filter-branch: work correctly with ambiguous refnames
'git-filter-branch branch' could fail producing the error:
"Which ref do you want to rewrite?" if existed another branch
or tag, which name was 'branch-something' or 'something/branch'.
[jc: original report and fix were done between Dmitry Potapov
and Dscho; I rewrote it using "rev-parse --symbolic-full-name"]
The plumbing level can understand that the user meant
"refs/heads/master" when the user says "master" or
"heads/master", but there is no easy way for the scripts to
figure it out without duplicating the dwim_ref() logic.
git-stash clear: refuse to work with extra parameter for now
Because it is so tempting to expect "git stash clear stash@{4}"
to remove the fourth element in the stash while leaving other
elements intact, we should not blindly throw away everything
upon seeing such a command.
This may change when we start using "git reflog delete" to
selectively nuke a single (or multiple, for that matter) stash
entries with such a command line.
git-stash: use stdout instead of stderr for non error messages
Some scripts and libraries check stderr to detect a failing command,
instead of checking the exit code. Because the output from git-status
is not primarily for machine consumption, it would not hurt to send
these messages to stdout instead and it will make it easier to drive
the command for such callers.
Signed-off-by: Marco Costalba <mcostalba@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t/t3800: do not use a temporary file to hold expected result.
It is a good practice to write program output to a temporary file
during the test, as it would allow easier postmortem when the tested
program does break. But there is no benefit in writing the expected
output out to the temporary.
This actually fixes a bug in check_verify_failure() routine.
The intention of the test seems to make sure the "git mktag" command
fails, and it spits out the expected error message. But if the
command did not fail as expected, the shell function as originally
written would not have detected the failure.
* git://repo.or.cz/git-gui:
git-gui: Make commit log messages end with a newline
Added Swedish translation.
git-gui: Unconditionally use absolute paths with Cygwin
git-gui: Handle file mode changes (644->755) in diff viewer
git-gui: Move frequently used commands to the top of the context menu.
1. The program has already loaded git_diff_ui_config, in
which case this is a noop.
2. The program didn't, which means it is plumbing that
does not _want_ git_diff_ui_config to be loaded.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The funcname patterns influence the "comment" on @@ lines of
the diff. They are safe to use with plumbing since they
don't fundamentally change the meaning of the diff in any
way.
Since all diff users call either diff_ui_config or
diff_basic_config, we can get rid of the lazy reading of the
config.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The diff porcelain uses git_diff_ui_config to set
porcelain-ish config options, like automatically turning on
color. The plumbing specifically avoids calling this
function, since it doesn't want things like automatic color
or rename detection.
However, some diff options should be set for both plumbing
and porcelain. For example, one can still turn on color in
git-diff-files using the --color command line option. This
means we want the color config from color.diff.* (so that
once color is on, we use the user's preferred scheme), but
_not_ the color.diff variable.
We split the diff config into "ui" and "basic", where
"basic" is suitable for use by plumbing (so _most_ things
affecting the output should still go into the "ui" part).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now the routine is an open-coded loop that avoids an extra
strlen() in the previous implementation, it got a bit too big to
be inlined. Uninlining it makes code footprint smaller but the
result still retains the avoidance of strlen() cost.
This updates send-pack and fast-import to use symbolic constants
for checking the return values from check_ref_format(), and also
futureproof the logic in lock_any_ref_for_update() to explicitly
name the case that is usually considered an error but is Ok for
this particular use.
git-svn: unlink index files that were globbed, too
commit 3157dd9e89a71e80673d0bc21b5c0630f3b1fe68 (git-svn: unlink
internal index files after operations) introduced unlinking
index files after fetching. However, this missed indices for
refs that were created by globbing branches and tags. This will
track all refs we ever touch during a fetch and unlink them at
exit time.
Signed-off-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Certain codepaths (notably "git log --pretty=format...") use
prefixcmp() extensively, with very short prefixes. In those cases,
calling strlen() is a wasteful operation, so avoid it.
Initial patch by Marco Costalba.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
lock_any_ref_for_update(): reject wildcard return from check_ref_format
Recent check_ref_format() returns -3 as well as -1 (general
error) and -2 (less than two levels). The caller was explicitly
checking for -1, to allow "HEAD" but still needed to disallow
bogus refs.
This introduces symbolic constants for the return values from
check_ref_format() to make them read better and more
meaningful. Normal ref creation codepath can still treat
non-zero return values as errors.
config: handle lack of newline at end of file better
The config parsing routines use the static global
'config_file' to store the FILE* pointing to the current
config file being parsed. The function get_next_char()
automatically converts an EOF on this file to a newline for
the convenience of its callers, and it sets config_file to
NULL to indicate that EOF was reached.
This throws away useful information, though, since some
routines want to call ftell on 'config_file' to find out
exactly _where_ the routine ended. In the case of a key
ending at EOF boundary, we ended up segfaulting in some
cases (changing that key or adding another key in its
section), or failing to provide the necessary newline
(adding a new section).
This patch adds a new flag to indicate EOF and uses that
instead of setting config_file to NULL. It also makes sure
to add newlines where necessary for truncated input. All
three included tests fail without the patch.
After replaying a single change, the code performed a number of checks,
but some of them were for sanity checking, failures from which should
make the command abort, and others were checks to see if it should make
a new commit object. Stringing them together with "&&" was wrong.
git-rebase -i behaves better on commits with incomplete messages
The commit message template when squashing multiple commits is
prepared by concatenating the messages of existing commits
together. If the messages from some of them end with incomplete
lines, this would result in a suboptimal message template. Make
sure that we add a terminating LF after each commit message.
git-gui: Make commit log messages end with a newline
Concatenating commit log messages from multiple commits works better
when all of the commits end with a clean line break.
Its good to be strict in what you create, and lenient in what you
accept, and since we're creating here, we should always try to
Do The Right Thing(tm).
Signed-off-by: Bernt Hansen <bernt@alumni.uwaterloo.ca> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
"git pull --tags": error out with a better message.
When "git pull --tags" is run without any other arguments, the
standard error message "You told me to fetch and merge stuff but
there is nothing to merge! You might want to fix your config"
is given.
While the error may be technically correct, fixing the config
would not help, as "git pull --tags" itself tells "git fetch"
not to use the configured refspecs.
This commit makes "git pull --tags" to issue a different error
message to avoid confusion. This is merely an interim solution.
In the longer term, it would be a better approach to change the
semantics of --tags option to make "git fetch" and "git pull"
to:
(1) behave as if no --tags was given (so an explicit refspec on
the command line overrides configured ones, or no explicit
refspecs on the command line takes configured ones); but
(2) no auto-following of tags is made even when using
configured refspecs; and
(3) fetch all tags as not-for-merge entries".
Then we would not need to have this separate error message, as
the ordinary merge will happen even with the --tags option.
When a commit message that does not have a terminating LF is
read in and the memory that was allocated to read it happens to
have a LF immediately after that, the code was not careful and
went past the terminating NUL.
* ar/commit-cleanup:
Allow selection of different cleanup modes for commit messages
builtin-commit: avoid double-negation in the code.
builtin-commit: fix amending of the initial commit
t7005: do not exit inside test.
This moves the logic to quote two paths (prefix + path) in
C-style introduced in the previous commit from the
dump_quoted_path() in combine-diff.c to quote.c, and uses it to
fix rewrite_diff() that never C-quoted the pathnames correctly.
Earlier when showing combined diff, the filenames on the ---/+++
header lines were quoted incorrectly. a/ (or b/) prefix was
output literally and then the path was output, with c-quoting.
This fixes the quoting logic, and while at it, adjusts the code
to use the customizable prefix (a_prefix and b_prefix)
introduced recently.
In commit b7bb760d5ed4881422673d32f869d140221d3564 (Fix revision
log diff setup, avoid unnecessary diff generation) an optimization was
made to avoid unnecessary diff generation. This was partly fixed in 99516e35d096f41e7133cacde8fbed8ee9a3ecd0 (Fix embarrassing "git log
--follow" bug). The '--diff-filter' option also needs the diff machinery
in action.
Signed-off-by: Arjen Laarhoven <arjen@yaph.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Improve error messages when int/long cannot be parsed from config
If a config file has become mildly corrupted due to a missing LF
we may discover some other option joined up against the end of a
numeric value. For example:
[section]
number = 1auto
where the "auto" flag was meant to occur on the next line, below
"number", but the missing LF has caused it to no longer be its
own option. Instead the word "auto" is parsed as a 'unit factor'
for the value of "number".
Before this change we got the confusing error message:
fatal: unknown unit: 'auto'
which told us nothing about where the problem appeared. Now we get:
fatal: bad config value for 'aninvalid.unit'
which at least points the user in the right direction of where to
search for the incorrectly formatted configuration file.
Noticed by erikh on #git, which received the original error from
a simple `git checkout -b` due to a midly corrupted config.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We were not previously checking the exit status of cvsps at
all. If it exited before producing any useful output, we
ended up with an empty import, which caused a spew of
confusing error messages from other parts of git:
$ git-cvsimport foo
Initialized empty Git repository in ...
some error from cvsps
fatal: refs/heads/origin: not a valid SHA1
fatal: master: not a valid SHA1
warning: You appear to be on a branch yet to be born.
warning: Forcing checkout of HEAD.
fatal: just how do you expect me to merge 0 trees?
checkout failed: 256
Now we get:
$ git-cvsimport foo
Initialized empty Git repository in ...
some error from cvsps
git-cvsimport: fatal: cvsps reported error
Signed-off-by: Jeff King <peff@peff.net> Acked-by: Martin Langhoff <martin@catalyst.net.nz> Signed-off-by: Junio C Hamano <gitster@pobox.com>
shortlog manpage documentation: work around asciidoc markup issues
We wanted to have a list in which one (and the sole, as it happen to
be) item in it is ".mailmap", but do not seem to be able to convince
AsciiDoc to format it correctly for manpages. Reformat it into a
paragraph that describes the said file to work around the issue.
Signed-off-by: Gustaf Hendeby <hendeby@isy.liu.se> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allow selection of different cleanup modes for commit messages
Although we traditionally stripped away excess blank lines, trailing
whitespaces and lines that begin with "#" from the commit log message,
sometimes the message just has to be the way user wants it.
For instance, a commit message template can contain lines that begin with
"#", the message must be kept as close to its original source as possible
if you are converting from a foreign SCM, or maybe the message has a shell
script including its comments for future reference.
The cleanup modes are default, verbatim, whitespace and strip. The
default mode depends on if the message is being edited and will either
strip whitespace and comments (if editor active) or just strip the
whitespace (for where the message is given explicitely).
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin-commit: avoid double-negation in the code.
The flag no_edit meant "we have got final message from the user
and will not editing it any further", but there were quite a few
places that needed to check !no_edit. Rename the variable to
use_editor and reverse the logic everywhere.
builtin-commit: fix amending of the initial commit
When amending initial commit without editor, the command
incorrectly barfed because the check to see if there is anything
to commit referenced the non-existent HEAD^1.
Emit helpful status for accidental "git stash" save
If the user types "git stash" mistakenly thinking that this will list
their stashes he/she may be surprised to see that it actually saved
a new stash and reset their working tree and index.
In the worst case they might not know how to recover the state. So
help them by telling them exactly what was saved and also how to
restore it immediately.
Signed-off-by: Wincent Colaiuta <win@wincent.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix $EDITOR regression introduced by rewrite in C.
When git-tag and git-commit launches the editor, they used to
honor EDITOR="editor -options args..." but recent rewrite in C
insisted on $EDITOR to be the path to the editor executable.
Currently git send-email does not accept $EDITOR with arguments, eg,
emacs -nw, when starting an editor to produce a cover letter. This
patch changes this by letting the shell handle the option parsing.
Signed-off-by: Gustaf Hendeby <hendeby@isy.liu.se> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The command itself takes an optional <pattern> argument that
limits the shown tags to the ones that match when in listing
mode that is triggered with '-l' option. The <pattern> is not
an optional option-argument to '-l'.
With this fix, "git tag -l -n 4 v0.99" works as expected.
It also removes a few bogus tests in t7004.
Signed-off-by: Pierre Habouzit <madcoder@debian.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The tar-ball and the git archive itself is fine, but yes, the diff from
2.6.23 to 2.6.24-rc6 is bad. It's the "trim_common_tail()" optimization
that has caused way too much pain.
Very interesting breakage. The patch was actually "correct" in a (rather
limited) technical sense, but the context at the end was missing because
while the trim_common_tail() code made sure to keep enough common context
to allow a valid diff to be generated, the diff machinery itself could
decide that it could generate the diff differently than the "obvious"
solution.
Thee sad fact is that the git optimization (which is very important for
"git blame", which needs no context), is only really valid for that one
case where we really don't need any context.
[jc: since this is shared with "git diff -U0" codepath, context recovery
to the end of line needs to be done even for zero context case.]
In everyday tasks, "repack -a -d -f" won't be used, so there
is not much point mentioning "repack". By showing the --prune
option to "gc", we can do without mentioning "git prune", too.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some entries in .gitignore are obselete. These should be cleaned up
just for the sake of general tidiness and so that any developers who
have a working tree that was moved forward without a clean know that
they have old stuff in their work tree.
Signed-off-by: Charles Bailey <charles@hashpling.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
shell-scripts usage(): consistently exit with non-zero
Earlier conversion of shell scripts to parse-options made usage()
to run "git cmd -h" which in turn emit LONG_USAGE and exit with 0
status. This is inconsistent with the scripts that do not use
parse-options, whose usage() died with the message, exiting with 1.
Reallow git-rebase --interactive --continue if commit is unnecessary
During git-rebase --interactive's --continue implementation we used
to silently restart the rebase if the user had made the commit
for us. This is common if the user stops to edit a commit and
does so by amending it. My recent change to watch git-commit's
exit status broke this behavior.
Thanks to Bernt Hansen for catching it in 1.5.4-rc1.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The prepare_log_message() function serves two purposes:
- Prepares the commit log message template, to be given to the end
user;
- Return true if there is something committable;
7168624c3530d8c7ee32f930f8fb2ba302b9801f (Do not generate full commit
log message if it is not going to be used) cheated to omit the former
when we know the log message template is not going to be used. However,
its replacement logic to see if there is something committable was
botched. When amending, it should compare the index with the parent of
the HEAD, not the current HEAD. Otherwise you cannot run --amend to
fix only the message without changing the tree.
The earlier test stripped away expected number of 'z' but the output
would have been very hard to read once somebody broke the common tail
optimization. Instead, count the number of 'z' and show it, to help
diagnosing the problem better in the future.
Teach diff machinery to display other prefixes than "a/" and "b/"
With the new options "--src-prefix=<prefix>", "--dst-prefix=<prefix>"
and "--no-prefix", you can now control the path prefixes of the diff
machinery. These used to by hardwired to "a/" for the source prefix
and "b/" for the destination prefix.
Initial patch by Pascal Obry. Sane option names suggested by Linus.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Catch and handle git-commit failures in git-rebase --interactive
If git-commit fails for any reason then git-rebase needs to stop
and not plow through the rest of the series. Its unlikely that
a future git-commit will succeed if the current attempt failed.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we are rebasing changes that contain potential whitespace
errors that our .git/hooks/pre-commit hook looks for and fails
on then git-commit will fail to commit that change. This causes
git-rebase--interactive to squash commits together, even though it
was not requested to do so by the todo file.
Passing --no-verify to git-commit makes git-rebase -i behave more
like git-rebase normally would in such conditions, providing more
consistent behavior between the different rebase implementations.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When merging conflicting submodule changes from a supermodule, generate
a conflict message saying what went wrong. Also leave the tree in a state
where git status shows the conflict, and git submodule status gives the user
enough information to do the merge manally. Previously this would just fail.
Signed-off-by: Finn Arne Gangstad <finnag@pvv.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix git-instaweb breakage on MacOS X due to the limited sed functionality
git-instaweb relied on a pipe in a sed script, but this is not supported
by MacOS X sed when using BREs. git-instaweb relies on a working perl
in any case, and perl re are more consistent between platforms, so
replace sed invocation with an equivalent perl invocation.
Also, fix the documented -b "" to work without giving a spurious 'command
not found' error.
Signed-off-by: Charles Bailey <charles@hashpling.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's possible for bad clients to commit symlinks without the
5-character "link " prefix in symlinks. So guard around this
bug in SVN and make a best effort to create symlinks if the
"link " prefix is missing.
More information on this SVN bug is described here:
http://subversion.tigris.org/issues/show_bug.cgi?id=2692
To be pedantic, there is still a corner case that neither we nor
SVN can handle: If somebody made a link using a broken SVN
client where "link " is the first part of its path, e.g.
"link sausage", then we'd end up having a symlink which points
to "sausage" because we incorrectly stripped the "link ".
Hopefully this hasn't happened in practice, but if it has,
it's not our fault SVN is broken :)
Thanks to Benoit Sigoure and Sverre Johansen for reporting
and feedback.
git-svn: avoid leaving leftover committer/author info in rebase
We set the 6 environment variables for controlling
committer/author email/name/time for every commit.
We do this in the parent process to be passed to
git-commit-tree, because open3() doesn't afford us the control
of doing it only in the child process. This means we leave them
hanging around in the main process until the next revision comes
around and all 6 environment variables are overwridden again.
Unfortunately, for the last commit, leaving them hanging around
means the git-rebase invocation will pick it up, rewriting the
rebased commit with incorrect author information. This should fix
it.
git-filter-branch.sh: more portable tr usage: use \012, not \n.
I hesitate to suggest this, since GNU tr has accepted \n for 15 years,
but there are supposedly a few crufty vendor-supplied versions of tr still
in use. Also, all of the other uses of tr-with-newline in git use \012.
Signed-off-by: Jim Meyering <meyering@redhat.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
filter-branch: Remove broken and unnecessary summary of rewritten refs.
There was an attempt to list the refs that were rewritten by filtering
the output of 'git show-ref' for 'refs/original'. But it got the
grep argument wrong, which did not account for the SHA1 that is listed
before the ref.
Moreover, right before this summary is the loop that actually does the
rewriting, and the rewritten refs are listed there anyway. So this extra
summary is plainly too verbose.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Junio C Hamano <gitster@pobox.com>
and that "merge" is a cache_entry pointer. If we have a non-zero
FLEX_ARRAY size, it will cause us to copy the first few bytes of the
name too.
That is technically wrong even for FLEX_ARRAY being 1, but you'll never
notice, since the filenames should always be the same with the current
code. But if we do the same thing for a rename, we'd be screwed.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>