HPA noticed that git-rebase fails when changes involve symlinks
in the middle of the hierarchy. Consider:
* The tree state before the patch is applied has arch/x86_64/boot
as a symlink pointing at ../i386/boot/
* The patch tries to remove arch/x86_64/boot symlink, and
create bunch of files there: .gitignore, Makefile, etc.
git-apply tries to be careful while applying patches; it never
touches the working tree until it is convinced that the patch
would apply cleanly. One of the check it does is that when it
knows a path is going to be created by the patch, it runs
lstat() on the path to make sure it does not exist.
This leads to a false alarm. Because we do not touch the
working tree before all the check passes, when we try to make
sure that arch/x86_64/boot/.gitignore does not exist yet, we
haven't removed the arch/x86_64/boot symlink. The lstat() check
ends up seeing arch/i386/boot/.gitignore through the
yet-to-be-removed symlink, and says "Hey, you already have a
file there, but what you fed me is a patch to create a new
file. I am not going to clobber what you have in the working
tree."
We have similar checks to see a file we are going to modify does
exist and match the preimage of the diff, which is done by
directly opening and reading the file.
For a file we are going to delete, we make sure that it does
exist and matches what is going to be removed (a removal patch
records the full preimage, so we check what you have in your
working tree matches it in full -- otherwise we would risk
losing your local changes), which again is done by directly
opening and reading the file.
These checks need to be adjusted so that they are not fooled by
symlinks in the middle.
- To make sure something does not exist, first lstat(). If it
does not exist, it does not, so be happy. If it _does_, we
might be getting fooled by a symlink in the middle, so break
leading paths and see if there are symlinks involved. When
we are checking for a path a/b/c/d, if any of a, a/b, a/b/c
is a symlink, then a/b/c/d does _NOT_ exist, for the purpose
of our test.
This would fix this particular case you saw, and would not
add extra overhead in the usual case.
- To make sure something already exists, first lstat(). If it
does not exist, barf (up to this, we already do). Even if it
does seem to exist, we might be getting fooled by a symlink
in the middle, so make sure leading paths are not symlinks.
This would make the normal codepath much more expensive for
deep trees, which is a bit worrisome.
This patch implements the first side of the check "making sure
it does not exist". The latter "making sure it exists" check is
not done yet, so applying the patch in reverse would still
fail, but we have to start from somewhere.
Signed-off-by: Junio C Hamano <junkio@cox.net>
When we are applying a patch that creates a blob at a path, or
when we are switching from a branch that does not have a blob at
the path to another branch that has one, we need to make sure
that there is nothing at the path in the working tree, as such a
file is a local modification made by the user that would be lost
by the operation.
Normally, lstat() on the path and making sure ENOENT is returned
is good enough for that purpose. However there is a twist. We
may be creating a regular file arch/x86_64/boot/Makefile, while
removing an existing symbolic link at arch/x86_64/boot that
points at existing ../i386/boot directory that has Makefile in
it. We always first check without touching filesystem and then
perform the actual operation, so when we verify the new file,
arch/x86_64/boot/Makefile, does not exist, we haven't removed
the symbolic link arc/x86_64/boot symbolic link yet. lstat() on
the file sees through the symbolic link and reports the file is
there, which is not what we want.
The function has_symlink_leading_path() function takes a path,
and sees if any of the leading directory component is a symbolic
link.
When files in a new directory are created, we tend to process
them together because both index and tree are sorted. The
function takes advantage of this and allows the caller to cache
and reuse which symbolic link on the filesystem caused the
function to return true.
The calling sequence would be:
char last_symlink[PATH_MAX];
*last_symlink = '\0';
for each index entry {
if (!lose)
continue;
if (lstat(it))
if (errno == ENOENT)
; /* happy */
else
error;
else if (has_symlink_leading_path(it, last_symlink))
; /* happy */
else
error; /* would lose local changes */
unlink_entry(it, last_symlink);
}
Signed-off-by: Junio C Hamano <junkio@cox.net>
The rebase and merge commands used diff-tree to display the summary stats of
what files had changed from the operation. diff-tree does not read the
diff ui configuration options, so the diff.color setting was not used.
Have rebase and merge call diff rather than diff-tree, which does read the
diff ui options.
Signed-off-by: James Bowes <jbowes@dangerouslyinc.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
git-checkout is also adapted to make use of this new option
instead of the handcrafted command sequence.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This option causes 'git gc' to more aggressively optimize the
repository at the cost of taking much more wall clock and CPU time.
Today this option causes git-pack-objects to use --no-use-delta
option, and it allows the --window parameter to be set via the
gc.aggressiveWindow configuration parameter.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Add config variables pack.compression and core.loosecompression ,
and switch --compression=level to pack-objects.
Loose objects will be compressed using core.loosecompression if set,
else core.compression if set, else Z_BEST_SPEED.
Packed objects will be compressed using --compression=level if seen,
else pack.compression if set, else core.compression if set,
else Z_DEFAULT_COMPRESSION. This is the "pack compression level".
Loose objects added to a pack undeltified will be recompressed
to the pack compression level if it is unequal to the current
loose compression level by the preceding rules, or if the loose
object was written while core.legacyheaders = true. Newly
deltified loose objects are always compressed to the current
pack compression level.
Previously packed objects added to a pack are recompressed
to the current pack compression level exactly when their
deltification status changes, since the previous pack data
cannot be reused.
In either case, the --no-reuse-object switch from the first
patch below will always force recompression to the current pack
compression level, instead of assuming the pack compression level
hasn't changed and pack data can be reused when possible.
This applies on top of the following patches from Nicolas Pitre:
[PATCH] allow for undeltified objects not to be reused
[PATCH] make "repack -f" imply "pack-objects --no-reuse-object"
Signed-off-by: Dana L. How <danahow@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Now that we encourage and actively preserve objects in a packed form
more agressively than we did at the time the new loose object format and
core.legacyheaders were introduced, that extra loose object format
doesn't appear to be worth it anymore.
Because the packing of loose objects has to go through the delta match
loop anyway, and since most of them should end up being deltified in
most cases, there is really little advantage to have this parallel loose
object format as the CPU savings it might provide is rather lost in the
noise in the end.
This patch gets rid of core.legacyheaders, preserve the legacy format as
the only writable loose object format and deprecate the other one to
keep things simpler.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Recomputing delta is much more expensive than recompressing
anyway, and when the user says 'repack -f', it is a sign that
the user is willing to spend CPU cycles.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Currently non deltified object data is always reused when possible.
This means that any change to core.compression has no effect on those
objects as they don't get recompressed when repacking them.
Let's add a --no-reuse-object flag to git-repack in order to force
recompression of all objects when desired.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
We now write to MERGE_MSG, not .msg. I missed this earlier
when I changed the target we write to.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
* 'master' of git://repo.or.cz/git-gui:
git gui 0.7.0
git-gui: Paperbag fix blame in subdirectory
git-gui: Format author/committer times in ISO format
git-gui: Cleanup minor nits in blame code
git-gui: Generate blame on uncommitted working tree file
git-gui: Smarter command line parsing for browser, blame
git-gui: Use prefix if blame is run in a subdirectory
git-gui: Convert blame to the "class" way of doing things
git-gui: Don't attempt to inline array reads in methods
git-gui: Convert browser, console to "class" format
git-gui: Define a simple class/method system
git-gui: Allow shift-{k,j} to select a range of branches to merge
git-gui: Call changes "Staged" and "Unstaged" in file list titles.
Avoid specifying font sizes in pixels, since that is just pure evil.
Pointed out by Chris Riddoch.
Note that this is pretty much just a proposal; I didn't test if everything
fits perfectly right, but things seem to be pretty much okay. repo.or.cz
uses it now as a test drive - if you find any visual quirks, please point
them out, with a patch if possible since I'm total CSS noob and debugging
CSS is an extremely painful experience for me.
Note that this patch actually does change visual look of gitweb in Firefox
with my resolution and default settings - everything is bigger and I can't
explain the joy of actually seeing gitweb text that is in _readable_ size;
also, my horizontal screen real estate feels better used now. But judging
from the look of most modern webpages on the 'net, most people prefer
reading the web with strained eyes and/or a magnifying glass (I wonder what
species of scientists should look into this mystifying phenomenon) - so,
please tell us what you think.
Maybe we might want to get rid of absolute sizes other than font sizes in
the CSS file too in the long term.
Signed-off-by: Petr Baudis <pasky@suse.cz>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This patch renames config_boolean() to config_bool() for consistency with
the commandline interface and because it is shorter but still obvious. ;-)
It also changes the return value from some obscure string to real Perl
boolean, allowing for clean user code.
Signed-off-by: Petr Baudis <pasky@suse.cz>
* maint:
.mailmap: add some aliases
SPECIFYING RANGES typo fix: it it => it is
git-clone: don't get fooled by $PWD
Fix documentation of tag in git-fast-import.txt
If you have /home/me/git symlink pointing at /pub/git/mine,
trying to clone from /pub/git/his/ using relative path would not
work as expected:
$ cd /home/me
$ cd git
$ ls ../
his mine
$ git clone -l -s -n ../his/stuff.git
This is because "cd ../his/stuff.git" done inside git-clone to
check if the repository is local is confused by $PWD, which is
set to /home/me, and tries to go to /home/his/stuff.git which is
different from /pub/git/his/stuff.git.
We could probably say "set -P" (or "cd -P") instead, if we know
the shell is POSIX, but the way the patch is coded is probably
more portable.
[jc: this is updated with Andy Whitcroft's improvements]
Signed-off-by: Junio C Hamano <junkio@cox.net>
gitweb URLs use the a= parameter for the view to use on the given path, such
as "blob" or "tree". Currently, if a gitweb URL omits the a= parameter,
gitweb just shows the top-level repository summary, regardless of the path
given. gitweb could instead choose an appropriate view based on the file
type: blob for blobs (files), tree for trees (directories), and summary if
no path given (the URL included no f= parameter, or an empty f= parameter).
Apart from making gitweb more robust and supporting URL editing more easily,
this change would aid the creation of shortcuts to git repositories using
simple substitution, such as:
http://example.org/git/?p=path/to/repo.git;hb=HEAD;f=%s
With this patch, if given the hash through the h= parameter, or the hash
base (hb=) and a filename (f=), gitweb uses cat-file -t to automatically set
the a= parameter.
This feature was requested by Josh Triplett through
http://bugs.debian.org/410465
Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
* jc/diffopt:
diff -S: release the image after looking for needle in it
diff -M: release the preimage candidate blobs after rename detection.
diff.c: do not use a separate "size cache".
diff: release blobs after generating textual diff.
* jn/gitweb:
gitweb: Show combined diff for merge commits in 'commit' view
gitweb: Show combined diff for merge commits in 'commitdiff' view
gitweb: Make it possible to use pre-parsed info in git_difftree_body
gitweb: Add combined diff support to git_patchset_body
gitweb: Add combined diff support to git_difftree_body
gitweb: Add parsing of raw combined diff format to parse_difftree_raw_line
Use the :fork: access method to force cvs to
call "$CVS_SERVER server" even when accessing a local
repository.
Add a basic test for checkout and some tests for update.
Signed-off-by: Frank Lichtenheld <frank@lichtenheld.de>
Signed-off-by: Junio C Hamano <junkio@cox.net>
Rather than storing the temporary commit message data in .msg (in
the working tree) we now store the message data in .git/MERGE_MSG.
By storing the message in the .git/ directory we are sure we will
never have a collision with a user file, should a project actually
have a ".msg" file in their top level tree. We also don't need to
worry about leaving this stale file behind during a `reset --hard`
and have it show up in the output of status.
We are using .git/MERGE_MSG here to store the temporary message as
it is an already established convention between git-merge, git-am
and git-rebase that git-commit will default the user's edit buffer
to the contents of .git/MERGE_MSG. If the user is going to need
to resolve this commit or wants to edit the message on their own
prepping that file with the desired message "just works".
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
RPM packages did not include howto files which causes broken
links in howto-index.html
Signed-off-by: Quy Tonthat <qtonthat@gmail.com>
Signed-off-by: Junio C Hamano <junkio@cox.net>
This is a simple change to match what gitk does when it shows
a commit; we format using ISO dates (yyyy-mm-dd HH:MM:SS).
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
We can use [list ...] rather than "", especially when we are talking
about values as then they are properly escaped if necessary. Small
nit, but probably not a huge deal as the only data being inlined here
is Tk paths.
Some of the lines in the parser code were longer than 80 characters
wide, and they actually were all the same value on the end part of
the line. Rather than keeping the mess copied-and-pasted around we
can set the last argument into a local variable and reuse it many
times.
The commit display code was also rather difficult to read on an 80
character wide terminal, so I'm moving it all into a double quoted
string that is easier to read.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If the user doesn't give us a revision parameter to our blame
subcommand then we can generate blame against the working tree
file by passing the file path off to blame with the --contents
argument. In this case we cannot obtain the contents of the
file from the ODB; instead we must obtain the contents by
reading the working directory file as-is.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The browser subcommand now optionally accepts a single revision
argument; if no revision argument is supplied then we use the
current branch as the tree to browse. This is very common, so
its a nice option.
Our blame subcommand now tries to perform the same assumptions
as the command line git-blame; both the revision and the file
are optional. We assume the argument is a filename if the file
exists in the working directory, otherwise we assume the argument
is a revision name. A -- can be supplied between the two to force
parsing, or before the filename to force it to be a filename.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
I think it was Andy Parkins who pointed out that git gui blame HEAD f
does not work if f is in a subdirectory and we are currently running
git-gui within that subdirectory. This is happening because we did
not take the user's prefix into account when we computed the file
path in the repository.
We now assume the prefix as returned by rev-parse --show-prefix is
valid and we use that during the command line blame subcommand when
we apply the parameters.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Our blame viewer code has historically been a mess simply
because the data for multiple viewers was all crammed into
a single pair of Tcl arrays. This made the code hard to
read and even harder to maintain.
Now that we have a slightly better way of tracking the data
for our "meta-widgets" we can make use of it here in the
blame viewer to cleanup the code and make it easier to work
with long term.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
If a variable reference to a field is to an array, and it is
the only reference to that field in that method we cannot make
it an inlined [set foo] call as the regexp was converting the
Tcl code wrong. We were producing "[set foo](x)" for "$foo(x)",
and that isn't valid Tcl when foo is an array. So we just punt
if the only occurance has a ( after it.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Now that we have a slightly easier method of working with per-widget
data we should make use of that technique in our browser and console
meta-widgets, as both have a decent amount of information that they
store on a per-widget basis and our current approach of handling
it is difficult to follow.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
As most of the git-gui interface is based upon "meta-widgets"
that need to carry around a good deal of state (e.g. console
windows, browser windows, blame viewer) we have a good deal
of messy code that tries to store this meta-widget state in
global arrays, where keys into the array are formed from a
union of a unique "object instance id" and the field name.
This is a simple class system for Tcl that allows us to
hide much of that mess by making Tcl do what it does best;
process strings to manipulate its own code during startup.
Each object instance is placed into its own namespace. The
namespace is created when the object instance is created and
the namespace is destroyed when the object instance is removed
from the system. Within that namespace we place variables for
each field within the class; these variables can themselves be
scalar values or full-blown Tcl arrays.
A simple class might be defined as:
class map {
field data
field size 0
constructor {} {
return $this
}
method set {name value} {
set data($name) $value
incr size
}
method size {} {
return $size
} ifdeleted { return 0 }
}
All fields must be declared before any constructors or methods. This
allows our class to generate a list of the fields so it can properly
alter the definition of the constructor and method bodies prior to
passing them off to Tcl for definition with proc. A field may optionally
be given a default/initial value. This can only be done for non-array
type fields.
Constructors are given full access to all fields of the class, so they
can initialize the data values. The default values of fields (if any)
are set before the constructor runs, and the implicit local variable
$this is initialized to the instance identifier.
Methods are given access to fields they actually use in their body.
Every method has an implicit "this" argument inserted as its first
parameter; callers of methods must be sure they supply this value.
Some basic optimization tricks are performed (but not much). We
try to only upvar (locally bind) fields that are accessed within a
method, but we err on the side of caution and may upvar more than
we need to. If a variable is accessed only once within a method
and that access is by $foo (read) we avoid the upvar and instead
use [set foo] to obtain the value. This is slightly faster as Tcl
does not need to lookup the variable twice.
We also offer some small syntatic sugar for interacting with Tk and
the fileevent callback system in Tcl. If a field (say "foo") is used
as "@foo" we insert instead the true global variable name of that
variable into the body of the constructor or method. This allows easy
binding to Tk textvariable options, e.g.:
label $w.title -textvariable @title
Proper namespace callbacks can also be setup with the special cb proc
that is defined in each namespace. [cb _foo a] will invoke the method
_foo in the current namespace, passing it $this as the first (implied)
parameter and a as the second parameter. This makes it very simple to
connect an object instance to a -command option for a Tk widget or to
a fileevent readable or writable for a file channel.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
