The ie_modified() function is the workhorse for refresh_cache_entry(),
i.e. checking if an index entry that is stat-dirty actually has changes.
After running quicker check to compare cached stat information with
results from the latest lstat(2) to answer "has modification" early, the
code goes on to check if there really is a change by comparing the staged
data with what is on the filesystem by asking ce_modified_check_fs().
However, this function always said "no change" for any gitlinks that has a
directory at the corresponding path. This made ie_modified() to miss
actual changes in the subproject.
The patch fixes this first by modifying an existing short-circuit logic
before calling the ce_modified_check_fs() function. It knows that for any
filesystem entity to which ie_match_stat() says its data has changed, if
its cached size is nonzero then the contents cannot match, which is a
correct optimization only for blob objects. We teach gitlink objects to
this special case, as we already know that any gitlink that
ie_match_stat() says is modified is indeed modified at this point in the
codepath.
With the above change, we could leave ce_modified_check_fs() broken, but
it also futureproofs the code by teaching it to use ce_compare_gitlink(),
instead of assuming (incorrectly) that any directory is unchanged.
Originally noticed by Alex Riesen on Cygwin.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This treats aborting a commit more like a feature.
Signed-off-by: Anders Melchiorsen <mail@cup.kalibalik.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Although it does not matter for Git itself, tools that
export to systems that explicitly track copies and
renames can benefit from such information.
This patch makes fast-export output correct action
logs when -M or -C are enabled.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch makes two small changes to improve the output of --inline
and --attach.
The first is to write a newline preceding the boundary. This is needed because
MIME defines the encapsulation boundary as including the preceding CRLF (or in
this case, just LF), so we should be writing one. Without this, the last
newline in the pre-diff content is consumed instead.
The second change is to always write the line termination character
(default: newline) even when using --inline or --attach. This is simply to
improve the aesthetics of the resulting message. When using --inline an email
client should render the resulting message identically to the non-inline
version. And when using --attach this adds a blank line preceding the
attachment in the email, which is visually attractive.
Signed-off-by: Kevin Ballard <kevin@sb.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-svnimport is no longer supported, so don't mention it in the
documentation. This also updates the description, removing the
historical discussion, since it mostly dealt with how it differed from
svnimport. The new description gives some starting points into the
rest of the documentation.
Noticed by Jurko Gospodnetić <jurko.gospodnetic@docte.hr>
Signed-off-by: Brian Gernhardt <benji@silverinsanity.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-gui needs bindir in PATH to be able to run 'git'. bindir
however is not necessarily in PATH if started directly through a
Windows shortcut. Therefore, we used to add the directory
git-gui is located in. But with the new 'libexec/git-core'
layout this directory is no longer identical to bindir.
This commit modifies the wrapper script to discover the bindir
and add it to PATH.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Instead of using an absolute path, git-gui can discover its
gui library using a relative path from execdir. We want to
use the relative path discovery on MinGW to avoid issues
with translation of absolute paths.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We always wanted the library for git-gui to install into the
$prefix/share directory, not $prefix/libexec/share. All of
the files in our library are platform independent and may
be reused across systems, like any other content stored in
the share directory.
Our computation of where our library should install to was broken
when git itself started installing to $prefix/libexec/git-core,
which was one level down from where we expected it to be.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Back in 15430be5a1 ("Look for gitk in $PATH, not $LIBEXEC/git-core")
git-gui learned to use [_which gitk] to locate where gitk's script
is as Git 1.6 will install gitk to $prefix/bin (in $PATH) and all
of the other tools are in $gitexecdir.
This failed on Windows because _which adds the ".exe" suffix as it
searches for the program on $PATH, under the assumption that we can
only execute something from Tcl if it is a proper Windows executable.
When scanning for gitk on Windows we need to omit the ".exe" suffix.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
A new configuration variable 'core.trustctime' is introduced to
allow ignoring st_ctime information when checking if paths
in the working tree has changed, because there are situations where
it produces too much false positives. Like when file system crawlers
keep changing it when scanning and using the ctime for marking scanned
files.
The default is to notice ctime changes.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Before this patch "git merge-base" just exited with error code 1
and without an error message in case it was passed a ref to an
object that is not a commit (for example a tree).
This patch makes it "die" in this case with an error message.
While at it, this patch also refactors the code to get the
commit reference from an argument into a new
"get_commit_reference" function.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The test assumed that we can keep the cached stat information fresh across
rename(2); many filesystems however update st_ctime (and POSIX allows them
to do so), and that assumption does not hold.
We can explicitly refresh the index for the purpose of these tests. The
only thing we are interested in is the staged contents and the mode bits
are preserved across "git mv".
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We prefer running the dashless form, and POSIX side already does so; we
should use it in MinGW's start_command(), too.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Acked-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
prepare_git_cmd(const char **argv) adds a first entry "git" to
the array argv. The new array is allocated on the heap. It's
the caller's responsibility to release it with free(). The code
was already present in execv_git_cmd() but could not be used from
outside. Now it can also be called for preparing the command list
in the MinGW codepath in run-command.c.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the C rewrite, "git verify-tag -v" just does nothing instead of
printing the usage message with an error. This patch fix the regression.
Signed-off-by: Olivier Marin <dkr@freesurf.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When run in a working copy subdirectory, git-ls-tree will automagically
add the prefix to the pathspec, which can result in an unexpected behavior
when the tree object accessed is not the root tree object.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In an earlier commit c70a8d9 (Makefile: Do not install a copy of 'git' in
$(gitexecdir), 2008-07-21), we tried to avoid installing two git, one in
/usr/bin/git and the other in /usr/libexec/git-core/git. It mistakenly
removed the only copy of git when gitexecdir and bindir are set to the
same directory, i.e. the traditional layout.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is a follow-up patch to 49fa65a (Allow the built-in exec path to be
relative to the command invocation path, 2008-07-23). Without specific
gitexecdir passed from the command line, git-gui's build procedure would
try to figure out the value for it by running an installed git.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a user passes "--template=", then our template parameter
is blank. Unfortunately, copy_templates() assumes it has at
least one character, and does all sorts of bad things like
reading from template[-1] and then proceeding to link all of
'/' into the .git directory.
This patch just checks for that condition in copy_templates
and aborts. As a side effect, this means that --template=
now has the meaning "don't copy any templates."
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The rewrite of git-mv from a shell script to a builtin was perhaps
a little too straightforward: the git add and git rm queues were
emulated directly, which resulted in a rather complicated code and
caused an inconsistent behaviour when moving dirty index entries;
git mv would update the entry based on working tree state,
except in case of overwrites, where the new entry would still have
sha1 of the old file.
This patch introduces rename_index_entry_at() into the index toolkit,
which will rename an entry while removing any entries the new entry
might render duplicate. This is then used in git mv instead
of all the file queues, resulting in a major simplification
of the code and an inevitable change in git mv -n output format.
Also the code used to refuse renaming overwriting symlink with a regular
file and vice versa; there is no need for that.
A few new tests have been added to the testsuite to reflect this change.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The path list builder had a branch for the case the source is not in index, but
this can happen only if the source was a directory. However, in that case we
have already expanded the list to the directory contents and set mode
to WORKING_DIRECTORY, which is tested earlier.
The patch removes the superfluous branch and adds an assert() instead. git-mv
testsuite still passes.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Not all temporary file creation routines will ensure 14 bytes are
used to generate the temporary file name. In C Git this may be
true, but alternate implementations such as jgit are not always
able to generate a temporary file name with a specific prefix and
also ensure the file name length is 14 bytes long.
Since temporary files in a directory we are fsck'ing should be
uncommon (as they are short lived only long enough for an active
writer to finish writing the file and rename it) we shouldn't see
these show up very often. Always using a prefixcmp() call and
ignoring the length opens up room for other implementations to use
different name generation schemes.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function list_commands_in_dir() tried to be lazy and just chdir()
to the directory which entries it listed, so that the check if the
file is executable could be done on dir->d_name.
However, there is no good reason to jump around wildly just to find
all Git commands.
Instead, have a strbuf and construct the full path dynamically.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After the optimization to --[no-]merged logic, the calculation of the
width of the longest refname to be shown might become inaccurate (since
the matching against merge_filter is performed after adding refs to
ref_list). This patch forces a recalculation of maxwidth when it might
be needed.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The logic for checking commits against merge_filter will be reused
when we recalculate the maxwidth of refnames.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous optimization to --[no-]merged ended up with some duplicated
code which this patch removes.
Signed-off-by: Lars Hjemli <hjemli@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Those shorthands are explained in the rev-parse documentation but were not
actually supported by rev-parse itself.
gitk internally uses rev-parse to interpret its command line arguments, and
being able to use these "limit with parents" syntax is handy there.
Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A wildcard refspec is internally parsed into a refspec structure with src
and dst strings. Many parts of the code assumed that these do not include
the trailing "/*" when matching the wildcard pattern with an actual ref we
see at the remote. What this meant was that we needed to make sure not
just that the prefix matched, and also that a slash followed the part that
matched.
But a codepath that scans the result from ls-remote and finds matching
refs forgot to check the "matching part must be followed by a slash" rule.
This resulted in "refs/heads/b1" from the remote side to mistakenly match
the source side of "refs/heads/b/*:refs/remotes/b/*" refspec.
Worse, the refspec crafted internally by "git-clone", and a hardcoded
preparsed refspec that is used to implement "git-fetch --tags", violated
this "parsed widcard refspec does not end with slash" rule; simply adding
the "matching part must be followed by a slash" rule then would have
broken codepaths that use these refspecs.
This commit changes the rule to require a trailing slash to parsed
wildcard refspecs. IOW, "refs/heads/b/*:refs/remotes/b/*" is parsed as
src = "refs/heads/b/" and dst = "refs/remotes/b/". This allows us to
simplify the matching logic because we only need to do a prefixcmp() to
notice "refs/heads/b/one" matches and "refs/heads/b1" does not.
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
t6030-bisect-porcelain.sh relies on "ls" exiting with nonzero
status when asked to list nonexistent files. Unfortunately,
/bin/ls on Mac OS X 10.3 exits with exit code 0. So look at
its output instead.
Signed-off-by: Jonathan Nieder <jrnieder@uchicago.edu>
Acked-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rename cached __git_commandlist to __git_porcelain_commandlist and
add __git_all_commandlist that only filters out *--* helpers.
Completions for 'git help' will use the __git_all_commandlist, while
the __git_porcelain_commandlist is used for git command completion.
Users who actually read man pages may want to see help for plumbing
commands.
Signed-off-by: Lee Marlow <lee.marlow@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The diff.external examples pass a flag to gnu-diff, but GNU diff
does not follow the GIT_EXTERNAL_DIFF interface.
Signed-off-by: Anders Melchiorsen <mail@cup.kalibalik.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://repo.or.cz/git-gui:
git-gui: "Stage Line": Treat independent changes in adjacent lines better
git-gui: Fix "Stage/Unstage Line" with one line of context.
git-gui: Correct 'Visualize Branches' on Mac OS X to start gitk
git-gui: Look for gitk in $PATH, not $LIBEXEC/git-core
Add a menu item to invoke full copy detection in blame.
Kill the blame back-end on window close.
Add options to control the search for copies in blame.
Fix pre-commit hooks under MinGW/MSYS
It is especially useful for Stage/Unstage Line, because
they invoke full state scan and diff reload, which originally
would reset the scroll position to the top of the file.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
It was positioned incorrectly (offset by one position)
if the menu had a tear-off handle.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Assume that we want to commit these states:
Old state == HEAD Intermediate state New state
--------------------------------------------------------
context before context before context before
old 1 new 1 new 1
old 2 old 2 new 2
context after context after context after
that is, want to commit two changes in this order:
1. transform "old 1" into "new 1"
2. transform "old 2" into "new 2"
[This discussion and this patch is about this very case and one other case
as outlined below; any other intermediate states that one could imagine are
not affected by this patch.]
Now assume further, that we have not staged and commited anything, but we
have already changed the working file to the new state. Then we will see
this hunk in the "Unstaged Changes":
@@ -1,4 +1,4 @@
context before
-old 1
-old 2
+new 1
+new 2
context after
The obvious way to stage the intermediate state is to apply "Stage This
Line" to "-old 1" and "+new 1". Unfortunately, this resulted in this
intermediate state:
context before
old 2
new 1
context after
which is not what we wanted. In fact, it was impossible to stage the
intermediate state using "Stage Line". The crux was that if a "+" line was
staged, then the "-" lines were converted to context lines and arranged
*before* the "+" line in the forged hunk that we fed to 'git apply'.
With this patch we now treat "+" lines that are staged differently. In
particular, the "-" lines before the "+" block are moved *after* the
staged "+" line. Now it is possible to get the correct intermediate state
by staging "-old 1" and "+new 1". Problem solved.
But there is a catch.
Noticing that we didn't get the right intermediate state by staging
"-old 1" and "+new 1", we could have had the idea to stage the complete
hunk and to *unstage* "-old 2" and "+new 2". But... the result is the same.
The reason is that there is the exact symmetric problem with unstaging the
last "-" and "+" line that are in adjacent blocks of "-" and "+" lines.
This patch does *not* change the way in which "-" lines are *unstaged*.
Why? Because if we did (i.e. move "+" lines before the "-" line after
converting them to context lines), then it would be impossible to stage
this intermediate state:
context before
old 1
new 2
context after
that is, it would be impossible to stage the two independet changes in the
opposite order.
Let's look at this case a bit further: The obvious way to get this
intermediate state would be to apply "Stage This Line" to "-old 2" and
"+new 2". Before this patch, this worked as expected. With this patch, it
does not work as expected, but it can still be achieved by first staging
the entire hunk, then *unstaging* "-old 1" and "+new 1".
In summary, this patch makes a common case possible, at the expense that
a less common case is made more complicated for the user.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
To "Stage/Unstage Line" we construct a patch that contains exactly one
change (either addition or removal); the hunk header was forged by counting
the old side and adjusting the count by +/-1 for the new side. But when we
counted the context we never counted the changed line itself. If the hunk
had only one removal line and one line of context, like this:
@@ -1,3 +1,2 @@
context 1
-removal
context 2
We had constructed this patch:
@@ -1,2 +1,1 @@
context 1
-removal
context 2
which does not apply because git apply deduces that it must apply at the
end of the file. ("context 2" is considered garbage and ignored.) The fix
is that removal lines must be counted towards the context of the old side.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
"git reset -q" is advertised to "only report errors", but "locally
modified" messages are still shown. They are not errors but diagnostics.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Subversion repositories often require files to have properties such as
svn:mime-type and svn:eol-style set when they are added. Users
typically set these properties automatically using the SVN auto-props
feature with 'svn add'. This commit teaches dcommit to look at the user
SVN configuration and apply matching auto-props entries for files added
by a diff as it is applied to the SVN remote.
Signed-off-by: Brad King <brad.king@kitware.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since the exec-path on Windows is derived from the program invocation path,
we must ensure that argv[0] always has a path. Unfortunately, if a program
is invoked from CMD, argv[0] has no path. But on the other hand, the
C runtime offers a global variable, _pgmptr, that always has the full path
to the program. We hook into main() with a preprocessor macro, where we
replace argv[0].
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function had used make_absolute_path(); but this function dies if
the directory that contains the entry whose relative path was supplied in
the argument does not exist. This is a problem if the argument is, for
example, "../libexec/git-core", and that "../libexec" does not exist.
Since the resolution of symbolic links is not required for elements in
PATH, we can fall back to using make_nonrelative_path(), which simply
prepends $PWD to the path.
We have to move make_nonrelative_path() alongside make_absolute_path() in
abspath.c so that git-shell can be linked. See 5b8e6f85f.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
