I have reviewed all occurrences of mmap() in git and fixed three types
of errors/defects:
1) The result is not checked.
2) The file descriptor is closed if mmap() succeeds, but not when it
fails.
3) Various casts applied to -1 are used instead of MAP_FAILED, which is
specifically defined to check mmap() return value.
[jc: This is a second round of Pavel's patch. He fixed up the problem
that close() potentially clobbering the errno from mmap, which
the first round had.]
Signed-off-by: Pavel Roskin <proski@gnu.org>
Signed-off-by: Junio C Hamano <junkio@cox.net>
... and make git-diff-files use it too. This all _should_ make the
diffcore-pathspec.c phase unnecessary, since the diff'ers now all do the
path matching early interally.
An earlier change to optimize directory-file conflict check
broke what "read-tree --emu23" expects. This is fixed by this
commit.
(1) Introduces an explicit flag to tell add_cache_entry() not to
check for conflicts and use it when reading an existing tree
into an empty stage --- by definition this case can never
introduce such conflicts.
(2) Makes read-cache.c:has_file_name() and read-cache.c:has_dir_name()
aware of the cache stages, and flag conflict only with paths
in the same stage.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This is (imho) more readable, and is also a lot faster. The expense of
looking up sub-directory beginnings was killing us on things like
"git-diff-cache", even though that one didn't even care at all about the
file vs directory conflicts.
We really only care when somebody tries to add a conflicting name to
stage 0.
We should go through the conflict rules more carefully some day.
When we choose to omit deleted entries, we should subtract
numbers of such entries from the total number in the header.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Oops.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
check_file_directory_conflict can give the wrong answers. This is
because the wrong length is passed to cache_name_pos. The length
passed should be the length of the whole path from the root, not
the length of each path subcomponent.
$ git-init-db
defaulting to local storage area
$ mkdir path && touch path/file
$ git-update-cache --add path/file
$ rm path/file
$ mkdir path/file && touch path/file/f
$ git-update-cache --add path/file/f <-- Conflict ignored
$
Signed-off-by: David Meybohm <dmeybohmlkml@bellsouth.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Thomas Glanzmann points out that it doesn't work well with different
clients accessing the repository over NFS - they have different views
on what the "device" for the filesystem is.
Of course, other filesystems may not even have stable inode numbers.
But we don't care. At least for now.
Add <limits.h> to the include files handled by "cache.h", and remove
extraneous #include directives from various .c files. The rule is that
"cache.h" gets all the basic stuff, so that we'll have as few system
dependencies as possible.
This one compares two pathnames that may be partial basenames, not
full paths. We need to get the path sorting right, since a directory
name will sort as if it had the final '/' at the end.
With -u flag, git-checkout-cache picks up the stat information
from newly created file and updates the cache. This removes the
need to run git-update-cache --refresh immediately after running
git-checkout-cache.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- Raw hashes should be unsigned char.
- String functions want signed char.
- Hash and compress functions want unsigned char.
Signed-off By: Brian Gerst <bgerst@didntduck.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When "path" exists as a file or a symlink in the index, an
attempt to add "path/file" is refused because it results in file
vs directory conflict. Similarly when "path/file1",
"path/file2", etc. exist, an attempt to add "path" as a file or
a symlink is refused. With git-update-cache --replace, these
existing entries that conflict with the entry being added are
automatically removed from the cache, with warning messages.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Some commands initialize sha1_file_directory by hand. There is no
need to do so; sha1_file.c knows how to handle it.
The next patch will remove the variable altogether.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
It didn't properly mark all cache updates as being dirty, and
causes merge errors due to that. In particular, it didn't notice
when a file was force-removed.
Besides, it was ugly as hell. I've put in place a slightly cleaner
version, but I've not enabled the optimization because I don't
want to be burned again.
Fix update-cache to compare the blob of a symlink against the link-target
and not the file it points to. Also ignore all permissions applied to
links.
Thanks to Greg for recognizing this while he added our list of symlinks
back to the udev repository.
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Allow to store and track symlink in the repository. A symlink is stored
the same way as a regular file, only with the appropriate mode bits set.
The symlink target is therefore stored in a blob object.
This will hopefully make our udev repository fully functional. :)
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Introduce xmalloc and xrealloc to die gracefully with a descriptive
message when out of memory, rather than taking a SIGSEGV.
Signed-off-by: Christopher Li<chrislgit@chrisli.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This one is about a million times simpler, and much more likely to be
correct too.
Instead of trying to match up a tree object against the index, we just
read in the tree object side-by-side into the index, and just walk the
resulting index file. This was what all the read-tree cleanups were
all getting to.
We use that to specify alternative index files, which can be useful
if you want to (for example) generate a temporary index file to do
some specific operation that you don't want to mess with your main
one with.
It defaults to the regular ".git/index" if it hasn't been specified.
I noticed this when I tried a non-trivial scsi merge and checked the
results against BK. The problem is that remove_entry_at() actually
decrements active_nr, so decrementing it in add_cache_entry() before
calling remove_entry_at() is a double decrement (hence we lose cache
entries at the end).
This fixes show-diff listing all +x files as differring.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
[ That's what I get for working on a G5 - my testing was all
big-endian in the first place. -- Linus ]
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When update-cache --remove is run, resolve unmerged state for
the path. This is consistent with the update-cache --add
behaviour. Essentially, the user is telling us how he wants to
resolve the merge by running update-cache.
Signed-off-by: Junio C Hamano <junkio@cox.net>
Fixed to do the right thing at the end.
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This allows using a git tree over NFS with different byte order, and
makes it possible to just copy a fully populated repository and have
the end result immediately usable (needing just a refresh to update
the stat information).
Now there is error() for "library" errors and die() for fatal "application"
errors. usage() is now used strictly only for usage errors.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
The nsec field of ctime/mtime is now checked only with -DNSEC defined during
compilation. nsec acts broken since it is stored in the icache but apparently
just gets to zero when flushed to filesystem not supporting it (e.g. ext3),
creating illusions of false changes. At least that's my impression.
Signed-off-by: Petr Baudis <pasky@ucw.cz>
When compiled with -DCOLLISION_CHECK, we will check against SHA1
collisions when writing to the object database.
From: Christopher Li <chrislgit@chrisli.org>
Signed-off-by: Petr Baudis <pasky@ucw.cz>
Also make the return value of "cache_name_pos()" be sane: positive
or zero if we found it (it's the index into the cache array), and
"-pos-1" to indicate where it should go if we didn't.
It now requires the "--add" flag before you add any new files, and
a "--remove" file if you want to mark files for removal. And giving
it the "--refresh" flag makes it just update all the files that it
already knows about.
It's got some debugging printouts etc still in it, but testing on the
kernel seems to show that it does indeed fix the issue with huge tree
files for each commit.
Pavel Roskin