The "trivial merge" codepath wants to optimize itself by making an
internal call to the read-tree machinery, but it does not read the index
before doing so, and the codepath is never exercised. Incidentally, this
failure to read the index upfront means that the safety to refuse doing
anything when the index is unmerged does not kick in, either.
These two problem are fixed by using read_cache_unmerged() that does read
the index before checking if it is unmerged at the beginning of
cmd_merge().
The primary logic of the merge, however, assumes that the process never
reads the index in-core, and the call to write_cache_as_tree() it makes
from write_tree_trivial() will always read from the on-disk index that is
prepared the strategy back-ends. This assumption is now broken by the
above fix. To fix this issue, we now call discard_cache() before calling
write_tree_trivial() when it wants to write the on-disk index as a tree.
When multiple strategies are tried, their results are evaluated by reading
the resulting index and inspecting it. The codepath needs to make a call
to read_cache() for each successful strategy, and for that to work, they
need to discard_cache() the one read by the previous round.
Also the "trivial merge" forgot that the current commit is one of the
parents of the resulting commit.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We used to refresh the index to clear stat-dirtyness before a fast-forward
merge. Recent C rewrite forgot to do this.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If you have a tag with a single, incomplete line as its payload, asking
git-for-each-ref for its %(body) element accessed a NULL pointer.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code remembered that the last diff output it saw was an empty line,
and tried to reset that state whenever it sees a context line, a non-blank
new line, or a new hunk. However, this codepath asks the underlying diff
engine to feed diff without any context, and the "just saw an empty line"
state was not reset if you added a new blank line in the last hunk of your
patch, even if it is not the last line of the file.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In handle_from, we calculate the end boundary of a section
to remove from a strbuf using strcspn like this:
el = strcspn(buf, set_of_end_boundaries);
strbuf_remove(&sb, start, el + 1);
This works fine if "el" is the offset of the boundary
character, meaning we remove up to and including that
character. But if the end boundary didn't match (that is, we
hit the end of the string as the boundary instead) then we
want just "el". Asking for "el+1" caught an out-of-bounds
assertion in the strbuf library.
This manifested itself when we got a 'From' header that had
just an email address with nothing else in it (the end of
the string was the end of the address, rather than, e.g., a
trailing '>' character), causing git-mailinfo to barf.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some old platforms have an old diff which doesn't have the -U option.
'git diff' can be used in its place. Adjust the comparison function to
strip git's additional header lines to make this possible.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Recent changes to is_multipart_boundary() caused git-mailinfo to segfault.
The reason was after handling the end of the boundary the code tried to look
for another boundary. Because the boundary list was empty, dereferencing
the pointer to the top of the boundary caused the program to go boom.
The fix is to check to see if the list is empty and if so go on its merry
way instead of looking for another boundary.
I also fixed a couple of increments and decrements that didn't look correct
relating to content_top.
The boundary test case was updated to catch future problems like this again.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The git-apply documentation says that --binary is a historical option.
This patch lets git-am ignore --binary and removes advertisements of this
option.
Signed-off-by: Stephan Beyer <s-beyer@gmx.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use square brackets instead.
And the prominent example of the deficiency are, as usual, the filesystems
of Microsoft house.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
test-chmtime can adjust the mtime of a file based on the file's mtime, or
based on the system time. For files accessed over NFS, the file's mtime is
set by the NFS server, and as such may vary a great deal from the NFS
client's system time if the clocks of the client and server are out of
sync. Since these tests are testing the expire feature of git-prune, an
incorrect mtime could cause a file to be expired or not expired incorrectly
and produce a test failure.
Avoid this NFS pitfall by modifying the calls to test-chmtime so that the
mtime is adjusted based on the system time, rather than the file's mtime.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
match_one implements an optimized pathspec match where it only uses
fnmatch if it detects glob special characters in the pattern. Unfortunately
it didn't treat \ as a special character, so attempts to escape a glob
special character would fail even though fnmatch() supports it.
Signed-off-by: Kevin Ballard <kevin@sb.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The existing parent rewriting did not handle the case where a previous
commit was amended (via edit or squash). Fix by always putting the
new sha1 of the last commit into the $REWRITTEN map.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
'git rebase -i -p' forgot to update the index and working directory
during fast forwards. Fix this. Makes 'GIT_EDITOR=true rebase -i -p
<ancestor>' a no-op again.
Also, it attempted to do a fast forward even if it was instructed not
to commit (via -n). Fall back to the cherry-pick code path and let
that handle the issue for us.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Recently "git diff --check" learned to detect new trailing blank lines
just like "git apply --whitespace" does. However this check should not
trigger unconditionally. This patch makes it honor the whitespace
settings from core.whitespace and gitattributes.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The post-update hook, which is required to be enabled in order for
the repository to be accessible over HTTP, is not enabled by
chmod a+x anymore, but instead by dropping the .sample suffix.
This patch emphasizes this change in the release notes (since
I believe this is rather noticeable backwards-incompatible change).
It also adjusts the documentation which still described the old way
and fixes t/t5540-http-push.sh, which was broken for 1.5 month
but apparently noone ever runs this test.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
Documentation: fix invalid reference to 'mybranch' in user manual
Fix deleting reflog entries from HEAD reflog
reflog test: add more tests for 'reflog delete'
Documentation: rev-list-options: Fix -g paragraph formatting
Conflicts:
Documentation/user-manual.txt
dwim_ref() used to resolve HEAD symbolic ref to its target (i.e. current
branch). This incorrectly removed the reflog entry from the current
branch when 'git reflog delete HEAD@{1}' was asked for.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This adds more tests for 'reflog delete' and marks it as
broken, as currently a call to 'git reflog delete HEAD@{1}'
deletes entries in the currently checked out branch's log,
not the HEAD log.
Noticed by John Wiegley
Signed-off-by: Pieter de Bie <pdebie@ai.rug.nl>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After finding a MIME multi-part message boundary line, the handle_body()
function is supposed to first flush any accumulated contents from the
previous part to the output stream. However, the code mistakenly output
the boundary line it found.
The old code that used one global, fixed-length buffer line[] used an
alternate static buffer newline[] for keeping track of this accumulated
contents and flushed newline[] upon seeing the boundary; when 3b6121f
(git-mailinfo: use strbuf's instead of fixed buffers, 2008-07-13)
converted a fixed-length buffer in this program to use strbuf,these two
buffers were converted to "line" and "prev" (the latter of which now has a
much more sensible name) strbufs, but the code mistakenly flushed "line"
(which contains the boundary we have just found), instead of "prev".
This resulted in the first boundary to be output in front of the first
line of the message.
The rewritten implementation of handle_boundary() lost the terminating
newline; this would then result in the second line of the message to be
stuck with the first line.
The is_multipart_boundary() was designed to catch both the internal
boundary and the terminating one (the one with trailing "--"); this also
was broken with the rewrite, and the code in the handle_boundary() to
handle the terminating boundary was never triggered.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some repositories use a deep branching strategy, such as:
branches/1.0/1.0.rc1
branches/1.0/1.0.rc2
branches/1.0/1.0.rtm
branches/1.0/1.0.gold
Only allowing a single glob stiffles this.
This change allows for a single glob 'set' to accept this deep
branching strategy.
The ref glob depth must match the branch glob depth. When using
the -b or -t options for init or clone, this is automatically
done.
For example, using the above branches:
svn-remote.svn.branches = branches/*/*:refs/remote/*/*
gives the following branch names:
1.0/1.0.rc1
1.0/1.0.rc2
1.0/1.0.rtm
1.0/1.0.gold
[ew:
* removed unrelated line-wrapping changes
* fixed line-wrapping in a few more places
* removed trailing whitespace
* fixed bashism in test
* removed unnecessary httpd startup in test
* changed copyright on tests to 2008 Marcus Griep
* added executable permissions to new tests
]
Signed-off-by: Marcus Griep <marcus@griep.us>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fixes bad regex match check for multiple globs (would always return
one glob regardless of actual number).
[ew: fixed a bashism in the test and some minor line-wrapping]
Signed-off-by: Marcus Griep <marcus@griep.us>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With --mirror, clone asks for refs/* already, so it does not need to
ask for ref/tags/*, too.
Noticed by Cesar Eduardo Barros.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When running "git commit -F file" and "git tag -F file" from a
subdirectory, we should take it as relative to the directory we started
from, not relative to the top-level directory.
This adds a helper function "parse_options_fix_filename()" to make it more
convenient to fix this class of issues. Ideally, parse_options() should
support a new type of option, "OPT_FILENAME", to do this uniformly, but
this patch is meant to go to 'maint' to fix it minimally.
One thing to note is that value for "commit template file" that comes from
the command line is taken as relative to $cwd just like other parameters,
but when it comes from the configuration varilable 'commit.template', it
is taken as relative to the working tree root as before. I think this
difference actually is sensible (not that I particularly think
commit.template itself is sensible).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I've tested this with svn 1.4.4
This also adds quoting to make it work odd characters
in the trash path.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Changed the ptouch bash function to use the "Text Last Updated"
date reported by 'svn info' when changing the modified time
(mtime) of the file/symlink/directory in the git working
directory. Previously it used the mtime of the item in the
svn working directory, which caused the race condition.
[ew: swapped argument order of ptouch() to minimize diff]
From: David D. Kilzer <ddkilzer@kilzer.net>
Signed-off-by: David D. Kilzer <ddkilzer@kilzer.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The tests requires anonymous write access. Therefore, "anon-access =
write" is added to conf/svnserve.conf. But because it was added to
the end of the file, it is impossible to guarantee in what section
it will be located. It turned out that on SVN 1.5, it was placed in
the wrong section and as result the test failed.
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Tested-by: Brad King <brad.king@kitware.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This prevents double output in case stdout is redirected.
Signed-off-by: Anders Melchiorsen <mail@cup.kalibalik.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We have a tradition that bare repositories live in directories ending
in ".git". To make this more a convention than just a tradition, teach
"git clone --bare" to add a ".git" suffix to the directory name.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The command line
$ git clone --mirror $URL
is now a short-hand for
$ git clone --bare $URL
$ (cd $(basename $URL) && git remote add --mirror origin $URL)
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some shells hang when parsing the script if the last statement is not
followed by a newline. So add one.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We used to set the TOPOSORT flag of commits during the topological
sorting, but we can just as well use the member "indegree" for it:
indegree is now incremented by 1 in the cases where the commit used
to have the TOPOSORT flag.
This is the same behavior as before, since indegree could not be
non-zero when TOPOSORT was unset.
Incidentally, this fixes the bug in show-branch where the 8th column
was not shown: show-branch sorts the commits in topological order,
assuming that all the commit flags are available for show-branch's
private matters.
But this was not true: TOPOSORT was identical to the flag corresponding
to the 8th ref. So the flags for the 8th column were unset by the
topological sorting.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The -u option to override the remote system's path to git-upload-pack was
being ignored by "git clone"; caused by a missing call to
transport_set_option to set TRANS_OPT_UPLOADPACK. Presumably this crept in
when git-clone was converted from shell to C.
Signed-off-by: Steve Haslam <shaslam@lastminute.com>
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>
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>
We recently let the user know explicitly that an empty
commit message will abort the commit. However, this adds yet
another line to the template; let's rephrase and re-wrap so
that this fits back on two lines.
This patch also makes the "fatal: empty commit message?"
warning a bit less scary, since this is now a "feature"
instead of an error. However, we retain the non-zero exit
status to indicate to callers that nothing was committed.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
OPT_INTEGER() works on an integer, not on an unsigned long. On a big
endian architecture with long larger than int, integer test gives bogus
results because of this bug.
Reported by H.Merijn Brand in HP-UX 64-bit environment.
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>
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>
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>
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>
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>
The -u option to override the remote system's path to git-upload-pack was
being ignored by "git clone"; caused by a missing call to
transport_set_option to set TRANS_OPT_UPLOADPACK. Presumably this crept in
when git-clone was converted from shell to C.
Signed-off-by: Steve Haslam <shaslam@lastminute.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
