While working on parallelizing the tests in Azure Pipelines, an issue
was discovered with the `is_msys2_sh()` function: it expects the path
components to be separated by exactly one dir separator. That does not
need to be the case, though, e.g. when the components in the `PATH`
variable have trailing slashes.
Let's make the code much more robust in this respect.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Every once in a while, the Azure Pipeline fails with some semi-random
error: timer thread did not terminate timely
This error message means that the thread that is used to emulate the
setitimer() function did not terminate within 1,000 milliseconds.
The most likely explanation (and therefore the one we should assume to
be true, according to Occam's Razor) is that the timeout of one second
is simply not enough because we try to run so many tasks in parallel.
So let's give it ten seconds instead of only one. That should be enough.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
For many Win32 functions, there actually exist two variants: one with
the `A` suffix that takes ANSI parameters (`char *` or `const char *`)
and one with the `W` suffix that takes Unicode parameters (`wchar_t *`
or `const wchar_t *`).
Let's be precise what we want to use.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Using FindFirstFileExW() requires the OS to allocate a 64K buffer for each
directory and then free it when we call FindClose(). Update fscache to call
the underlying kernel API NtQueryDirectoryFile so that we can do the buffer
management ourselves. That allows us to allocate a single buffer for the
lifetime of the cache and reuse it for each directory.
This change improves performance of 'git status' by 18% in a repo with ~200K
files and 30k folders.
Documentation for NtQueryDirectoryFile can be found at:
https://docs.microsoft.com/en-us/windows-hardware/drivers/ddi/content/ntifs/nf-ntifs-ntquerydirectoryfilehttps://docs.microsoft.com/en-us/windows/desktop/FileIO/file-attribute-constantshttps://docs.microsoft.com/en-us/windows/desktop/fileio/reparse-point-tags
To determine if the specified directory is a symbolic link, inspect the
FileAttributes member to see if the FILE_ATTRIBUTE_REPARSE_POINT flag is
set. If so, EaSize will contain the reparse tag (this is a so far
undocumented feature, but confirmed by the NTFS developers). To
determine if the reparse point is a symbolic link (and not some other
form of reparse point), test whether the tag value equals the value
IO_REPARSE_TAG_SYMLINK.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
We cannot rely on `uname -m` in Git for Windows' SDK to tell us what
architecture we are compiling for, as we can compile both 32-bit and
64-bit `git.exe` from a 64-bit SDK, but the `uname -m` in that SDK will
always report `x86_64`.
So let's go back to our original design. And make it explicitly
Windows-specific.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The recent change to make fscache thread specific relied on fscache_enable()
being called first from the primary thread before being called in parallel
from worker threads. Make that more robust and protect it with a critical
section to avoid any issues.
Helped-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Ben Peart <benpeart@microsoft.com>
To support IPv6, Git provided fall back functions for Windows versions that
did not support IPv6. However, as Git dropped support for Windows XP and
prior, those functions are not needed anymore.
Removed those fallbacks by reverting commit[1] and using the functions
directly (without 'ipv6_' prefix).
[1] fe3b2b7b82.
Signed-off-by: tanushree27 <tanushreetumane@gmail.com>
When debugging Git, the criss-cross spawning of processes can make
things quite a bit difficult, especially when a Unix shell script is
thrown in the mix that calls a `git.exe` that then segfaults.
To help debugging such things, we introduce the `open_in_gdb()` function
which can be called at a code location where the segfault happens (or as
close as one can get); This will open a new MinTTY window with a GDB
that already attached to the current process.
Inspired by Derrick Stolee.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Now that the fscache is single threaded, take advantage of the mem_pool as
the allocator to significantly reduce the cost of allocations and frees.
With the reduced cost of free, in future patches, we can start freeing the
fscache at the end of commands instead of just leaking it.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
The threading model for fscache has been to have a single, global cache.
This puts requirements on it to be thread safe so that callers like
preload-index can call it from multiple threads. This was implemented
with a single mutex and completion events which introduces contention
between the calling threads.
Simplify the threading model by making fscache thread specific. This allows
us to remove the global mutex and synchronization events entirely and instead
associate a fscache with every thread that requests one. This works well with
the current multi-threading which divides the cache entries into blocks with
a separate thread processing each block.
At the end of each worker thread, if there is a fscache on the primary
thread, merge the cached results from the worker into the primary thread
cache. This enables us to reuse the cache later especially when scanning for
untracked files.
In testing, this reduced the time spent in preload_index() by about 25% and
also reduced the CPU utilization significantly. On a repo with ~200K files,
it reduced overall status times by ~12%.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Update enable_fscache() to take an optional initial size parameter which is
used to initialize the hashmap so that it can avoid having to rehash as
additional entries are added.
Add a separate disable_fscache() macro to make the code clearer and easier
to read.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Track fscache hits and misses for lstat and opendir requests. Reporting of
statistics is done when the cache is disabled for the last time and freed
and is only reported if GIT_TRACE_FSCACHE is set.
Sample output is:
11:33:11.836428 compat/win32/fscache.c:433 fscache: lstat 3775, opendir 263, total requests/misses 4052/269
Signed-off-by: Ben Peart <benpeart@microsoft.com>
Use FindFirstFileExW with FindExInfoBasic to avoid forcing NTFS to look up
the short name. Also switch to a larger (64K vs 4K) buffer using
FIND_FIRST_EX_LARGE_FETCH to minimize round trips to the kernel.
In a repo with ~200K files, this drops warm cache status times from 3.19
seconds to 2.67 seconds for a 16% savings.
Signed-off-by: Ben Peart <benpeart@microsoft.com>
This fixes the issue identified in
https://github.com/git-for-windows/git/issues/1498
where Git would not fall back to reading credentials from a Win32
Console when the credentials could not be read from the terminal via the
Bash hack (that is necessary to support running in a MinTTY).
Tested in a Powershell window.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This topic branch brings slightly experimental changes supporting Git
for Windows to use BusyBox-w32 to execute its shell scripts as well as
its test suite.
The test suite can be run by installing the test artifacts into a MinGit
that has busybox.exe (and using Git for Windows' SDK's Perl for now, as
the test suite requires Perl even when NO_PERL is set, go figure) by
using the `install-mingit-test-artifacts` Makefile target with the
DESTDIR variable pointing to the top-level directory of the MinGit
installation.
To facilitate running the test suite (without having `make` available,
as `make.exe` is not part of MinGit), this branch brings an experimental
patch to the `test-run-command` helper to run Git's test suite. It is
still very experimental, though: in this developer's tests it seemed
that the `poll()` emulation required for `run_parallel_processes()` to
work sometimes hiccups on Windows, causing infinite "hangs". It is also
possible that BusyBox itself has problems writing to the pipes opened by
`test-run-command` (and merging this branch will help investigate
further). Caveat emptor.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When spawning child processes, we do want them to inherit the standard
handles so that we can talk to them. We do *not* want them to inherit
any other handle, as that would hold a lock to the respective files
(preventing them from being renamed, modified or deleted), and the child
process would not know how to access that handle anyway.
Happily, there is an API to make that happen. It is supported in Windows
Vista and later, which is exactly what we promise to support in Git for
Windows for the time being.
This also means that we lift, at long last, the target Windows version
from Windows XP to Windows Vista.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This branch introduces support for reading the "Windows-wide" Git
configuration from `%PROGRAMDATA%\Git\config`. As these settings are
intended to be shared between *all* Git-related software, that config
file takes an even lower precedence than `$(prefix)/etc/gitconfig`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
It turns out that when running in a Powershell window, we need to turn
on ENABLE_ECHO_INPUT because the default would be *not* to echo
anything.
This also ensures that we use the input mode where all input is read
until the user hits the Return key.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
To support Git Bash running in a MinTTY, we use a dirty trick to access
the MSYS2 pseudo terminal: we execute a Bash snippet that accesses
/dev/tty.
The idea was to fall back to writing to/reading from CONOUT$/CONIN$ if
that Bash call failed because Bash was not found.
However, we should fall back even in other error conditions, because we
have not successfully read the user input. Let's make it so.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When trying to query the user directly via /dev/tty, both WSL's bash and
BusyBox' bash emulation seem to have problems printing the value that
they just read. The bash just stops in those instances, does not even
execute any commands after the echo command.
Let's just work around this by running the Bash snippet only in MSYS2's
Bash: its `SHELL` variable has the `.exe` suffix, and neither WSL's nor
BusyBox' bash set the `SHELL` variable to a path with that suffix. In
the latter case, we simply exit with code 127 (indicating that the
command was not found) and fall back to the CONIN$/CONOUT$ method
quietly.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
BusyBox comes with a ton of applets ("applet" being the identical
concept to Git's "builtins"). And similar to Git's builtins, the applets
can be called via `busybox <command>`, or the BusyBox executable can be
copied/hard-linked to the command name.
The similarities do not end here. Just as with Git's builtins, it is
problematic that BusyBox' hard-linked applets cannot easily be put into
a .zip file: .zip archives have no concept of hard-links and therefore
would store identical copies (and also extract identical copies,
"inflating" the archive unnecessarily).
To counteract that issue, MinGit already ships without hard-linked
copies of the builtins, and the plan is to do the same with BusyBox'
applets: simply ship busybox.exe as single executable, without
hard-linked applets.
To accommodate that, Git is being taught by this commit a very special
trick, exploiting the fact that it is possible to call an executable
with a command-line whose argv[0] is different from the executable's
name: when `sh` is to be spawned, and no `sh` is found in the PATH, but
busybox.exe is, use that executable (with unchanged argv).
Likewise, if any executable to be spawned is not on the PATH, but
busybox.exe is found, parse the output of `busybox.exe --help` to find
out what applets are included, and if the command matches an included
applet name, use busybox.exe to execute it.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The main idea of this patch is that even if we have to look up the
absolute path of the script, if only the basename was specified as
argv[0], then we should use that basename on the command line, too, not
the absolute path.
This patch will also help with the upcoming patch where we automatically
substitute "sh ..." by "busybox sh ..." if "sh" is not in the PATH but
"busybox" is: we will do that by substituting the actual executable, but
still keep prepending "sh" to the command line.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
By default, CreateProcess() does not inherit any open file handles,
unless the bInheritHandles parameter is set to TRUE. Which we do need to
set because we need to pass in stdin/stdout/stderr to talk to the child
processes. Sadly, this means that all file handles (unless marked via
O_NOINHERIT) are inherited.
This lead to problems in GVFS Git, where a long-running read-object hook
is used to hydrate missing objects, and depending on the circumstances,
might only be called *after* Git opened a file handle.
Ideally, we would not open files without O_NOINHERIT unless *really*
necessary (i.e. when we want to pass the opened file handle as standard
handle into a child process), but apparently it is all-too-easy to
introduce incorrect open() calls: this happened, and prevented updating
a file after the read-object hook was started because the hook still
held a handle on said file.
Happily, there is a solution: as described in the "Old New Thing"
https://blogs.msdn.microsoft.com/oldnewthing/20111216-00/?p=8873 there
is a way, starting with Windows Vista, that lets us define precisely
which handles should be inherited by the child process.
And since we bumped the minimum Windows version for use with Git for
Windows to Vista with v2.10.1 (i.e. a *long* time ago), we can use this
method. So let's do exactly that.
We need to make sure that the list of handles to inherit does not
contain duplicates; Otherwise CreateProcessW() would fail with
ERROR_INVALID_ARGUMENT.
While at it, stop setting errno to ENOENT unless it really is the
correct value.
Also, fall back to not limiting handle inheritance under certain error
conditions (e.g. on Windows 7, which is a lot stricter in what handles
you can specify to limit to).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This topic branch conflicts with the next change that will change the
way we call `CreateProcessW()`. So let's merge it early, to avoid merge
conflicts during a merge (because we would have to resolve this with
every single merging-rebase).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This branch introduces support for reading the "Windows-wide" Git
configuration from `%PROGRAMDATA%\Git\config`. As these settings are
intended to be shared between *all* Git-related software, that config
file takes an even lower precedence than `$(prefix)/etc/gitconfig`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Previously, we did not install any handler for Ctrl+C, but now we really
want to because the MSYS2 runtime learned the trick to call the
ConsoleCtrlHandler when Ctrl+C was pressed.
With this, hitting Ctrl+C while `git log` is running will only terminate
the Git process, but not the pager. This finally matches the behavior on
Linux and on macOS.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
From the documentation of said setting:
This boolean will enable fsync() when writing object files.
This is a total waste of time and effort on a filesystem that
orders data writes properly, but can be useful for filesystems
that do not use journalling (traditional UNIX filesystems) or
that only journal metadata and not file contents (OS X’s HFS+,
or Linux ext3 with "data=writeback").
The most common file system on Windows (NTFS) does not guarantee that
order, therefore a sudden loss of power (or any other event causing an
unclean shutdown) would cause corrupt files (i.e. files filled with
NULs). Therefore we need to change the default.
Note that the documentation makes it sound as if this causes really bad
performance. In reality, writing loose objects is something that is done
only rarely, and only a handful of files at a time.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
For some reason, when being called via TortoiseGit the standard handles,
or at least what is returned by _get_osfhandle(0) for standard input,
can take on the value (HANDLE)-2 (which is not a legal value, according
to the documentation).
Even if this value is not documented anywhere, CreateProcess() seems to
work fine without complaints if hStdInput set to this value.
In contrast, the upcoming code to restrict which file handles get
inherited by spawned processes would result in `ERROR_INVALID_PARAMETER`
when including such handle values in the list.
To help this, special-case the value (HANDLE)-2 returned by
_get_osfhandle() and replace it with INVALID_HANDLE_VALUE, which will
hopefully let the handle inheritance restriction work even when called
from TortoiseGit.
This fixes https://github.com/git-for-windows/git/issues/1481
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The TerminateProcess() function does not actually leave the child
processes any chance to perform any cleanup operations. This is bad
insofar as Git itself expects its signal handlers to run.
A symptom is e.g. a left-behind .lock file that would not be left behind
if the same operation was run, say, on Linux.
To remedy this situation, we use an obscure trick: we inject a thread
into the process that needs to be killed and to let that thread run the
ExitProcess() function with the desired exit status. Thanks J Wyman for
describing this trick.
The advantage is that the ExitProcess() function lets the atexit
handlers run. While this is still different from what Git expects (i.e.
running a signal handler), in practice Git sets up signal handlers and
atexit handlers that call the same code to clean up after itself.
In case that the gentle method to terminate the process failed, we still
fall back to calling TerminateProcess(), but in that case we now also
make sure that processes spawned by the spawned process are terminated;
TerminateProcess() does not give the spawned process a chance to do so
itself.
Please note that this change only affects how Git for Windows tries to
terminate processes spawned by Git's own executables. Third-party
software that *calls* Git and wants to terminate it *still* need to make
sure to imitate this gentle method, otherwise this patch will not have
any effect.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
... even if they may look like them.
As looking up the target of the "symbolic link" (just to see whether it
starts with `/ContainerMappedDirectories/`) is pretty expensive, we
do it when we can be *really* sure that there is a possibility that this
might be the case.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: JiSeop Moon <zcube@zcube.kr>
This is retry of #1419.
I added flush_fscache macro to flush cached stats after disk writing
with tests for regression reported in #1438 and #1442.
git checkout checks each file path in sorted order, so cache flushing does not
make performance worse unless we have large number of modified files in
a directory containing many files.
Using chromium repository, I tested `git checkout .` performance when I
delete 10 files in different directories.
With this patch:
TotalSeconds: 4.307272
TotalSeconds: 4.4863595
TotalSeconds: 4.2975562
Avg: 4.36372923333333
Without this patch:
TotalSeconds: 20.9705431
TotalSeconds: 22.4867685
TotalSeconds: 18.8968292
Avg: 20.7847136
I confirmed this patch passed all tests in t/ with core_fscache=1.
Signed-off-by: Takuto Ikuta <tikuta@chromium.org>
In preparation for making this function a bit more complicated (to allow
for special-casing the `ContainerMappedDirectories` in Windows
containers, which look like a symbolic link, but are not), let's move it
out of the header.
Signed-off-by: JiSeop Moon <zcube@zcube.kr>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
It is a known issue that a rename() can fail with an "Access denied"
error at times, when copying followed by deleting the original file
works. Let's just fall back to that behavior.
Signed-off-by: JiSeop Moon <zcube@zcube.kr>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This branch introduces support for reading the "Windows-wide" Git
configuration from `%PROGRAMDATA%\Git\config`. As these settings are
intended to be shared between *all* Git-related software, that config
file takes an even lower precedence than `$(prefix)/etc/gitconfig`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Make fscache_enabled() function public rather than static.
Remove unneeded fscache_is_enabled() function.
Change is_fscache_enabled() macro to call fscache_enabled().
is_fscache_enabled() now takes a pathname so that the answer
is more precise and mean "is fscache enabled for this pathname",
since fscache only stores repo-relative paths and not absolute
paths, we can avoid attempting lookups for absolute paths.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Teach FSCACHE to remember "not found" directories.
This is a performance optimization.
FSCACHE is a performance optimization available for Windows. It
intercepts Posix-style lstat() calls into an in-memory directory
using FindFirst/FindNext. It improves performance on Windows by
catching the first lstat() call in a directory, using FindFirst/
FindNext to read the list of files (and attribute data) for the
entire directory into the cache, and short-cut subsequent lstat()
calls in the same directory. This gives a major performance
boost on Windows.
However, it does not remember "not found" directories. When STATUS
runs and there are missing directories, the lstat() interception
fails to find the parent directory and simply return ENOENT for the
file -- it does not remember that the FindFirst on the directory
failed. Thus subsequent lstat() calls in the same directory, each
re-attempt the FindFirst. This completely defeats any performance
gains.
This can be seen by doing a sparse-checkout on a large repo and
then doing a read-tree to reset the skip-worktree bits and then
running status.
This change reduced status times for my very large repo by 60%.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Teach read_directory_recursive() and add_excludes() to
be aware of optional fscache and avoid trying to open()
and fstat() non-existant ".gitignore" files in every
directory in the worktree.
The current code in add_excludes() calls open() and then
fstat() for a ".gitignore" file in each directory present
in the worktree. Change that when fscache is enabled to
call lstat() first and if present, call open().
This seems backwards because both lstat needs to do more
work than fstat. But when fscache is enabled, fscache will
already know if the .gitignore file exists and can completely
avoid the IO calls. This works because of the lstat diversion
to mingw_lstat when fscache is enabled.
This reduced status times on a 350K file enlistment of the
Windows repo on a NVMe SSD by 0.25 seconds.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
