diff: use large integers for diffstat calculations
The diffstat "added" and "changed" fields generally store
line counts; however, for binary files, they store file
sizes. Since we store and print these values as ints, a
diffstat on a file larger than 2G can show a negative size.
Instead, let's use uintmax_t, which should be at least 64
bits on modern platforms.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t1010-mktree: Adjust expected result to code and documentation
The last two tests here were always supposed to fail in the sense
that, according to code and documentation, mktree should read non-recursive
ls-tree output, but not recursive one, and therefore explicitely refuses
to deal with slashes.
Adjust the test (must_fail) so that it succeeds when mktree dies on
slashes.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Consider an evil merge of two commits A and B, both of which have a
file 'foo', but the merge result does not have that file.
The combined-diff code learned in 4462731 (combine-diff: do not punt
on removed or added files., 2006-02-06) to concisely show only the
removal, since that is the evil part and the previous contents are
presumably uninteresting.
However, to diagnose an empty merge result, it overloaded the variable
that holds the file's length. This means that the check also triggers
for truncated files. Consequently, such files were not shown in the
diff at all despite the merge being clearly evil.
Fix this by adding a new variable that distinguishes whether the file
was deleted (which is the case 4462731 handled) or truncated. In the
truncated case, we show the full combined diff again, which is rather
spammy but at least does not hide the evilness.
Reported-by: David Martínez Martí <desarrollo@gestiweb.com> Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since recently "git diff --submodule" prints out extra lines when the
submodule contains untracked or modified files. Show all those lines of
one submodule under the same header.
Also for newly added or removed submodules the submodule name contained
trailing garbage because the extraction of the name was not done right.
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de> Signed-off-by: Paul Mackerras <paulus@samba.org>
exec_cmd.c: replace hard-coded path list with one from <paths.h>
The default executable path list used by exec_cmd.c is hard-coded to
be "/usr/local/bin:/usr/bin:/bin". Use an appropriate value for the
system from <paths.h> when available.
Add HAVE_PATHS_H make variables and enable it on Linux, FreeBSD,
NetBSD, OpenBSD and GNU where it is known to exist for now. Somebody
else may want to do an autoconf support later.
Signed-off-by: Chris Webb <chris@arachsys.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
v1.7.0-rc0~18^2 (branch -d: base the "already-merged" safety on the
branch it merges with, 2009-12-29) taught ‘git branch’ a new heuristic
for when it is safe to delete a branch without forcing the issue. It
is safe to delete a branch "topic" without second thought if:
- the branch "topic" is set up to pull from a (remote-tracking,
usually) branch and is fully merged in that "upstream" branch, or
- there is no branch.topic.merge configuration and branch "topic" is
fully merged in the current HEAD.
Update the man page to acknowledge the new rules.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The makefile snippets that would land in these directories are already
being ignored. Ignore the directories instead so they don’t show up
in ‘git clean -n’ output.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Most of these tests are removing files, environment variables, and
configuration that might interfere outside the test. Putting these
clean-up commands in the test (in the same spirit as v1.7.1-rc0~59,
2010-03-20) means that errors during setup will be caught quickly and
non-error text will be suppressed without -v.
While at it, apply some other minor fixes:
- do not rely on the shell to export variables defined with the same
command as a function call
- avoid whitespace immediately after the > redirection operator, for
consistency with the style of other tests
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3cb22b8 (Per-ref reflog expiry configuration, 2008-06-15) added support
for setting the expiry parameters differently for different reflog, but
it was never documented.
am -3: recover the diagnostic messages for corrupt patches
"git am -3" first tries to apply the patch without any extra trick, and
applies it to a synthesized tree for 3-way merge after the first attempt
fails. "git apply" exits with status 1 for a patch that is well-formed
but is not applicable (and it dies on other errors with non-zereo, non-1
status) and has an optimization to fall back to the 3-way merge only in
the case.
An earlier patch 3ddd170 (am: suppress apply errors when using 3-way,
2009-06-16) squelched diagnostic messages from the first attempt, not to
be shown to the end user. This worked reasonably well if the reason the
first application failed was because the patch was made against a wrong
version.
When the patch is corrupt (e.g. line-wrapped or leading whitespaces got
dropped), however, because the second patch application is not even
attempted, the error message from the first application is never shown
and is forever lost. This message is necessary to locate where the patch
is corrupt and fix it up.
We could fix this issue by reverting 3dd170, or keeping the error message
to somewhere and showing it, but because this is an error codepath, the
easiest is to disable the optimization. The second patch application is
attempted even when the input is corrupt, and it will notice, diagnose,
and stop with an error message.
diff: add --word-diff option that generalizes --color-words
This teaches the --color-words engine a more general interface that
supports two new modes:
* --word-diff=plain, inspired by the 'wdiff' utility (most similar to
'wdiff -n <old> <new>'): uses delimiters [-removed-] and {+added+}
* --word-diff=porcelain, which generates an ad-hoc machine readable
format:
- each diff unit is prefixed by [-+ ] and terminated by newline as
in unified diff
- newlines in the input are output as a line consisting only of a
tilde '~'
Both of these formats still support color if it is enabled, using it
to highlight the differences. --color-words becomes a synonym for
--word-diff=color, which is the color-only format. Also adds some
compatibility/convenience options.
Thanks to Junio C Hamano and Miles Bader for good ideas.
Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using git log --pretty='%N' without an explicit --show-notes, git
would segfault. This patches fixes this behaviour by loading the needed
notes datastructures if --pretty is used and the format contains %N.
When --pretty='%N' is used together with --no-notes, %N won't be
expanded.
This is an extension to a proposed patch by Jeff King.
Signed-off-by: Johannes Gilger <heipei@hackvalue.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If more than one annotated tag points at the same commit, use the
tag whose tagger field has a more recent date stamp. This resolves
non-deterministic cases where the maintainer has done:
$ git tag -a -m "2.1-rc1" v2.1-rc1 deadbeef
$ git tag -a -m "2.1" v2.1 deadbeef
If the tag is an older-style annotated tag with no tagger date, we
assume a date stamp at the UNIX epoch. This will cause us to prefer
an annotated tag that has a valid date.
We could also try to consider the tag object chain, favoring a tag
that "includes" another one:
$ git tag -a -m "2.1-rc0" v2.1-rc1 deadbeef
$ git tag -a -m "2.1" v2.1 v2.1-rc1
However traversing the tag's object chain looking for inclusion
is much more complicated. Its already very likely that even in
these cases the v2.1 tag will have a more recent tagger date than
v2.1-rc1, so with this change describe should still resolve this
by selecting the more recent v2.1.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Just like with committer dates, we parse the tagger date into the
struct tag so its available for further downstream processing.
However since the tagger header was not introduced until Git 0.99.1
we must consider it optional. For tags missing this header we use
the default date of 0.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
tag.c: Refactor parse_tag_buffer to be saner to program
This code was horribly ugly to follow. The structure of the headers
in an annotated tag object must follow a prescribed order, and most
of these are required. Simplify the entire parsing logic by going
through the headers in the order they are supposed to appear in,
acting on each header as its identified in the buffer.
This change has the same behavior as the older version, its just
easier to read and maintain.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
These lines were incorrectly indented with spaces, violating our
coding style. Its annoying to read with 4 position tab stops, so
fix the indentation to be correct.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
index-pack: smarter memory usage when appending objects
In the same spirit as commit 9892bebafe, let's avoid allocating the full
buffer for the deflated data in write_compressed() in order to write it.
Let's deflate and write the data in chunks instead to reduce memory
usage.
Signed-off-by: Nicolas Pitre <nico@fluxnic.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
index-pack: smarter memory usage when resolving deltas
In the same spirit as commit 9892bebafe, let's avoid allocating the full
buffer for the deflated data in get_data_from_pack() in order to inflate
it. Let's read and inflate the data in chunks instead to reduce memory
usage.
Signed-off-by: Nicolas Pitre <nico@fluxnic.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Windows: start_command: Support non-NULL dir in struct child_process
A caller of start_command can set the member 'dir' to a directory to
request that the child process starts with that directory as CWD. The first
user of this feature was added recently in eee49b6 (Teach diff --submodule
and status to handle .git files in submodules).
On Windows, we have been lazy and had not implemented support for this
feature, yet. This fixes the shortcoming.
Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In particular, add a missing && to the update --init test.
The goal is to make it clearer what happened when one of these
tests fails. The update --init test is currently (consistently)
failing on a few unusual machines.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
A new reader may not realize what properties the $submodurl
repository needs to have.
One of the tests is checking that ‘submodule add -b foo’ creates
a ‘foo’ branch. Put this test in context by checking that
without -b, no ‘foo’ branch is created.
While at it, make sure each added submodule is a reasonable
repository, with clean index, no stray files, and so on.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The setup in t7400-submodule-basic does a number of different
things to support different tests. Splitting it up makes the
test a little easier to read and should provide an opportunity
to move each piece of setup closer to the tests that require it.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using macros it is otherwise hard to know whether an
attribute set by the macro should override an already set
attribute. Consider the following .gitattributes file:
Without this patch both foo.bin and bar.bin will have
the ident attribute set, which is probably not what
the user expects. With this patch foo.bin will have an
unset ident attribute, while bar.bin will have it set.
Signed-off-by: Henrik Grubbström <grubba@grubba.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* mr/gitweb-jsmin:
gitweb: update INSTALL to use shorter make target
gitweb: add documentation to INSTALL regarding gitweb.js
instaweb: add minification awareness
Gitweb: add autoconfigure support for minifiers
Gitweb: add support for minifying gitweb.css
Gitweb: add ignore and clean rules for minified files
The way the code stored --smtp-domain was unlike its handling of other
similar options. Bring it in line with the others by:
- Renaming $mail_domain to $smtp_domain to match the command line
option. Also move its declaration from near the top of the file to
near other option variables.
- Removing $mail_domain_default. The variable was used once and only
served to move the default away from where it gets used.
- Adding a sendemail.smtpdomain config option. smtp-domain was the
only SMTP configuration option that couldn't be set in the user's
.gitconfig.
Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Although Net::Domain::domainname attempts to be very thorough, the
host's configuration can still refuse to give a FQDN. Check to see if
what we receive contains a dot as a basic sanity check.
Since the same condition is used twice and getting complex, let's move
it to a new function.
Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
As Jakub Narebski pointed out on the list, Perl code usually prefers
sub func {
}
over
sub func
{
}
git-send-email.perl is somewhat inconsistent in its style, with 23
subroutines using the first style and 6 using the second. Convert the
few odd subroutines so that the code matches normal Perl style.
Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach diff --submodule and status to handle .git files in submodules
The simple test for an existing .git directory gives an incorrect result
if .git is a file that records "gitdir: overthere". So for submodules that
use a .git file, "git status" and the diff family - when the "--submodule"
option is given - did assume the submodule was not populated at all when
a .git file was used, thus generating wrong output or no output at all.
This is fixed by using read_gitfile_gently() to get the correct location
of the .git directory. While at it, is_submodule_modified() was cleaned up
to use the "dir" member of "struct child_process" instead of setting the
GIT_WORK_TREE and GIT_DIR environment variables.
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Let check_preimage() use memset() to initialize "struct checkout"
Every code site except check_preimage() uses either memset() or declares
a static instance of "struct checkout" to achieve proper initialization.
Lets use memset() instead of explicit initialization of all members here
too to be on the safe side in case this structure is expanded someday.
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
- use "<options>" instead of just "options".
- use "[<repository> [<refspec>...]]" to indicate that <repository> and
<refspec> are optional, and that <refspec> cannot be specified
without specifying <repository>.
Note that when called without specifying <repository> (eg. "git fetch
-f"), it is accurate to say that the "git fetch [<options>]
[<repository> ...]" case takes precedence over "git fetch [<options>]
<group>".
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
remote-curl: ensure that URLs have a trailing slash
Previously, we blindly assumed that URLs passed to the remote-curl
helper did not end with a trailing slash.
Use the convenience function end_url_with_slash() from http.[ch] to
ensure that URLs have a trailing slash on invocation of the remote-curl
helper, and use the URL as one with a trailing slash throughout.
It is possible for users to pass a URL with a trailing slash to
remote-curl, by, say, setting it in remote.<name>.url in their git
config. The resulting requests have an empty path component (//) and may
break implementations of the http git protocol.
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The sed script intended to add a standard opening to python scripts
was non-compatible and overly complex. Simplifying it down to a set
of one-liners removes the compatibility issues of newlines. Moving
the environment alterations from the Makefile to the python scripts
makes also makes the scripts easier to run in-place.
Specifically, the new sed script:
- Alters the shebang line to use the configured Python.
- Alters any os.getenv("GITPYTHONLIB") calls to use @@INSTLIBDIR@@ as the
default. This will replace any existing default or add a default if
none is provided.
- Replaces the @@INSTLIBDIR@@ placeholder with the directory git installs
its python libraries to.
The last two steps could be combined into a single step, but is left
separate in case someone has another need for @@INSTLIBDIR@@ in their
script.
Suggested-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Brian Gernhardt <brian@gernhardtsoftware.com> Acked-by: Sverre Rabbelier <srabbelier@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
branch: say "Reset to" in reflog entries for 'git branch -f' operations
In 5f856dd (fix reflog entries for "git-branch"), it is mentioned that
'git branch -f' is intended to be equivalent to 'git reset'. Since we
usually say "reset to <commit>" in the git-reset Documentation and
elsewhere, it would make sense to say "Reset to" here as well, instead
of "Reset from" previously.
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
log.decorate: only ignore it under "log --pretty=raw"
Unlike notes that are often multi-line and disrupting to be placed in many
output formats, a decoration is designed to be a small token that can be
tacked after an existing line of the output where a commit object name sits.
Disabling log.decorate for something like "log --oneline" would defeat the
purpose of the configuration.
We _might_ want to change it further in the future to force scripts that
do not want to be broken by random end user configurations to explicitly
say "log --no-decorate", but that would be an incompatible change that
needs the usual multi-release-cycle deprecation process.
Because log.decorate now shows decorations for --pretty=oneline,
we must explicitly turn it off when scripting. Otherwise,
users with log.decorate set will get cruft like:
$ git stash
Saved working directory and index state WIP on master: 2c1f7f5 (HEAD, master) commit subject
Instead of adding --no-decorate to the log command line,
let's just use the rev-list plumbing interface instead,
which does the right thing.
git-submodule has a similar call. Since it just counts the
commit lines, nothing is broken, but let's switch it, too,
for the sake of consistency and cleanliness.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
ls-remote: fall-back to default remotes when no remote specified
Instead of breaking execution when no remote (as specified in the
variable dest) is specified when git-ls-remote is invoked, continue on
and let remote_get() handle it.
This way, we are able to use the default remotes (eg. "origin",
branch.<name>.remote), as git-fetch, git-push, and other users of
remote_get(), do.
If no suitable remote is found, exit with a message describing the
issue, instead of just the usage text, as we do previously.
Add several tests to check that git-ls-remote handles the
no-remote-specified situation.
Also add a test that "git ls-remote <pattern>" does not work; we are
unable to guess the remote in that situation, as are git-fetch and
git-push.
In that test, we are testing for messages coming from two separate
processes, but we should be OK, because the second message is triggered
by closing the fd which must happen after the first message is printed.
(analysis by Jeff King.)
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Thread-safe xmalloc and xrealloc needs a recursive mutex
The mutex used to protect object access (read_mutex) may need to be
acquired recursively. Introduce init_recursive_mutex() helper function
in thread-utils.c that constructs a mutex with the PHREAD_MUTEX_RECURSIVE
attribute.
pthread_mutex_init() emulation on Win32 is already recursive as it is
implemented on top of the CRITICAL_SECTION type, which is recursive.
apply: Allow blank *trailing* context lines to match beyond EOF
In 51667147be, "git apply --whitespace=fix" was extended to
allow a blank context line to match beyond the end of the file,
but only if the context line was in the leading part of the
hunk (i.e. the hunk inserted additional contents at the end
of the file).
Drop the restriction that the context line must be in the
leading part of the hunk, thus allowing a file to be changed
from:
a
(blank line)
to:
b
a
(blank line)
Note that the blank line will be kept, because "--whitespace=fix"
only removes trailing blank lines that a hunk would add, never
trailing blank lines in the context.
Signed-off-by: Björn Gustavsson <bgustavsson@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Even if you use imap-send to throw your drafts in the outbox, using their
web interface will mangle your patches. Clarify that the imap-send is
meant to be used together with a real MUA that can use IMAP drafts, and
remove instructions related to the web interface, which is irrelevant.
Add description of send-email as an alternative.
Use --cover-letter, and do not use -C nor --no-color, on the example
command line for format-patch.
Add option to git-commit to allow empty log messages
Change git-commit(1) to accept the --allow-empty-message option
to allow a commit with an empty message. This is analogous to the
existing --allow-empty option which allows a commit that records
no changes. As these are mainly for interoperating with foreign SCM
systems, and are not meant for normal use, ensure that "git commit -h"
does not talk about them.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The option tells the command to expire older reflog entries that refer to
commits that are no longer reachable from the tip of the ref the reflog is
associated with. To avoid repeated merge_base() invocations, we used to
mark commits that are known to be reachable by walking the history from
the tip until we hit commits that are older than expire-total (which is
the timestamp before which all the reflog entries are expired).
However, it is a different matter if a commit is _not_ known to be
reachable and the commit is known to be unreachable. Because you can
rewind a ref to an ancient commit and then reset it back to the original
tip, a recent reflog entry can point at a commit that older than the
expire-total timestamp and we shouldn't expire it. For that reason, we
had to run merge-base computation when a commit is _not_ known to be
reachable.
This introduces a lazy/on-demand traversal of the history to mark
reachable commits in steps. As before, we mark commits that are newer
than expire-total to optimize the normal case before walking reflog, but
we dig deeper from the commits the initial step left off when we encounter
a commit that is not known to be reachable.
In many places in test suite we have "sleep"s that do not have to be
there.
- I do not simply see the point of the one in t3500. It may be making
sure that the timestamp order of commits generated during the test is
stable, in which case test_tick is the right ingredient to use without
wasting tester's time.
- The one in t4011 is to make sure that the plumbing diff-index notices
the stat-dirtyness of a removed then identically recreated symlink.
Keeping the old symlink around to make sure that a newly created
symlink gets different ino would be sufficient for that purpose.
- The one in t7600 is to make sure that "git merge" does not get confused
by stat-dirty "file" in the working tree. Again, keeping the old file
around and creating an identical copy to ensure a different ino would
be sufficient for that purpose.
The "racy git" tests in t0010 are inherently about mtime between the index
itself and index entries. The "sleep" in that test must stay as they are.
* sb/fmt-merge-msg:
fmt-merge-msg: hide summary option
fmt-merge-msg: remove custom string_list implementation
string-list: add unsorted_string_list_lookup()
fmt-merge-msg: use pretty.c routines
t6200: test fmt-merge-msg more
t6200: modernize with test_tick
fmt-merge-msg: be quiet if nothing to merge
Many scripts, most notably gitk, rely on output from the log family of
command not to be molested by random user configuration. This is
especially true when --pretty=raw is given.
Just like we disable notes output unless the command line explicitly
asks for --show-notes, disable the decoration code unless --decorate is
given explicitly from the command line and --pretty or --oneline is
given.
Could it be that you are using a third party iconv library?
I've experienced this on another system and the problem was related
to curl. In that case, curl was linked against an external iconv and
not the native library, so if I tried to build with curl support, I had
to also build against the external iconv library.
Rename ONE_FILESYSTEM to DISCOVERY_ACROSS_FILESYSTEM
If a missing ONE_FILESYSTEM defaults to true, the only users who set this
variable set it to false to tell git not to limit the discovery to one
filesystem; there are too many negations in one sentence to make a simple
panda brain dizzy.
Use the variable GIT_DISCOVERY_ACROSS_FILESYSTEM that changes the
behaviour from the default "limit to one filesystem" to "cross the
boundary as I ask you to"; makes the semantics much more straight
forward.