git-clone: aggressively optimize local clone behaviour.
This changes the behaviour of cloning from a repository on the
local machine, by defaulting to "-l" (use hardlinks to share
files under .git/objects) and making "-l" a no-op. A new
option, --no-hardlinks, is also added to cause file-level copy
of files under .git/objects while still avoiding the normal
"pack to pipe, then receive and index pack" network transfer
overhead. The old behaviour of local cloning without -l nor -s
is availble by specifying the source repository with the newly
introduced file:///path/to/repo.git/ syntax (i.e. "same as
network" cloning).
* With --no-hardlinks (i.e. have all .git/objects/ copied via
cpio) would not catch the source repository corruption, and
also risks corrupted recipient repository if an
alpha-particle hits memory cell while indexing and resolving
deltas. As long as the recipient is created uncorrupted, you
have a good back-up.
* same-as-network is expensive, but it would catch the breakage
of the source repository. It still risks corrupted recipient
repository due to hardware failure. As long as the recipient
is created uncorrupted, you have a good back-up.
* The new default on the same filesystem, as long as the source
repository is healthy, it is very likely that the recipient
would be, too. Also it is very cheap. You do not get any
back-up benefit, though.
None of the method is resilient against the source repository
corruption, so let's discount that from the comparison. Then
the difference with and without --no-hardlinks matters primarily
if you value the back-up benefit or not. If you want to use the
cloned repository as a back-up, then it is cheaper to do a clone
with --no-hardlinks and two git-fsck (source before clone,
recipient after clone) than same-as-network clone, especially as
you are likely to do a git-fsck on the recipient if you are so
paranoid anyway.
Which leads me to believe that being able to use file:/// is
probably a good idea, if only for testability, but probably of
little practical value. We default to hardlinked clone for
everyday use, and paranoids can use --no-hardlinks as a way to
make a back-up.
When calling "git rebase -i <upstream> <branch>", git should switch
to <branch> first. This worked before, but I broke it by my
"Shut git rebase -i up" patch.
Fix that, and add a test to make sure that it does not break again.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Acked-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-sh-setup.sh: make GIT_EDITOR/core.editor/VISUAL/EDITOR accept commands
The previous code only allowed specifying a single executable rather
than a complete command like "emacsclient --alternate-editor vi" in
those variables. Since VISUAL/EDITOR appear to be traditionally
passed to a shell for interpretation (as corroborated with "less",
"mail" and "mailx", while the really ancient "more" indeed allows only
an executable name), the shell function git_editor has been amended
appropriately.
"eval" is employed to have quotes and similar interpreted _after_
expansion, so that specifying
EDITOR='"/home/dak/My Commands/notepad.exe"'
can be used for actually using commands with blanks.
Instead of passing just the first argument of git_editor on, we pass
all of them (so that +lineno might be employed at a later point of
time, or so that multiple files may be edited when appropriate).
Strictly speaking, there is a change in behavior: when
git config core.editor
returns a valid but empty string, the fallbacks are still searched.
This is more consistent, and the old code was problematic with regard
to multiple blanks. Putting in additional quotes might have worked,
but quotes inside of command substitution inside of quotes is nasty
enough to not reliably work the same across "Bourne shells".
Signed-off-by: David Kastrup <dak@gnu.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
get_relative_cwd(): clarify why it handles dir == NULL
The comment did not make a good case why it makes sense.
Clarify, and remove stale comment about the caller being lazy.
The behaviour on NULL input is pretty much intentional.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The old version of work-tree support was an unholy mess, barely readable,
and not to the point.
For example, why do you have to provide a worktree, when it is not used?
As in "git status". Now it works.
Another riddle was: if you can have work trees inside the git dir, why
are some programs complaining that they need a work tree?
IOW it is allowed to call
$ git --git-dir=../ --work-tree=. bla
when you really want to. In this case, you are both in the git directory
and in the working tree. So, programs have to actually test for the right
thing, namely if they are inside a working tree, and not if they are
inside a git directory.
Also, GIT_DIR=../.git should behave the same as if no GIT_DIR was
specified, unless there is a repository in the current working directory.
It does now.
The logic to determine if a repository is bare, or has a work tree
(tertium non datur), is this:
--work-tree=bla overrides GIT_WORK_TREE, which overrides core.bare = true,
which overrides core.worktree, which overrides GIT_DIR/.. when GIT_DIR
ends in /.git, which overrides the directory in which .git/ was found.
In related news, a long standing bug was fixed: when in .git/bla/x.git/,
which is a bare repository, git formerly assumed ../.. to be the
appropriate git dir. This problem was reported by Shawn Pearce to have
caused much pain, where a colleague mistakenly ran "git init" in "/" a
long time ago, and bare repositories just would not work.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add functions get_relative_cwd() and is_inside_dir()
The function get_relative_cwd() works just as getcwd(), only that it
takes an absolute path as additional parameter, returning the prefix
of the current working directory relative to the given path. If the
cwd is no subdirectory of the given path, it returns NULL.
is_inside_dir() is just a trivial wrapper over get_relative_cwd().
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch adds convenience functions to work with absolute paths.
The function is_absolute_path() should help the efforts to integrate
the MinGW fork.
Note that make_absolute_path() returns a pointer to a static buffer.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
I switched git.kernel.dk to --base-path a few minutes ago, to get rid of
a /data/git postfix in the posted urls. But transitioning is tricky,
since now all old paths will fail miserably.
So I added this --base-path-relaxed option, that will make git-daemon
try the absolute path without prefixing --base-path before giving up.
With this in place and --base-path-relaxed added, both my new url of
git://git.kernel.dk/linux-2.6-block.git
and the old
git://git.kernel.dk/data/git/linux-2.6-block.git
work fine.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are (really!) systems where using environment variables is very
cumbersome (yes, Windows, it has problems unsetting them). Besides this
form is shorter.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
make the name of the library directory a config option
Introduce new makefile variable lib to hold the name of the lib
directory ("lib" by default). Also introduce a switch for configure
to specify this name with --with-lib=ARG. This is useful for systems
that use a different name than "lib" (like "lib64" on some 64 bit
Linux architectures).
Signed-off-by: Robert Schiele <rschiele@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some systems do not provide zlib development headers and libraries in
default search path of the compiler. For these systems we should allow
specifying the location by --with-zlib=PATH or by setting ZLIB_PATH in
the makefile.
Signed-off-by: Robert Schiele <rschiele@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
add_file_to_index: skip rehashing if the cached stat already matches
An earlier commit 366bfcb6 broke git-add by moving read_cache()
call down, because it wanted the directory walking code to grab
paths that are already in the index. The change serves its
purpose, but introduces a regression because the responsibility
of avoiding unnecessary reindexing by matching the cached stat
is shifted nowhere.
This makes it the job of add_file_to_index() function.
This makes rebase/am keep the original commit log message
better, even when it does not conform to "single line paragraph
to say what it does, then explain and defend why it is a good
change in later paragraphs" convention.
This change is a two-edged sword. While the earlier behaviour
would make such commit log messages more friendly to readers who
expect to get the birds-eye view with oneline summary formats,
users who primarily use git as a way to interact with foreign
SCM systems would not care much about the convenience of oneline
git log tools, but care more about preserving their own
convention. This changes their commits less useful to readers
who read them with git tools while keeping them more consistent
with the foreign SCM systems they interact with.
symbolic-ref, update-ref: do not refuse reflog message with LF
Earlier these tools refused to create a reflog entry when the
message given by the calling Porcelain had a LF in it, partially
to keep the file format integrity of reflog file, which is
one-entry-per-line. These tools should not be dictating such a
policy.
Instead, let the codepath to write out the reflog entry worry
about the format integrity and allow messages with LF in them.
log_ref_write() -- do not chomp reflog message at the first LF
A reflog file is organized as one-line-per-entry records, and we
enforced the file format integrity by chomping the given message
at the first LF. This changes it to convert them to SP, which
is more in line with the --pretty=oneline format.
* bs/lock:
Add test for symlinked configuration file updates.
use lockfile.c routines in git_commit_set_multivar()
fully resolve symlinks when creating lockfiles
* 'master' of git://repo.or.cz/git-gui: (50 commits)
git-gui: Minor refactoring of merge command line in merge support
git-gui: Use more modern looking icons in the tree browser
git-gui: Don't offer to stage hunks from untracked files
git-gui: Make sure remotes are loaded when picking revisions
git-gui: Use progress bar while resetting/aborting files
git-gui: Honor core.excludesfile when listing extra files
git-gui: Unify wording to say "to stage" instead of "to add"
git-gui: Don't kill modified commit message buffer with merge templates
git-gui: Remove usernames from absolute SSH urls during merging
git-gui: Format tracking branch merges as though they were pulls
git-gui: Cleanup bindings within merge dialog
git-gui: Replace merge dialog with our revision picker widget
git-gui: Show ref last update times in revision chooser tooltips
git-gui: Display commit/tag/remote info in tooltip of revision picker
git-gui: Save remote urls obtained from config/remotes setup
git-gui: Avoid unnecessary symbolic-ref call during checkout
git-gui: Refactor current branch menu items to make i18n easier
git-gui: Refactor diff popup into a procedure to ease i18n work
git-gui: Paper bag fix quitting crash after commit
git-gui: Clarify meaning of add tracked menu option
...
git-gui: Minor refactoring of merge command line in merge support
This is just a small code movement to cleanup how we generate
the command line for a merge. I'm only doing it to make the
next series of changes slightly more readable.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Use more modern looking icons in the tree browser
This is a replacement of all of the icons in our tree browser
window, as the prior icons just looked too 1980s Tk-ish. The
icons used here are actually from a KDE themed look, so they
might actually be familiar to some users of git-gui.
Aside from using more modern looking icons we now have a special
icon for executable blobs, to make them stand out from the normal
non-executable blobs. We also denote symlinks now with a different
icon, so they stand out from the other types of objects in the tree.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Don't offer to stage hunks from untracked files
If the user looks at an untracked file in our diff pane we used
to offer "Stage Hunk For Commit" in the context menu when they
right-clicked in that pane. The problem is we don't actually
have any diff hunks in untracked files, so there is nothing to
really select for staging. So we now grey out the menu item,
so the user cannot invoke it and think its broken when it does
not perform any useful action.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Make sure remotes are loaded when picking revisions
If we are started for only a blame/browser/citool run we don't
usually initialize the list of remotes, or determine which refs
are tracking branches and which are local branch heads. This is
because some of that work is relatively expensive and is usually
not going to be needed if we are started only for a blame, or to
make a single commit.
However by not loading the remote configuration we were crashing
if the user tried to open a browser for another branch through
the Repository menu, as our load_all_heads procedure was unable
to decide which refs/heads/ items were actually local heads. We
now force all remote configuration data to be loaded if we have
not done so already and we are trying to create a revision mega
widget.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The tree recursion behavior of git-diff may appear
inconsistent to the user because it depends on the format of
the patch as well as whether one is diffing between trees or
against the index.
Since git-diff is a porcelain wrapper for low-level diff
commands, it makes sense for its behavior to be consistent
no matter what is being diffed. This patch turns on
recursion in all cases.
git-gui: Use progress bar while resetting/aborting files
Resetting a large number of files on a slow filesystem can take
considerable time, just as switching branches in such a case can
take more than two seconds. We now take advantage of the progress
meter output by read-tree and show it in the main window status
bar, just like we do during checkout (branch switch).
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Honor core.excludesfile when listing extra files
Recent git versions have a git-status that honors the core.excludesfile
configuration option when it reports on untracked files. Unfortunately
I missed the introduction of this configuration option in the core
porcelain implementation, so it was not reflected here in git-gui.
Found and reported by Lars Noschinski <lars@public.noschinski.de>.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
gitweb: Simplify 'opt' parameter validation, add "no merges" feeds
Simplify and make more readable validation of 'opt' (extra options)
parameter, using exists($hash{key}) instead of grepping keys of a hash
for value.
Move 'opt' parameter to be the last (for now) in the URL.
Make use of '--no-merges' extra option ('opt') by adding "no merges"
RSS and Atom feeds to the HTML header. Note that alternate format
links in the RSS and Atom views do not use '--no-merges' option yet!
Adds tests for the 'opt' parameter to t9500-gitweb-standalone-no-errors.sh
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add S_ISGITLINK subroutine and S_IFGITLINK, S_IFINVALID constants.
Add support for "commit" (submodule) entries in the tree object to
mode_str ('m---------', following cgit), file_type and file_type_long
('submodule') subroutines.
There is only link to the history of submodule entry in the
supermodule (current repository) for now, because gitweb doesn't know
where to search for submodule repository objects.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Originally, "apply --index" codepath was bolted on to the
"update working tree files and index, but then revert the
changes we make to the index except for added files so that we
do not forget about them" codepath, almost as an afterthought.
Because "apply --index" first prepares the final index state
upfront, "revert except the added paths" postprocessing does not
have to be done.
Two bugs that made the command practically unusable were fixed
with this.
- A stash created with a clean index does not have any
difference between the base tree and the index tree.
Trying to apply the diff between them to the index would
error out with "No changes". Even when the user asked to
unstash with --index, do not bother with --index action if
the base tree and the index tree match.
- After successfully performing the working tree merge, the
index was reloaded from an earlier state of unstashed index
with "read-tree"; this left all the paths cache dirty. By
moving the call to git-status after this read-tree, match the
cached stat information in the index.
* 'master' of git://git.kernel.org/pub/scm/gitk/gitk:
gitk: Fix bugs in the Find function
gitk: Wait for the window to become visible after creating it
[PATCH] gitk: Bind keyboard actions to the command key on Mac OS
[PATCH] gitk: Ignore ctrl-z as EOF on windows
gitk: Make the fake commit for the index changes green rather than magenta
gitk: Show changes in index and changes in working directory separately
At the moment this just has two entries, which allow you to add the file
that you clicked on to the list of filenames to highlight, or replace
the list with the file.
git-gui: Don't kill modified commit message buffer with merge templates
If the user is in the middle of a merge and has already started to
modify their commit message we were losing the user's changes when
they pressed 'Rescan' after resolving issues or making changes in
the working directory.
The problem here was our background timer that saves the commit
message buffer. It marks the commit message buffer as not being
modified when it writes it out to disk, so during the rescan we
assumed the buffer should be replaced with what we read from the
MERGE_MSG file. So we now only read these files from .git if we
have a valid backup file. Since we clear it on commit this will
only have an impact while the user is actively editing the current
commit.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This replaces "git-verify-tag.sh" with "builtin-verify-tag.c".
Testing relies on the "git tag -v" tests calling this command.
A temporary file is needed when calling to gpg, because git is
already creating detached signatures (gpg option -b) to sign tags
(instead of leaving gpg to add the signature to the file by itself),
and those signatures need to be supplied in a separate file to be
verified by gpg.
The program uses git_mkstemp to create that temporary file needed by
gpg, instead of the previously used "$GIT_DIR/.tmp-vtag", in order to
allow the command to be used in read-only repositories, and also
prevent other instances of git to read or remove the same file.
Signal SIGPIPE is ignored because the program sometimes was
terminated because that signal when writing the input for gpg.
The command now can receive many tag names to be verified.
Documentation is also updated here to reflect this new behaviour.
Signed-off-by: Carlos Rica <jasampler@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This fixes the problem reported by Brian Downing where searching for
a string that doesn't exist would give a Tcl error. The basic problem
was that we weren't reading the data for the last commit since it
wasn't terminated with a null. This effectively adds a null on the end
(if there isn't one already) to make sure we process the last commit.
This also makes the yellow background behind instances of the search
string appear more consistently, and fixes a bug where the "/" key
would just find the same commit again and again instead of advancing.
When squashing, you no longer saw what the editor had to say to you
after commit 'Shut "git rebase -i" up when no --verbose was given'
(if you used a console based editor, at least).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git_mkstemp(): be careful not to overflow the path buffer.
If user's TMPDIR is insanely long, return negative after
setting errno to ENAMETOOLONG, pretending that the underlying
mkstemp() choked on a temporary file path that is too long.
When locking in a symlinked repository, try to lock the original.
In a working tree prepared in new-workdir (in contrib/), some files in .git/
directory are symbolic links to the original repository. The usual sequence of
lock-write-rename would break the symbolic link.
Ideally we should resolve relative symbolic link with maxdepth, but I do not
want to risk too elaborate patch before 1.5.3 release, so this is a minimum
and trivially obvious fix. new-workdir creates its symbolic links absolute,
and does not link from a symlinked workdir, so this fix should suffice for now.
git-submodule: remove redundant call to git-describe
The code to find a more descriptive name given a commit in a
submodule were improved in bffe71f, but it forgot to remove the
older logic the patch replaced.
Signed-off-by: Simon Hausmann <shausman@trolltech.com> Signed-off-by: Marius Storm-Olsen <marius@trolltech.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-submodule module_name: avoid using unwieldy "value_regexp" feature.
"module_name $path" function wants to look up a configuration
variable "submodule.<modulename>.path" whose value is $path, and
return the <modulename> found. "git-config --get-regexp" is the
natural thing to use for this, but (1) its value matching has an
unfortunate "feature" that takes leading '!' specially, and (2)
its output needs to be parsed with sed to extract <modulename>
part anyway.
This changes the call to "git-config --get-regexp" not to use
the value-regexp part, and moves the "pick the one whose value
is $path" part to the downstream sed.
We say "SUBDIRECTORY_OK" but we did not chdir to toplevel; this
is fine as long as everything we use can be started from a
subdirectory, but unfortunately "merge-recursive" is not one of
the programs you can safely use from a subdirectory.
git-gui: Remove usernames from absolute SSH urls during merging
If we are being asked to merge a tracking branch that comes from a
remote repository accessed by the very common SSH URL format of
"user@host:/path/to/repo" then we really don't need the username
as part of the merge message, it only clutters up the history and
makes things more confusing. So we instead clip the username part
off if the local filesystem path is absolute, as its probably not
going to be an ambiguous URL even when it is missing the username.
On the other hand we cannot clip the username off if the URL is
not absolute, because in such cases (e.g. "user@host:myrepo") the
directory that the repository path is resolved in is relative to
the user's home directory, and the username becomes important.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Format tracking branch merges as though they were pulls
If we are merging a tracking branch we know exactly what remote URL
that branch is fetched from, and what its name is on that remote
repository. In this case we can setup a merge message that looks
just like a standard `git-pull $remote $branch` operation by filling
out FETCH_HEAD before we start git-merge, and then run git-merge just
like git-pull does.
I think the result of this behavior is that merges look a lot nicer
when the came off of local tracking branches, because they no longer
say "commit 'origin/...'" to describe the commit being merged but
instead now mention the specific repository we fetched those commits
from.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Replace merge dialog with our revision picker widget
Now that we only support merging one branch we can offer the user
a better user interface experience by allowing them to select the
revision they want to merge through our revision picking widget.
This change neatly solves the problem of locating a branch out of
a sea of 200 tracking branches, and of dealing with very long branch
names that all have a common prefix.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Show ref last update times in revision chooser tooltips
If we can we now show the last modification date of a loose ref as
part of the tooltip information shown in the revision picker. This
gives the user an indication of when was the last time that the ref
was modified locally, and may especially be of interest when looking
at a tracking branch.
If we cannot find the loose ref file than we try to fallback on the
reflog and scan it for the date of the last record. We don't start
with the reflog however as scanning it backwards from the end is not
an easy thing to do in Tcl. So I'm being lazy here and just going
through the entire file, line by line. Since that is less efficient
than a single stat system call, its our fallback strategy.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Display commit/tag/remote info in tooltip of revision picker
Our revision chooser mega-widget now sets up tooltips for itself so
that it displays details about a commit (or a tag and the commit
it refers to) when the user mouses over that line in the filtered
ref list. If the item is from a remote tracking branch then we also
show the remote url and what branch on that remote we fetch from, so
the user has a clear concept of where that revision data originated.
To help the merge dialog I've also added a new constructor that
makes the dialog only offer unmerged revisions (those not in HEAD),
as this allows users to avoid performing merges only to get "Already
up to date" messages back from core Git.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
git-gui: Save remote urls obtained from config/remotes setup
I'm storing the URLs of any pre-configured remote repositories
that we happen to come across so that we can later use these
URLs to show to the user in parts of the UI that might care.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
These are useful in organizations that enforce particular formats
for commit messages, e.g., to specify bug IDs or test plans.
Use of the template is not enforced; it is simply used as the
initial content when the editor is invoked.
Signed-off-by: Steven Grimm <koreth@midwinter.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If git cvsexportcommit is executed fast enough in sequence, the CVS
timestamps could end up being the same. CVS tries to fix this
by sleeping until the CPU clock changes seconds. Unfortunately,
the CPU clock and the file system clock are not necessarily the same, so
the timestamps could be the same anyway. When that happens CVS may not
recognize changed files and cvs will forget to commit some files.
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Remove the two write-only fields executable and symlink from struct
tree_entry_list. Also replace usage of the field directory with
S_ISDIR checks on the mode field, and then remove this now obsolete
field, too. Noticed by David Kastrup.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>