectedly.
This allows us to use finish_command() and finish_async() instead of
an explicit waitpid(2) call.

The change is smaller than it looks because most of it only reduces the
indentation of a large part of the inner loop.

Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
13 years agobuiltin-reset: avoid forking "update-index --refresh"
Johannes Schindelin [Sat, 3 Nov 2007 15:21:21 +0000 (15:21 +0000)]
builtin-reset: avoid forking "update-index --refresh"

Instead of forking update-index, call refresh_cache() directly.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>