git help -a scans the PATH for git commands. On Windows it failed for two
reasons:
- The PATH separator is ';', not ':' on Windows.
- stat() does not set the executable bit.
We now open the file and guess whether it is executable.
The result of the guess is good enough for the list of git commands, but
it is of no use for a general stat() implementation because (1) it is a
guess, (2) the user has no way to influence the outcome (via chmod or
similar), and (3) it would reduce stat() performance by an unacceptable
amount. Therefore, this strategy is a special-case local to help.c.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
On Windows, write() is implemented using WriteFile(). After the reader
closed its end of the pipe, the first WriteFile() returns
ERROR_BROKEN_PIPE (which translates to EPIPE), subsequent WriteFile()s
return ERROR_NO_DATA, which is translated to EINVAL.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Since we have neither fork() nor exec(), we have to spawn the pager and
feed it with the program's output.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Since the Makefile in the template/ subdirectory is only used to install
the templates, we do not simply pass down the setting of template_dir
when it is relative, but construct the intended destination in a new
variable: A relative template_dir is relative to gitexecdir.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
With this definition the templates and system config file will be found
irrespective of the installation location.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Since on Windows the user is fairly free where to install programs, we
cannot rely on a hard-coded path. We use the program name to derive the
installation directory and use that as exec_path.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
builtin_exec_path returns the hard-coded installation path, which is used
as the ultimate fallback to look for git commands. Making it into a function
enables us in a follow-up patch to return a computed value instead of just
a constant string.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Windows's struct stat does not have a st_blocks member. Since we already
have our own stat/lstat/fstat implementations, we can just as well use
a customized struct stat. This patch introduces just that, and also fills
in the st_blocks member. On the other hand, we don't provide members that
are never used.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
This is a necessary pendant to our lstat implementation: MSVCRT's
implementations of lstat and utime do some adjustments if daylight
saving time is in effect, but our lstat implementation doesn't do these
adjustments and report the correct UTC time. With this implementation
we omit the adjustments in utime() as well and always write UTC.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
This gives us a significant speedup when adding, committing and stat'ing files.
Also, since Windows doesn't really handle symlinks, we let stat just uses lstat.
We also need to replace fstat, since our implementation and the standard stat()
functions report slightly different timestamps, possibly due to timezones.
We simply report UTC in our implementation, and do our FILETIME to time_t
conversion based on the document at http://support.microsoft.com/kb/167296.
With Moe's repo structure (100K files in 100 dirs, containing 2-4 bytes)
mkdir bummer && cd bummer; for ((i=0;i<100;i++)); do
mkdir $i && pushd $i;
for ((j=0;j<1000;j++)); do echo "$j" >$j; done;
popd;
done
We get the following performance boost:
With normal lstat & stat Custom lstat/fstat
------------------------ ------------------------
Command: git init Command: git init
------------------------ ------------------------
real 0m 0.047s real 0m 0.063s
user 0m 0.031s user 0m 0.015s
sys 0m 0.000s sys 0m 0.015s
------------------------ ------------------------
Command: git add . Command: git add .
------------------------ ------------------------
real 0m19.390s real 0m12.031s 1.6x
user 0m 0.015s user 0m 0.031s
sys 0m 0.030s sys 0m 0.000s
------------------------ ------------------------
Command: git commit -a.. Command: git commit -a..
------------------------ ------------------------
real 0m30.812s real 0m16.875s 1.8x
user 0m 0.015s user 0m 0.015s
sys 0m 0.000s sys 0m 0.015s
------------------------ ------------------------
3x Command: git-status 3x Command: git-status
------------------------ ------------------------
real 0m11.860s real 0m 5.266s 2.2x
user 0m 0.015s user 0m 0.015s
sys 0m 0.015s sys 0m 0.015s
real 0m11.703s real 0m 5.234s
user 0m 0.015s user 0m 0.015s
sys 0m 0.000s sys 0m 0.000s
real 0m11.672s real 0m 5.250s
user 0m 0.031s user 0m 0.015s
sys 0m 0.000s sys 0m 0.000s
------------------------ ------------------------
Command: git commit... Command: git commit...
(single file) (single file)
------------------------ ------------------------
real 0m14.234s real 0m 7.735s 1.8x
user 0m 0.015s user 0m 0.031s
sys 0m 0.000s sys 0m 0.000s
Signed-off-by: Marius Storm-Olsen <mstormo_git@storm-olsen.com>
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
The problem with Windows's own implementation is that it tries to be
clever when a console program is invoked from a GUI application: In this
case it sometimes automatically allocates a new console window. As a
consequence, the IO channels of the spawned program are directed to the
console, but the invoking application listens on channels that are now
directed to nowhere.
In this implementation we use the lowlevel facilities of CreateProcess(),
which offers a flag to tell the system not to open a console. As a side
effect, only stdin, stdout, and stderr channels will be accessible from
C programs that are spawned. Other channels (file handles, pipe handles,
etc.) are still inherited by the spawned program, but it doesn't get
enough information to access them.
Johannes Schindelin integrated path quoting and unified the various
*execv* and *spawnv* helpers. Eric Raible suggested to also quote '{'.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
gethostbyname() is the first function that calls into the Winsock library,
and it is wrapped only to initialize the library.
socket() is wrapped for two reasons:
- Windows's socket() creates things that are like low-level file handles,
and they must be converted into file descriptors first.
- And these handles cannot be used with plain ReadFile()/WriteFile()
because they are opened for "overlapped IO". We have to use WSASocket()
to create non-overlapped IO sockets.
connect() must be wrapped because Windows's connect() expects the low-level
sockets, not file descriptors, and we must first unwrap the file descriptor
before we can pass it on to Windows's connect().
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
If the PATH lists the Windows system directories before the MSYS
directories, Windows's own incompatible sort and find commands would be
picked up. We implement these commands as functions and call the real
tools by absolute path.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
In upload-pack we must explicitly close the output channel of rev-list.
(On Unix, the channel is closed automatically because process that runs
rev-list terminates.)
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
If on Windows a path is specified as C:/path, then this is also a valid
SSH URL. To disambiguate between the two interpretations we take an URL
that looks like a path with a drive letter as a local URL.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
This emulation of poll() is by far not general. It assumes that the
fds that are to be waited for are connected to pipes. The pipes are
polled in a loop until data becomes available in at least one of them.
If only a single fd is waited for, the implementation actually does
not wait at all, but assumes that a subsequent read() will block.
In order not to needlessly burn CPU time, the CPU is yielded to other
processes before the next round in the poll loop using Sleep(0). Note that
any sleep timeout greater than zero will reduce the efficiency by a
magnitude.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Sometimes it is desirable to have non-fast-forward branches in a
shared repository. A typical example of that is the 'pu' branch.
This patch extends the format of allowed-users and allow-groups
files by using the '+' sign at the beginning as the mark that
non-fast-forward pushes are permitted to the branch.
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do this by handing over the Net::SMTP instance to Net::SMTP::SSL,
which avoids Net::SMTP::TLS and its weird error checking. This trick
is due to Brian Evins.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a fifo is given, validation must be skipped because we can't
read the fifo twice. Ideally git-send-email would cache the read
data instead of attempting to read twice, but for now just skip
validation.
Signed-off-by: Kevin Ballard <kevin@sb.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As announced for 1.6.0.
Git older than version 1.5.2 (or any other git version with this option
set to 1) may revert to version 1 of the pack index by manually deleting
all .idx files and recreating them using 'git index-pack'. Communication
over the git native protocol is unaffected since the pack index is never
transferred.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
GIT 1.5.6.1
fix update-hook-example to work with packed tag references
clone: create intermediate directories of destination repo
for-each-ref: implement missing tag values
git-rebase.sh: Add check if rebase is in progress
Resetting a selected set of index entries is done with
"git reset -- paths" syntax, but we did not allow -- to be omitted
even when the command is unambiguous.
This updates the command to follow the general rule:
* When -- appears, revs come before it, and paths come after it;
* When there is no --, earlier ones are revs and the rest are paths, and
we need to guess. When lack of -- marker forces us to guess, we
protect from user errors and typoes by making sure what we treat as
revs do not appear as filenames in the work tree, and what we treat as
paths do appear as filenames in the work tree, and by erroring out if
that is not the case. We tell the user to disambiguate by using -- in
such a case.
which is employed elsewhere in the system.
When this rule is applied to "reset", because we can have only zero or one
rev to the command, the check can be slightly simpler than other programs.
We have to check only the first one or two tokens after the command name
and options, and when they are:
-- A:
no explicit rev given; "A" and whatever follows it are paths.
A --:
explicit rev "A" given and whatever follows the "--" are paths.
A B:
"A" could be rev or path and we need to guess. "B" could
be missing but if exists that (and everything that follows) would
be paths.
So we apply the guess only in the last case and only to "A" (not "B" and
what comes after it).
* As long as "A" is unambiguously a path, index entries for "A", "B" (and
everything that follows) are reset to the HEAD revision.
* If "A" is unambiguously a rev, on the other hand, the index entries for
"B" (and everything that follows) are reset to the "A" revision.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The perldiag(1) has following to say about this:
"Can't do inplace edit without backup"
(F) You're on a system such as MS-DOS that gets confused if
you try reading from a deleted (but still opened) file. You
have to say -i.bak, or some such.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Johannes Sixt noticed that if the last file in the list was staged, my
earlier patch would display the diff for the penultimate file, but show
the file _before_ that as being selected.
This was due to my misunderstanding the lno argument to show_diff.
This patch fixes the problem: lno is not decremented in the special case
to handle the last item in the list (though we still need to use $lno-1
to find the right path for the next diff).
Signed-off-by: Abhijit Menon-Sen <ams@toroid.org>
Tested-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
* lt/config-fsync:
Add config option to enable 'fsync()' of object files
Split up default "i18n" and "branch" config parsing into helper routines
Split up default "user" config parsing into helper routine
Split up default "core" config parsing into helper routine
* sr/tests:
Hook up the result aggregation in the test makefile.
A simple script to parse the results from the testcases
Modify test-lib.sh to output stats to t/test-results/*
Conflicts:
t/test-lib.sh
* jh/clone-packed-refs:
Teach "git clone" to pack refs
Prepare testsuite for a "git clone" that packs refs
Move pack_refs() and friends into libgit
Incorporate fetched packs in future object traversal
The update-hook-example used 'test -f' to check the tag present, which
does not work if the checked reference is packed. This check has been
changed to use 'git rev-parse $tag' instead.
Signed-off-by: Dmitry Potapov <dpotapov@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The shell version used to use "mkdir -p" to create the repo
path, but the C version just calls "mkdir". Let's replicate
the old behavior. We have to create the git and worktree
leading dirs separately; while most of the time, the
worktree dir contains the git dir (as .git), the user can
override this using GIT_WORK_TREE.
We can reuse safe_create_leading_directories, but we need to
make a copy of our const buffer to do so. Since
merge-recursive uses the same pattern, we can factor this
out into a global function. This has two other cleanup
advantages for merge-recursive:
1. mkdir_p wasn't a very good name. "mkdir -p foo/bar" actually
creates bar, but this function just creates the leading
directories.
2. mkdir_p took a mode argument, but it was completely
ignored.
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "type" and "object" fields for tags were accepted as
valid atoms, but never implemented. Consequently, they
simply returned the empty string, even for valid tags.
Noticed by Lea Wiemann.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As announced for 1.6.0.
Access over the native protocol by old git versions is unaffected as
this capability is negociated by the protocol. Otherwise setting this
config option to "false" and doing a 'git repack -a -d' is enough to
remain compatible with ancient git versions (older than 1.4.4).
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To do so, check_pack_crc() moved from builtin-pack-objects.c to
pack-check.c where it is more logical to share.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is called when verify_pack() has its verbose argument set, and
verbose in this context makes sense only for the actual 'git verify-pack'
command. Therefore let's move show_pack_info() to builtin-verify-pack.c
instead and remove useless verbose argument from verify_pack().
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Using find_pack_entry_one() to get object offsets is rather suboptimal
when nth_packed_object_offset() can be used directly.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The shell version used to use "mkdir -p" to create the repo
path, but the C version just calls "mkdir". Let's replicate
the old behavior. We have to create the git and worktree
leading dirs separately; while most of the time, the
worktree dir contains the git dir (as .git), the user can
override this using GIT_WORK_TREE.
We can reuse safe_create_leading_directories, but we need to
make a copy of our const buffer to do so. Since
merge-recursive uses the same pattern, we can factor this
out into a global function. This has two other cleanup
advantages for merge-recursive:
1. mkdir_p wasn't a very good name. "mkdir -p foo/bar" actually
creates bar, but this function just creates the leading
directories.
2. mkdir_p took a mode argument, but it was completely
ignored.
Acked-by: Daniel Barkalow <barkalow@iabervon.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Otherwise remote executions directly over ssh won't find them as they used
to. --upload-pack and --receive-pack options _could_ be used on the
client side, but things should keep working out-of-box for older clients.
Later versions of clients (fetch-pack and send-pack) probably could start
asking for these programs with dashless form, but that is a different
topic.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We used to mark hooks we ship as samples by making them unexecutable, but
some filesystems cannot tell what is executable and what is not.
This makes it much more explicit. The hooks are suffixed with .sample
(but now are made executable), so enabling it is still one step operation
(instead of "chmod +x $hook", you would do "mv $hook.sample $hook") but
now they won't get accidentally enabled on systems without executable bit.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This hook is what I have been using to manage topic branches in git.git,
but have not been updated to the Real Thing for a while.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
New pack structures are currently allocated in 2 different places
and all members have to be initialized explicitly. This is prone
to errors leading to segmentation faults as found by Teemu Likonen.
Let's have a common place where this structure is allocated, and have
all members explicitly initialized to zero.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, execv_git_cmd() always try running the dashed form, which
means we cannot easily remove the git-foo hardlinks for built-in
commands. This updates the function to always exec "git foo" form, and
makes sure "git" potty does not infinitely recurse to itself.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
