read_packed_sha1 expectes its caller to free the buffer it returns, which
force_object_loose didn't do.
This leak is eventually triggered by "git gc", when it is manually invoked
or there are too many packs around, making gc totally unusable when there
are lots of unreachable objects.
Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de> Acked-by: Nicolas Pitre <nico@cam.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is the result of running make update-po and removing or fixing
the strings that were fuzzily matched. The ones that were fixed were
the ones where the only change was "git rev-list" to "git log", and
the "about gitk" message where the copyright year got updated.
To get xgettext to see the menu labels as needing translation, it
was necessary for arrange for them to be preceded by "mc". This
therefore changes makemenu to ignore the first element in each
menu item so that it can be "mc" in the makemenu call.
Fix "checkout --track -b newbranch" on detached HEAD
The test to make sure that checkout fails when --track was asked for and
we cannot set up tracking information in t7201 was wrong, and it turns out
that the implementation for that feature itself was buggy. This fixes it.
refactor handling of "other" files in ls-files and status
When the "git status" display code was originally converted
to C, we copied the code from ls-files to discover whether a
pathname returned by read_directory was an "other", or
untracked, file.
Much later, 5698454e updated the code in ls-files to handle
some new cases caused by gitlinks. This left the code in
wt-status.c broken: it would display submodule directories
as untracked directories. Nobody noticed until now, however,
because unless status.showUntrackedFiles was set to "all",
submodule directories were not actually reported by
read_directory. So the bug was only triggered in the
presence of a submodule _and_ this config option.
This patch pulls the ls-files code into a new function,
cache_name_is_other, and uses it in both places. This should
leave the ls-files functionality the same and fix the bug
in status.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
describe: Make --tags and --all match lightweight tags more often
If the caller supplies --tags they want the lightweight, unannotated
tags to be searched for a match. If a lightweight tag is closer
in the history, it should be matched, even if an annotated tag is
reachable further back in the commit chain.
The same applies with --all when matching any other type of ref.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Acked-By: Uwe Kleine-König <ukleinek@strlen.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git pull" has known about misguided "pull" that lets the underlying fetch
update the current branch for a long time. It also has known about
"git pull origin master" into a branch yet to be born.
These two workarounds however were not aware of the existence of each
other and did not work well together. This fixes it.
gitk: New way of constructing menus that allows for Alt+letter accelerators
This is inspired by patches from Robin Rosenberg but takes a different
approach. This adds a "makemenu" procedure for constructing menus
that allows the menu layout to be specified in a clear fashion, and
provides one place where the alt+letter accelerators can be detected
and handled.
The alt+letter accelerator is specified by putting an ampersand (&)
before the letter for the accelerator in the menu item name. (Two
ampersands in succession produce one ampersand in the menu item as
it appears on screen.) This is handled in makemenu.
We also add an mca procedure which is like mc but also does the
ampersand translation, for use when we want to refer to a menu item
by name. The mca name and the locations where we use it were
shamelessly stolen from Robin Rosenberg's patch.
This doesn't actually add any alt+letter accelerators yet.
* maint:
t1301-shared-repo.sh: don't let a default ACL interfere with the test
git-check-attr(1): add output and example sections
xdiff-interface.c: strip newline (and cr) from line before pattern matching
t4018-diff-funcname: demonstrate end of line funcname matching flaw
t4018-diff-funcname: rework negated last expression test
Typo "does not exists" when git remote update remote.
remote.c: correct the check for a leading '/' in a remote name
Add testcase to ensure merging an early part of a branch is done properly
t1301-shared-repo.sh: don't let a default ACL interfere with the test
This test creates files with several different umasks and expects their
permissions to be initialized according to the umask, so a default ACL on the
trash directory (which overrides the umask for files created in that directory)
causes the test to fail. To avoid that, remove the default ACL if possible with
setfacl(1).
Signed-off-by: Matt McCutchen <matt@mattmccutchen.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
rebase--interactive: fix parent rewriting for dropped commits
`rebase -i -p` got its rev-list of commits to keep by --left-right and
--cherry-pick. Adding --cherry-pick would drop commits that duplicated changes
already in the rebase target.
The dropped commits were then forgotten about when it came to rewriting the
parents of their descendents, so the descendents would get cherry-picked with
their old, unwritten parents and essentially make the rebase a no-op.
This commit adds a $DOTEST/dropped directory to remember dropped commits and
rewrite their children's parent as the dropped commit's possibly-rewritten
first-parent.
Signed-off-by: Stephen Haberman <stephen@exigencecorp.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
xdiff-interface.c: strip newline (and cr) from line before pattern matching
POSIX doth sayeth:
"In the regular expression processing described in IEEE Std 1003.1-2001,
the <newline> is regarded as an ordinary character and both a period and
a non-matching list can match one. ... Those utilities (like grep) that
do not allow <newline>s to match are responsible for eliminating any
<newline> from strings before matching against the RE."
Thus far git has not been removing the trailing newline from strings matched
against regular expression patterns. This has the effect that (quoting
Jonathan del Strother) "... a line containing just 'FUNCNAME' (terminated by
a newline) will be matched by the pattern '^(FUNCNAME.$)' but not
'^(FUNCNAME$)'", and more simply not '^FUNCNAME$'.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
t4018-diff-funcname: demonstrate end of line funcname matching flaw
Since the newline is not removed from lines before pattern matching, a
pattern cannot match to the end of the line using the '$' operator without
using an additional operator which will indirectly match the '\n' character.
Introduce a test which should pass, but which does not due to this flaw.
Signed-off-by: Brandon Casey <drafnel@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t4018-diff-funcname: rework negated last expression test
This test used the non-zero exit status of 'git diff' to indicate that a
negated funcname pattern, when placed last, was correctly rejected.
The problem with this is that 'git diff' always returns non-zero if it
finds differences in the files it is comparing, and the files must
contain differences in order to trigger the funcname pattern codepath.
Instead of checking for non-zero exit status, make sure the expected
error message is printed.
Signed-off-by: Brandon Casey <drafnel@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Fix binding for <Return> in sha1 entry field
This adds a break so that gitk doesn't go and execute the global
binding for <Return> (i.e. find next) when the user presses the
return key in the sha1 entry field to indicate that gitk should
jump to the commit identified by what they just put into the
sha1 field.
gitk: Clean up file encoding code and add enable/disable option
This adds an option allowing the user to select whether gitk should
look up per-file encoding settings using git check-attr or not. If
not, gitk uses the global encoding set in the git config (as reported
by git config --get gui.encoding) for all files, or if that is not
set, then the system encoding.
The option is controlled by a checkbox in the Edit->Preferences
window, and defaults to off for now because git check-attr is so
slow. When the user turns it on we discard any cached diff file
lists in treediffs, because we may not have encodings cached for
the files listed in those lists, meaning that getblobdiffline will
do it for each file, which will be really really slow.
This adjusts the limit of how many paths cache_gitattr passes to each
instance of git check-attr depending on whether we're running under
windows or not. Passing only 30 doesn't effectively amortize the
startup costs of git check-attr, but it's all we can do under windows
because of the 32k limit on arguments to a command. Under other OSes
we pass up to 1000.
Similarly we adjust how many lines gettreediffline processes depending
on whether we are doing per-file encodings so that we don't run for
too long. When we are, 500 seems to be a reasonable limit, leading
to gettreediffline taking about 60-70ms under Linux (almost all of
which is in cache_gitattr, unfortunately). This means that we can
take out the update call in cache_gitattr.
This adds a simple cache on [tclencoding]. Now that we get repeated
calls to translate the same encoding, this is useful.
This reindents the new code added in the last couple of commits to
conform to the gitk 4-space indent and makes various other improvements:
use regexp in gitattr and cache_gitattr instead of split + join + regsub,
make gui_encoding be the value from [tclencoding] to avoid having to
do [tcl_encoding $gui_encoding] in each call to get_path_encoding,
and print a warning message at startup if $gui_encoding isn't
supported by Tcl.
remote.c: correct the check for a leading '/' in a remote name
This test is supposed to disallow remote entries in the config file of the
form:
[remote "/foobar"]
...
The leading slash in '/foobar' is not acceptable.
Instead it was incorrectly testing that the subkey had no leading '/', which
had no effect since the subkey pointer was made to point at a '.' in the
preceding lines.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Acked-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Implement batch lookup and caching of encoding attrs
When the diff contains thousands of files, calling git-check-attr once
per file is very slow. With this patch gitk does attribute lookup in
batches of 30 files while reading the diff file list, which leads to a
very noticeable speedup.
It may be possible to reimplement this even more efficiently, if
git-check-attr is modified to support a --stdin-paths option.
Additionally, it should quote the ':' character in file paths, or
provide a more robust way of column separation.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com> Tested-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Paul Mackerras <paulus@samba.org>
This allows the encoding to be specified for file contents and used
when displaying files and diffs in the bottom-left pane. When
displaying diffs, the encoding for each diff hunk is that for the file
that the diff hunk is from, so it can change through the course of the
diff.
The encoding for file contents is determined as follows:
- File encoding defaults to the system encoding.
- It can be overridden by setting the gui.encoding option.
- Finally, the 'encoding' attribute is checked on
per-file basis; it has the last word.
Note: Since git-check-attr does not provide support for reading
attributes from trees, attribute lookup is done using files from the
working directory.
This also extends the range of supported encoding names, adding
ShiftJIS and Shift-JIS as aliases for Shift_JIS, and allowing
cp-*, cp_*, ibm-*, ibm_*, jis-* and jis_* as aliases for cp*,
ibm* and jis* respectively.
This also fixes some bugs in handling of non-ASCII filenames. Core
git apparently supports only locale-encoded filenames, so processing
is done using the system encoding.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com> Tested-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Paul Mackerras <paulus@samba.org>
compat/cygwin.c - Use cygwin's stat if core.filemode == true
Cygwin's POSIX emulation allows use of core.filemode true, unlike native
Window's implementation of stat / lstat, and Cygwin/git users who have
configured core.filemode true in various repositories will be very
unpleasantly surprised to find that git is no longer honoring that option.
So, this patch forces use of Cygwin's stat functions if core.filemode is
set true, regardless of any other considerations.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some confusing tutorials suggested that it would be a good idea to fetch
into the current branch with something like this:
git fetch origin master:master
(or even worse: the same command line with "pull" instead of "fetch").
While it might make sense to store what you want to pull, it typically is
plain wrong when the current branch is "master". This should only be
allowed when (an incorrect) "git pull origin master:master" tries to work
around by giving --update-head-ok to underlying "git fetch", and otherwise
we should refuse it, but somewhere along the lines we lost that behavior.
The check for the current branch is now _only_ performed in non-bare
repositories, which is an improvement from the original behaviour.
Some newer tests were depending on the broken behaviour of "git fetch"
this patch fixes, and have been adjusted.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Acked-by: Shawn O. Pearce <spearce@spearce.org> Acked-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since input parameters can be obtained both from CGI parameters and
PATH_INFO, we would like most of the code to be agnostic about the way
parameters were retrieved. We thus collect all the parameters into the
new %input_params hash, delaying validation after the collection is
completed.
Although the kludge removal is minimal at the moment, it makes life much
easier for future expansions such as more extensive PATH_INFO use or
other form of input such as command-line support.
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
"git diff <tree>{3,}": do not reverse order of arguments
According to the message of commit 0fe7c1de16f71312e6adac4b85bddf0d62a47168,
"git diff" with three or more trees expects the merged tree first followed by
the parents, in order. However, this command reversed the order of its
arguments, resulting in confusing diffs. A comment /* Again, the revs are all
reverse */ suggested there was a reason for this, but I can't figure out the
reason, so I removed the reversal of the arguments. Test case included.
Signed-off-by: Matt McCutchen <matt@mattmccutchen.net> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Replace calls to strbuf_init(&foo, 0) with STRBUF_INIT initializer
Many call sites use strbuf_init(&foo, 0) to initialize local
strbuf variable "foo" which has not been accessed since its
declaration. These can be replaced with a static initialization
using the STRBUF_INIT macro which is just as readable, saves a
function call, and takes up fewer lines.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If verification of path failed, it is always better to print an
error message saying this than relying on the caller function to
print a meaningful error message (especially when the callee already
prints error message for another situation).
Because the callers of add_index_entry_with_check() did not print
any error message, it resulted that the user would not notice the
problem when checkout of an invalid path failed.
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
t9001: use older Getopt::Long boolean prefix '--no' rather than '--no-'
Since dbf5e1e9, the '--no-validate' option is a Getopt::Long boolean
option. The '--no-' prefix (as in --no-validate) for boolean options
is not supported in Getopt::Long version 2.32 which was released with
Perl 5.8.0. This version only supports '--no' as in '--novalidate'.
More recent versions of Getopt::Long, such as version 2.34, support
either prefix. So use the older form in the tests.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We carefully verify that the input to git-apply is sane,
including cross-checking that the filenames we see in "+++"
headers match what was provided on the command line of "diff
--git". When --directory is used, however, we ended up
comparing the unadorned name to one with the prepended root,
causing us to complain about a mismatch.
We simply need to prepend the root directory, if any, when
pulling the name out of the git header.
Signed-off-by: Jeff King <peff@peff.net> Acked-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Show a round number of bytes of large untracked text files
If an untracked text file is selected, then its contents are displayed
instead of a diff. If the file is large, then the following hint is
inserted at the top:
* Untracked file is 14774881 bytes.
* Showing only first 131072 bytes.
Why exactly 131072 bytes? With this patch it is 100000 bytes.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Add a search command to the blame viewer.
One of the largest deficiencies in the blame viewer at
the moment is the impossibility to search for a text
string. This commit fixes it by adding a Firefox-like
search panel to the viewer.
The panel can be shown by pressing F7 or clicking a
menu entry, and is hidden by pressing Esc. Find Next
is available through the F3 key.
Implementation is based on the gitk code, but heavily
refactored. It now also supports case-insensitive
searches, and uses the text box background color to
signal success or failure of the search.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
On modern high-resolution monitors the blame viewer
window is very high, yet too narrow. This patch
makes it gravitate to a more sane resolution, which
takes the font size into account.
It also changes the default text view size to 80% of
the window, and slightly modifies the border decorations
for better appearance.
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
0aea2842 (Make Ctrl-T safe to use for conflicting files) introduced a new
case, but forgot the '-' to indicate that it shares the body with the
subsequent case label.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
rebase -i: do not fail when there is no commit to cherry-pick
In case there is no commit to apply (for example because you rebase to
upstream and all your local patches have been applied there), do not
fail. The non-interactive rebase already behaves that way.
Do this by introducing a new command, "noop", which is substituted for
an empty commit list, so that deleting the commit list can still abort
as before.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Since v1.6.0.2~13^2~ the completion of a thin pack uses sha1write() for
its ability to compute a SHA1 on the written data. This also provides
data buffering which, along with commit 92392b4a45, will confuse pread()
whenever an appended object is 1) freed due to memory pressure because
of the depth-first delta processing, and 2) needed again because it has
many delta children, and 3) its data is still buffered by sha1write().
Let's fix the issue by simply forcing cached data out when such an
object is written so it can be pread()'d at leisure.
Signed-off-by: Nicolas Pitre <nico@cam.org> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The new -v option forces the progressbar, even in case the output
is not a terminal. This can be useful if the caller is an IDE or
wrapper which wants to scrape the progressbar from stderr and show
its information in a different format.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
* pb/gitweb:
gitweb: Support for simple project search form
gitweb: Make the by_tag filter delve in forks as well
gitweb: Support for tag clouds
gitweb: Add support for extending the action bar with custom links
gitweb: Sort the list of forks on the summary page by age
gitweb: Clean-up sorting of project list
* mw/sendemail:
bash completion: Add --[no-]validate to "git send-email"
send-email: signedoffcc -> signedoffbycc, but handle both
Docs: send-email: Create logical groupings for man text
Docs: send-email: Create logical groupings for --help text
Docs: send-email: Remove unnecessary config variable description
Docs: send-email: --chain_reply_to -> --[no-]chain-reply-to
send-email: change --no-validate to boolean --[no-]validate
Docs: send-email: Man page option ordering
Docs: send-email usage text much sexier
Docs: send-email's usage text and man page mention same options
* maint:
builtin-apply: fix typo leading to stack corruption
git-stash.sh: fix flawed fix of invalid ref handling (commit da65e7c1)
builtin-merge.c: allocate correct amount of memory
Makefile: do not set NEEDS_LIBICONV for Solaris 8
rebase -i: remove leftover debugging
rebase -i: proper prepare-commit-msg hook argument when squashing
git-stash.sh: fix flawed fix of invalid ref handling (commit da65e7c1)
The referenced commit tried to fix a flaw in stash's handling of a user
supplied invalid ref. i.e. 'git stash apply fake_ref@{0}' should fail
instead of applying stash@{0}. But, it did so in a naive way by avoiding the
use of the --default option of rev-parse, and instead manually supplied the
default revision if the user supplied an empty command line. This prevented
a common usage scenario of supplying flags on the stash command line (i.e.
non-empty command line) which would be parsed by lower level git commands,
without supplying a specific revision. This should fall back to the default
revision, but now it causes an error. e.g. 'git stash show -p'
The correct fix is to use the --verify option of rev-parse, which fails
properly if an invalid ref is supplied, and still allows falling back to a
default ref when one is not supplied.
Convert stash-drop to use --verify while we're at it, since specifying
multiple revisions for any of these commands is also an error and --verify
makes it so.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
But while we're at it, change the allocation to reference the
variable it is allocating memory for to try to prevent a similar
mistake, for example if the type is changed, in the future.
* maint:
Do not use errno when pread() returns 0
git init: --bare/--shared overrides system/global config
git-push.txt: Describe --repo option in more detail
git rm: refresh index before up-to-date check
Fix a few typos in relnotes
If we use pread() while at the end of the file, it will return 0, which is
not an error from the operating system point of view. In this case, errno
has not been set and must not be used.
Signed-off-by: Samuel Tardieu <sam@rfc1149.net> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If core.bare or core.sharedRepository are set in /etc/gitconfig or
~/.gitconfig, then 'git init' will read the values when constructing a
new config file; reading them, however, will override the values
specified on the command line. In the case of --bare, this ends up
causing a segfault, without the repository being properly initialised;
in the case of --shared, the permissions are set according to the
existing config settings, not what was specified on the command line.
This fix saves any specified values for --bare and --shared prior to
reading existing config settings, and restores them after reading but
before writing the new config file. core.bare is ignored in all
situations, while core.sharedRepository will only be used if --shared
is not specified to git init.
Also includes testcases which use a specified global config file
override, demonstrating the former failure scenario.
Signed-off-by: Deskin Miller <deskinm@umich.edu> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-push.txt: Describe --repo option in more detail
The --repo option was described in a way that the reader would have to
assume that it is the same as the <repository> parameter. But it actually
servers a purpose, which is now written down.
Furthermore, the --mirror option was missing from the synopsis.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
It is sometimes desirable to disable the safety net of pre-rebase hook
when the user knows what he is doing (for example, when the original
changes on the branch have not been shown to the public yet).
This teaches --no-verify option to git-rebase, which is similar to the way
pre-commit hook is bypassed by git-commit.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Add git-svn branch to allow branch creation in SVN repositories
[ew: fixed a warning to stderr causing t9108 to fail]
Signed-off-by: Florian Ragwitz <rafl@debian.org> Signed-off-by: Deskin Miller <deskinm@umich.edu> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
* maint:
Update release notes for 1.6.0.3
Teach rebase -i to honor pre-rebase hook
docs: describe pre-rebase hook
do not segfault if make_cache_entry failed
make prefix_path() never return NULL
fix bogus "diff --git" header from "diff --no-index"
Fix fetch/clone --quiet when stdout is connected
builtin-blame: Fix blame -C -C with submodules.
bash: remove fetch, push, pull dashed form leftovers
The original git-rebase honored pre-rebase hook so that public branches
can be protected from getting rebased, but rebase --interactive ignored
the hook entirely. This fixes it.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
There are 9 places where prefix_path is called, and only in one of
them the returned pointer was checked to be non-zero and only to
call exit(128) as it is usually done by die(). In other 8 places,
the returned value was not checked and it caused SIGSEGV when a
path outside of the working tree was used. For instance, running
git update-index --add /some/path/outside
caused SIGSEGV.
This patch changes prefix_path() to die if the path is outside of
the repository, so it never returns NULL.
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
fix bogus "diff --git" header from "diff --no-index"
When "git diff --no-index" is given an absolute pathname, it
would generate a diff header with the absolute path
prepended by the prefix, like:
diff --git a/dev/null b/foo
Not only is this nonsensical, and not only does it violate
the description of diffs given in git-diff(1), but it would
produce broken binary diffs. Unlike text diffs, the binary
diffs don't contain the filenames anywhere else, and so "git
apply" relies on this header to figure out the filename.
This patch just refuses to use an invalid name for anything
visible in the diff.
Now, this fixes the "git diff --no-index --binary a
/dev/null" kind of case (and we'll end up using "a" as the
basename), but some other insane cases are impossible to
handle. If you do
git diff --no-index --binary a /bin/echo
you'll still get a patch like
diff --git a/a b/bin/echo
old mode 100644
new mode 100755
index ...
and "git apply" will refuse to apply it for a couple of
reasons, and the diff is simply bogus.
And that, btw, is no longer a bug, I think. It's impossible
to know whethe the user meant for the patch to be a rename
or not. And as such, refusing to apply it because you don't
know what name you should use is probably _exactly_ the
right thing to do!
Original problem reported by Imre Deak. Test script and problem
description by Jeff King.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Commit c85db254 changed the format of the message produced
by "git commit" when creating a commit. This patch updates
the example session in the tutorial to the new format.
It also adds in the missing diffstat summary lines, which
should have been added long ago.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
builtin-commit: use reduce_heads() only when appropriate
Since commit 6bb6b034 (builtin-commit: use commit_tree(), 2008-09-10),
builtin-commit performs a reduce_heads() unconditionally. However,
it's not always needed, and in some cases even harmful.
reduce_heads() is not needed for the initial commit or for an
"ordinary" commit, because they don't have any or have only one
parent, respectively.
reduce_heads() must be avoided when 'git commit' is run after a 'git
merge --no-ff --no-commit', otherwise it will turn the
non-fast-forward merge into fast-forward. For the same reason,
reduce_heads() must be avoided when amending such a merge commit.
To resolve this issue, 'git merge' will write info about whether
fast-forward is allowed or not to $GIT_DIR/MERGE_MODE. Based on this
info, 'git commit' will only perform reduce_heads() when it's
committing a merge and fast-forward is enabled.
Also add test cases to ensure that non-fast-forward merges are
committed and amended properly.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: SZEDER Gábor <szeder@ira.uka.de> Signed-off-by: Shawn O. Pearce <spearce@spearce.org>