This makes the traversal of index be in sync with the tree traversal.
When unpack_callback() is fed a set of tree entries from trees, it
inspects the name of the entry and checks if the an index entry with
the same name could be hiding behind the current index entry, and
(1) if the name appears in the index as a leaf node, it is also
fed to the n_way_merge() callback function;
(2) if the name is a directory in the index, i.e. there are entries in
that are underneath it, then nothing is fed to the n_way_merge()
callback function;
(3) otherwise, if the name comes before the first eligible entry in the
index, the index entry is first unpacked alone.
When traverse_trees_recursive() descends into a subdirectory, the
cache_bottom pointer is moved to walk index entries within that directory.
All of these are omitted for diff-index, which does not even want to be
fed an index entry and a tree entry with D/F conflicts.
This fixes 3-way read-tree and exposes a bug in other parts of the system
in t6035, test #5. The test prepares these three trees:
O = HEAD^
100644 blob e69de29bb2 a/b-2/c/d
100644 blob e69de29bb2 a/b/c/d
100644 blob e69de29bb2 a/x
A = HEAD
100644 blob e69de29bb2 a/b-2/c/d
100644 blob e69de29bb2 a/b/c/d
100644 blob 587be6b4c3f93f93c489c0111bba5596147a26cb a/x
B = master
120000 blob a36b77384451ea1de7bd340ffca868249626bc52 a/b
100644 blob e69de29bb2 a/b-2/c/d
100644 blob e69de29bb2 a/x
With a clean index that matches HEAD, running
git read-tree -m -u --aggressive $O $A $B
now yields
120000 a36b77384451ea1de7bd340ffca868249626bc52 3 a/b
100644 e69de29bb2 0 a/b-2/c/d
100644 e69de29bb2 1 a/b/c/d
100644 e69de29bb2 2 a/b/c/d
100644 587be6b4c3f93f93c489c0111bba5596147a26cb 0 a/x
which is correct. "master" created "a/b" symlink that did not exist,
and removed "a/b/c/d" while HEAD did not do touch either path.
Before this series, read-tree did not notice the situation and resolved
addition of "a/b" and removal of "a/b/c/d" independently. If A = HEAD had
another path "a/b/c/e" added, this merge should conflict but instead it
silently resolved "a/b" and then immediately overwrote it to add
"a/b/c/e", which was quite bogus.
Tests in t1012 start to work with this.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This prepares but does not yet implement a look-ahead in the index entries
when traverse-trees.c decides to give us tree entries in an order that
does not match what is in the index.
A case where a look-ahead in the index is necessary happens when merging
branch B into branch A while the index matches the current branch A, using
a tree O as their common ancestor, and these three trees looks like this:
O A B
t t
t-i t-i t-i
t-j t-j
t/1
t/2
The traverse_trees() function gets "t", "t-i" and "t" from trees O, A and
B first, and notices that A may have a matching "t" behind "t-i" and "t-j"
(indeed it does), and tells A to give that entry instead. After unpacking
blob "t" from tree B (as it hasn't changed since O in B and A removed it,
it will result in its removal), it descends into directory "t/".
The side that walked index in parallel to the tree traversal used to be
implemented with one pointer, o->pos, that points at the next index entry
to be processed. When this happens, the pointer o->pos still points at
"t-i" that is the first entry. We should be able to skip "t-i" and "t-j"
and locate "t/1" from the index while the recursive invocation of
traverse_trees() walks and match entries found there, and later come back
to process "t-i".
While that look-ahead is not implemented yet, this adds a flag bit,
CE_UNPACKED, to mark the entries in the index that has already been
processed. o->pos pointer has been renamed to o->cache_bottom and it
points at the first entry that may still need to be processed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When rewriting commits on a topic branch, sometimes it is easier to
compare the version of commits before and after the rewrite if they are
based on the same commit that forked from the upstream. An earlier commit
by Junio (fixed up by the previous commit) gives "--onto A...B" syntax to
rebase command, and rebases on top of the merge base between A and B;
teach the same to the interactive version, too.
Signed-off-by: しらいし ななこ <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous patch didn't parse "rebase --onto A...B" correctly when A
isn't an empty string. It also tried to be careful to notice a case in
which there are more than one merge bases, but forgot to give --all option
to merge-base, making the test pointless.
Fix these problems and add a test script to verify. Improvements to the
script to parse A...B syntax was taken from review comments by Johannes
Schindelin.
Signed-off-by: しらいし ななこ <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If switching from an unborn branch (= empty tree) to a valid commit failed
without -m, it would fail with -m option as well.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add another test to set prerequisite EXTGREP if the current build supports
external grep. This can be used to skip external grep only tests on builds
that do not support this optimization.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Providing multiple targets to force a rebuild is unnecessary
complication.
Avoid using a name that could conflict with future special
targets in GNU make (a leading period followed by uppercase
letters).
Cc: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Teach a new option, --autosquash, to the interactive rebase.
When the commit log message begins with "!fixup ...", and there
is a commit whose title begins with the same ..., automatically
modify the todo list of rebase -i so that the commit marked for
squashing come right after the commit to be modified, and change
the action of the moved commit from pick to squash.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We should tell ll_merge() that the 3-way merge between stages #2 and #3 is
an outermost merge, not a virtual-ancestor creation.
Back when this code was originally written, users couldn't write custom
merge drivers easily, so the bug didn't matter, but these days it does.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Providing multiple targets to force a rebuild is unnecessary
complication.
Avoid using a name that could conflict with future special
targets in GNU make (a leading period followed by uppercase
letters).
The corresponding change to the git-gui Makefile is left for
another patch.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'make git.s' to debug code generation of main() fails because
git.c makes use of preprocessor symbols such as GIT_VERSION that
are not set. make does not generate code listings for
builtin_help.c, exec_cmd.c, builtin-init-db.c, config.c, http.c,
or http-walker.c either, for the same reason.
So pass the flags used to generate each .o file when generating
the corresponding assembler listing.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This allows reusing the standard %.o: %.c pattern rule even for
targets that require special flags to be set. Thus after this
change, any changes in the command for compilation only have to
be performed in one place.
Target-specific variables have been supported in GNU make since
version 3.77, which has been available since 1998.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'make var.s' fails to regenerate an assembler listing if var.c
has not changed but a header it includes has:
$ make var.s
CC var.s
$ touch cache.h
$ make var.s
$
The corresponding problem for 'make var.o' does not occur because
the Makefile lists dependencies for each .o target explicitly;
analogous dependency rules for the .s targets are not present.
Rather than add some, it seems better to force 'make' to always
regenerate assembler listings, since the assembler listing
targets are only invoked when specifically requested on the make
command line.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This reuses many of the tests from the old t5560 but runs those tests
without curl or a webserver. This will hopefully increase the testing
coverage for http-backend because it does not require users to set
GIT_TEST_HTTPD.
Signed-off-by: Tarmigan Casebolt <tarmigan+git@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This should introduce no functional change in the tests or the amount
of test coverage.
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Tarmigan Casebolt <tarmigan+git@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 34b6cb8bb ("http-backend: Protect GIT_PROJECT_ROOT from /../
requests") added the path_info helper function to test t5560 but did
not use it. We should use it as it provides another level of error
checking.
The /etc/.../passwd case is one that is not special (and the test
fails for reasons other than being aliased), so we remove that test
case.
Also rename the function from 'path_info' to 'expect_aliased'.
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Tarmigan Casebolt <tarmigan+git@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Similar to how git-daemon checks whether a repository is OK to be
exported, smart-http should also check. This check can be satisfied
in two different ways: the environmental variable GIT_HTTP_EXPORT_ALL
may be set to export all repositories, or the individual repository
may have the file git-daemon-export-ok.
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Tarmigan Casebolt <tarmigan+git@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On CentOS 5, httpd is located at /usr/sbin/httpd, and the modules are
located at /usr/lib64/httpd/modules. To enable easy testing of httpd,
we would like those locations to be detected automatically.
uname might not be the best way to determine the default location for
httpd since different Linux distributions apparently put httpd in
different places, so we test a couple different locations for httpd,
and use the first one that we come across. We do the same for the
modules directory.
cc: Jay Soffian <jaysoffian@gmail.com>
Signed-off-by: Tarmigan Casebolt <tarmigan+git@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Recall that MSYS bash converts POSIX style absolute paths to Windows style
absolute paths. Unfortunately, it converts a program argument that begins
with a double-quote and otherwise looks like an absolute POSIX path, but
in doing so, it strips everything past the second double-quote[*]. This
case is triggered in the two test scripts. The work-around is to place the
Windows style path returned by $(pwd) between the quotes to avoid the path
conversion.
[*] It is already bogus that a conversion is even considered when a program
argument begins with a double-quote because it cannot be an absolute POSIX
path.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is mostly to make it more consistent with the rest of
git, which uses the shell to exec helpers.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently textconv helpers are run directly. Running through
the shell is useful because the user can provide a program
with command line arguments, like "antiword -f".
It also makes textconv more consistent with other parts of
git, most of which run their helpers using the shell.
The downside is that textconv helpers with shell
metacharacters (like space) in the filename will be broken.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that run_command implements the same code in a more
general form, we can make use of it.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If there are no metacharacters in the program to be run, we
can just skip running the shell entirely and directly exec
the program.
The metacharacter test is pulled verbatim from
launch_editor, which already implements this optimization.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that we have the use_shell feature, these callsites can
all be converted with small changes.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On Windows, we need the shbang line to correctly invoke shell scripts via
a POSIX shell, except when the script is invoked via 'sh -c' because sh (a
bash) does "the right thing". But the clean and smudge filters will not
always be invoked via 'sh -c'; to futureproof, we should mark the the one
in t0021-conversion with #!$SHELL_PATH.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The %.o: %.S pattern rule should depend on GIT-CFLAGS to avoid
trouble when ALL_CFLAGS changes.
The pattern only applies to one file (ppc/sha1ppc.S) and that
file does not use any #ifdefs, so leaving the dependency out is
probably harmless. Nevertheless, it is safer to include the
dependency in case future code's behavior does depend on the
build flags.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
and while at it also explain why --merge option is disallowed in some
cases.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the ancestor used to have a blob "P", your tree removed it, and the
tree you are merging with also removed it, the agressive three-way cleanly
merges to remove that blob. If the other tree added a new blob "P/Q"
while removing "P", it should also merge cleanly to remove "P" and create
"P/Q" (since neither the ancestor nor your tree could have had it, so it
is a typical "created in one").
The "aggressive" rule is not new anymore. Reword the stale comment.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
traverse_trees() is supposed to call its callback with all the matching
entries from the given trees. The current algorithm keeps a pointer to
each of the tree being traversed, and feeds the entry with the earliest
name to the callback.
This breaks down if the trees being traversed looks like this:
A B
t-1 t
t-2 u
t/a v
When we are currently looking at an entry "t-1" in tree A, and tree B has
returned "t", feeding "t" from the B and not feeding anything from A, only
because "t-1" sorts later than "t", will miss an entry for a subtree "t"
behind the current entry in tree A.
This introduces extended_entry_extract() helper function that gives what
name is expected from the tree, and implements a mechanism to look-ahead
in the tree object using it, to make sure such a case is handled sanely.
Traversal in tree A in the above example will first return "t" to match
that of B, and then the next request for an entry to A then returns "t-1".
This roughly corresponds to what Linus's "prepare for one-entry lookahead"
wanted to do, but because this does implement look ahead, t6035 and one more
test in t1012 reveal that the approach would not work without adjusting the
side that walks the index in unpack_trees() as well.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 9e8ecea (Add 'merge' mode to 'git reset', 2008-12-01) disallowed
"git reset --merge" when there was unmerged entries. But it wished if
unmerged entries were reset as if --hard (instead of --merge) has been
used. This makes sense because all "mergy" operations makes sure that
any path involved in the merge does not have local modifications before
starting, so resetting such a path away won't lose any information.
The previous commit changed the behavior of --merge to accept resetting
unmerged entries if they are reset to a different state than HEAD, but it
did not reset the changes in the work tree, leaving the conflict markers
in the resulting file in the work tree.
Fix it by doing three things:
- Update the documentation to match the wish of original "reset --merge"
better, namely, "An unmerged entry is a sign that the path didn't have
any local modification and can be safely resetted to whatever the new
HEAD records";
- Update read_index_unmerged(), which reads the index file into the cache
while dropping any higher-stage entries down to stage #0, not to copy
the object name from the higher stage entry. The code used to take the
object name from the a stage entry ("base" if you happened to have
stage #1, or "ours" if both sides added, etc.), which essentially meant
that you are getting random results depending on what the merge did.
The _only_ reason we want to keep a previously unmerged entry in the
index at stage #0 is so that we don't forget the fact that we have
corresponding file in the work tree in order to be able to remove it
when the tree we are resetting to does not have the path. In order to
differentiate such an entry from ordinary cache entry, the cache entry
added by read_index_unmerged() is marked as CE_CONFLICTED.
- Update merged_entry() and deleted_entry() so that they pay attention to
cache entries marked as CE_CONFLICTED. They are previously unmerged
entries, and the files in the work tree that correspond to them are
resetted away by oneway_merge() to the version from the tree we are
resetting to.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch makes "reset_index_file()" call "unpack_trees()" directly
instead of forking and execing "git read-tree". So the code is more
efficient.
And it's also easier to see which unpack_tree() options will be used,
as we don't need to follow "git read-tree"'s command line parsing
which is quite complex.
As Daniel Barkalow found, there is a difference between this new
version and the old one. The old version gives an error for
"git reset --merge" with unmerged entries, and the new version does
not when we reset the entries to some states that differ from HEAD.
Instead, it resets the index entry and succeeds, while leaving the
conflict markers in the corresponding file in the work tree (which
will be corrected by the next patch).
The code comes from the sequencer GSoC project:
git://repo.or.cz/git/sbeyer.git
(at commit 5a78908b70ceb5a4ea9fd4b82f07ceba1f019079)
Mentored-by: Daniel Barkalow <barkalow@iabervon.org>
Mentored-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Users might prefer to have git-difftool use a different
tool when run from a Git GUI.
This teaches git-difftool to honor 'diff.guitool' when
the '--gui' option is specified. This allows users to
configure their preferred command-line diff tool in
'diff.tool' and a GUI diff tool in 'diff.guitool'.
Reference: http://article.gmane.org/gmane.comp.version-control.git/133386
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a difftool test has an error then running the git test suite
may end up invoking a GUI diff tool. We now guard against this
by setting a difftool.bogus-tool.cmd variable.
The tests already used --tool=bogus-tool in various places so
this is simply ensuring that nothing ever falls back and
finds a real diff tool.
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many callsites run "sh -c $CMD" to run $CMD. We can make it
a little simpler for them by factoring out the munging of
argv.
For simple cases with no arguments, this doesn't help much, but:
1. For cases with arguments, we save the caller from
having to build the appropriate shell snippet.
2. We can later optimize to avoid the shell when
there are no metacharacters in the program.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
branch: die explicitly why when calling "git branch [-a|-r] branchname".
fast-import: Document author/committer/tagger name is optional
SubmittingPatches: hints to know the status of a submitted patch.
The -a and -r options used to be silently ignored in such a command.
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The fast-import parser does not validate that the author, committer
or tagger name component contains both a name and an email address.
Therefore the name component has always been optional. Correct the
documentation to match the implementation.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"What happened to my patch" is pretty much a FAQ on the Git mailing list,
it deserves a few paragraphs in SubmittingPatches...
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the implementation of GIT_PS1_SHOWDIRTYSTATE in 738a94a (bash:
offer to show (un)staged changes, 2009-02-03), I cut&pasted the
git-diff invocations from dirty-worktree checks elsewhere, carrying
along the --ignore-submodules option.
As pointed out by Kevin Ballard, this doesn't really make sense: to
the _user_, a changed submodule counts towards uncommitted changes.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
