Symlinks on Windows don't work the same way as on Unix systems. E.g. there
are different types of symlinks for directories and files, creating
symlinks requires administrative privileges etc.
By default, disable symlink support on Windows. I.e. users explicitly have
to enable it with 'git config [--system|--global] core.symlinks true'.
The test suite ignores system / global config files. Allow testing *with*
symlink support by checking if native symlinks are enabled in MSys2 (via
'MSYS=winsymlinks:nativestrict').
Reminder: This would need to be changed if / when we find a way to run the
test suite in a non-MSys-based shell (e.g. dash).
Signed-off-by: Karsten Blees <blees@dcon.de>
The retry pattern is duplicated in three places. It also seems to be too
hard to use: mingw_unlink() and mingw_rmdir() duplicate the code to retry,
and both of them do so incompletely. They also do not restore errno if the
user answers 'no'.
Introduce a retry_ask_yes_no() helper function that handles retry with
small delay, asking the user, and restoring errno.
mingw_unlink: include _wchmod in the retry loop (which may fail if the
file is locked exclusively).
mingw_rmdir: include special error handling in the retry loop.
Signed-off-by: Karsten Blees <blees@dcon.de>
Dynamic loading of DLL functions is duplicated in several places.
Add a set of macros to simplify the process.
Signed-off-by: Karsten Blees <blees@dcon.de>
Git typically doesn't trust the stat.st_size member of symlinks (e.g. see
strbuf_readlink()). However, some functions take shortcuts if st_size is 0
(e.g. diff_populate_filespec()).
In mingw_lstat() and fscache_lstat(), make sure to return an adequate size.
The extra overhead of opening and reading the reparse point to calculate
the exact size is not necessary, as git doesn't rely on the value anyway.
Signed-off-by: Karsten Blees <blees@dcon.de>
Move S_IFLNK detection to file_attr_to_st_mode() and reuse it in fscache.
Implement DT_LNK detection in dirent.c and the fscache readdir version.
Signed-off-by: Karsten Blees <blees@dcon.de>
When obtaining lstat information for reparse points, we need to call
FindFirstFile() in addition to GetFileInformationEx() to obtain the type
of the reparse point (symlink, mount point etc.). However, currently there
is no error handling whatsoever if FindFirstFile() fails.
Call FindFirstFile() before modifying the stat *buf output parameter and
error out if the call fails.
Note: The FindFirstFile() return value includes all the data that we get
from GetFileAttributesEx(), so we could replace GetFileAttributesEx() with
FindFirstFile(). We don't do that because GetFileAttributesEx() is about
twice as fast for single files. I.e. we only pay the extra cost of calling
FindFirstFile() in the rare case that we encounter a reparse point.
Note: The indentation of the remaining reparse point code will be fixed in
the next patch.
Signed-off-by: Karsten Blees <blees@dcon.de>
With the new mingw_stat() implementation, do_lstat() is only called from
mingw_lstat() (with follow == 0). Remove the extra function and the old
mingw_stat()-specific (follow == 1) logic.
Signed-off-by: Karsten Blees <blees@dcon.de>
With respect to symlinks, the current stat() implementation is almost the
same as lstat(): except for the file type (st_mode & S_IFMT), it returns
information about the link rather than the target.
Implement stat by opening the file with as little permissions as possible
and calling GetFileInformationByHandle on it. This way, all link resoltion
is handled by the Windows file system layer.
If symlinks are disabled, use lstat() as before, but fail with ELOOP if a
symlink would have to be resolved.
Signed-off-by: Karsten Blees <blees@dcon.de>
GetFileAttributes cannot handle paths with trailing dir separator. The
current [l]stat implementation calls GetFileAttributes twice if the path
has trailing slashes (first with the original path passed to [l]stat, and
and a second time with a path copy with trailing '/' removed).
With Unicode conversion, we get the length of the path for free and also
have a (wide char) buffer that can be modified.
Remove trailing directory separators before calling the Win32 API.
Signed-off-by: Karsten Blees <blees@dcon.de>
strbuf_readlink() refuses to read link targets that exceed PATH_MAX (even
if a sufficient size was specified by the caller).
As some platforms support longer paths, remove this restriction (similar
to strbuf_getcwd()).
Signed-off-by: Karsten Blees <blees@dcon.de>
strbuf_readlink() calls readlink() twice if the hint argument specifies the
exact size of the link target (e.g. by passing stat.st_size as returned by
lstat()). This is necessary because 'readlink(..., hint) == hint' could
mean that the buffer was too small.
Use hint + 1 as buffer size to prevent this.
Signed-off-by: Karsten Blees <blees@dcon.de>
POSIX semantics requires lstat() to fail with ENOTDIR when "[a]
component of the path prefix names an existing file that is neither a
directory nor a symbolic link to a directory".
See http://pubs.opengroup.org/onlinepubs/9699919799/functions/lstat.html
This behavior is expected by t1404-update-ref-df-conflicts now.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Using different code pages for console input (SetConsoleCP()) and console
output (SetConsoleOutputCP()) doesn't make much sense and may be hazardous
for native Windows programs.
Git uses UTF-8 internally, so it actually needs 'SetConsoleCP(CP_UTF8)'
rather than 'SetConsoleCP(GetACP())'. However, ReadFile() / ReadConsoleA()
are broken with CP_UTF8 (and thus any higher level APIs such as fgetc(),
getchar() etc.). Unicode-aware console input would have to be implemented
via mingw_* wrappers using ReadConsoleW(). As Git typically launches an
editor for anything more complex than ASCII-only, yes/no-style questions,
this is currently not a problem.
Drop 'SetConsoleCP()' from the git-wrapper, so that input and output code
pages stay in sync.
Signed-off-by: Karsten Blees <blees@dcon.de>
When we fall back to starting the Git Bash in the regular Windows
console, we need to show said console's window... So let's introduce yet
another configuration knob for use via string resources.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
With a recent change in Cygwin (which is the basis of the msys2-runtime),
a GUI process desiring to launch an MSys2 executable needs to allocate a
console for the new process (otherwise the process will just hang on
Windows XP). _Git Bash_ is such a GUI process.
While at it, use correct handles when inheriting the stdin/stdout/stderr
handles: `GetStdHandle()` returns NULL for invalid handles, but the
STARTUPINFO must specify `INVALID_HANDLE_VALUE` instead.
Originally, the hope was that only this `NULL` => `INVALID_HANDLE_VALUE`
conversion would be required to fix the Windows XP issue mentioned above
(extensive debugging revealed that starting _Git Bash_ on Windows XP would
yield invalid handles for `stdin` and `stderr`, but *not* for `stdout`).
However, while _Git Bash_ eventually showed a `mintty` when not allocating
a new console, it took around one second to show it, and several seconds
to close it. So let's just include both fixes.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
With the resource-driven command-line configuration, we introduced the
option to ensure that only the PATH environment variable is edited only
minimally, i.e. only /cmd/ is added (as appropriate for _Git CMD_).
We are about to add another option, so let's refactor the equivalent of
Git's `strip_prefix()` function; It is not *quite* the same because we
have to `memmove()` the remainder to the beginning of the buffer.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
git-wrapper fails to initialize HOME correctly if $HOMEDRIVE$HOMEPATH
points to a disconnected network drive.
Check if the directory exists before using $HOMEDRIVE$HOMEPATH.
Signed-off-by: Karsten Blees <blees@dcon.de>
This change accompanies the `--no-cd` option when configured via
resources. It is required to support `Git Bash Here`: when
right-clicking an icon in the Explorer to start a Bash, the working
directory is actually the directory that is displayed in the Explorer.
That means if the clicked icon actually refers to a directory, the
working directory would be its *parent* directory.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
To avoid that ugly Console window when calling \cmd\git.exe gui...
To avoid confusion with builtins, we need to move the code block
handling gitk (and now git-gui, too) to intercept before git-gui is
mistaken for a builtin.
Unfortunately, git-gui is in libexec/git-core/ while gitk is in bin/,
therefore we need slightly more adjustments than just moving and
augmenting the gitk case.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When *Git for Windows* is installed into a directory that has spaces in
it, e.g. `C:\Program Files\Git`, the `git-wrapper` appends this directory
unquoted when fixing up the command line. To resolve this, just quote the
provided `execpath`.
Signed-off-by: nalla <nalla@hamal.uberspace.de>
The idea of having the Git wrapper in the /cmd/ directory is to allow
adding only a *tiny* set of executables to the search path, to allow
minimal interference with other software applications. It is quite
likely, for example, that other software applications require their own
version of zlib1.dll and would not be overly happy to find the version
Git for Windows ships.
The /cmd/ directory also gives us the opportunity to let the Git wrapper
handle the `gitk` script. It is a Tcl/Tk script that is not recognized
by Windows, therefore calling `gitk` in `cmd.exe` would not work, even
if we add all of Git for Windows' bin/ directories.
So let's use the /cmd/ directory instead of adding /mingw??/bin/ and
/usr/bin/ to the PATH when launching Git CMD.
The way we implemented Git CMD is to embed the appropriate command line
as string resource into a copy of the Git wrapper. Therefore we extended
that syntax to allow for configuring a minimal search path.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
We recently added the ability to configure copies of the Git wrapper to
launch custom command-lines, configured via plain old Windows resources.
The main user is Git for Windows' `git-bash.exe`, of course. When the
user double-clicks the `git bash` icon, it makes sense to start the Bash
in the user's home directory.
Third-party software, such as TortoiseGit or GitHub for Windows, may
want to start the Git Bash in another directory, though.
Now, when third-party software wants to call Git, they already have to
construct a command-line, and can easily pass a command-line option
`--no-cd` (which this commit introduces), and since that option is not
available when the user double-clicks an icon on the Desktop or in the
Explorer, let's keep the default to switch to the home directory if the
`--no-cd` flag was not passed along.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When we rewrite the command-line to call the *real* Git, we want to skip
the first command-line parameter. The previous code worked in most
circumstances, but was a bit fragile because it assumed that no fancy
quoting would take place.
In the next commit, we will want to have the option to skip more than
just one command-line parameter, so we have to be much more careful with
the command-line handling.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
In the meantime, Git for Windows learned to handle those subcommands
quite well itself; There is no longer a need to special-case them in the
wrapper.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
In a push to polish Git for Windows more, we are moving away from
scripts toward proper binaries.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The Git wrapper does one thing, and does it well: setting up the
environment required to run Git and its scripts, and then hand off to
another program.
We already do this for the Git executable itself; in Git for Windows'
context, we have exactly the same need also when calling the Git Bash or
Git CMD. However, both are tied to what particular shell environment you
use, though: MSys or MSys2 (or whatever else cunning developers make
work for them). This means that the Git Bash and Git CMD need to be
compiled in the respective context (e.g. when compiling the
mingw-w64-git package in the MSys2 context).
Happily, Windows offers a way to configure compiled executables:
resources. So let's just look whether the current executable has a
string resource and use it as the command-line to execute after the
environment is set up. To support MSys2's Git Bash better (where
`mintty` should, but might not, be available), we verify whether the
specified executable exists, and keep looking for string resources if it
does not.
For even more flexibility, we expand environment variables specified as
`@@<VARIABLE-NAME>@@`, and for convenience `@@EXEPATH@@` expands into
the directory in which the executable resides.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Otherwise the output of Git commands cannot be caught by, say, Git GUI
(because it is running detached from any console, which would make
`git.exe` inherit the standard handles implicitly).
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
We are about to use the Git wrapper to call the Git Bash of Git for
Windows. All the wrapper needs to do for that is to set up the
environment variables, use the home directory as working directory and
then hand off to a user-specified command-line.
We prepare the existing code for this change by introducing flags to set
up the environment variables, to launch a non-Git program, and to use
the home directory as working directory.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
The original purpose of the Git wrapper is to run from inside Git for
Windows' /cmd/ directory, to allow setting up some environment variables
before Git is allowed to take over.
Due to differences in the file system layout, MSys2 requires some
changes for that to work.
In addition, we must take care to set the `MSYSTEM` environment variable
to `MINGW32` or `MINGW64`, respectively, to allow MSys2 to be configured
correctly in case Git launches a shell or Perl script.
We also need to change the `TERM` variable to `cygwin` instead of
`msys`, otherwise the pager `less.exe` (spawned e.g. by `git log`) will
simply crash with a message similar to this one:
1 [main] less 9832 cygwin_exception::open_stackdumpfile:
Dumping stack trace to less.exe.stackdump
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This reduces the disk footprint of a full Git for Windows setup
dramatically because on Windows, one cannot assume that hard links are
supported.
The net savings are calculated easily: the 32-bit `git.exe` file weighs
in with 7662 kB while the `git-wrapper.exe` file (modified to serve as a
drop-in replacement for builtins) weighs a scant 21 kB. At this point,
there are 109 builtins which results in a total of 813 MB disk space
being freed up by this commit.
Yes, that is really more than half a gigabyte.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Git started out as a bunch of separate commands, in the true Unix spirit.
Over time, more and more functionality was shared between the different
Git commands, though, so it made sense to introduce the notion of
"builtins": programs that are actually integrated into the main Git
executable.
These builtins can be called in two ways: either by specifying a
subcommand as the first command-line argument, or -- for backwards
compatibility -- by calling the Git executable hardlinked to a filename
of the form "git-<subcommand>". Example: the "log" command can be called
via "git log <parameters>" or via "git-log <parameters>". The latter
form is actually deprecated and only supported for scripts; calling
"git-log" interactively will not even work by default because the
libexec/git-core/ directory is not in the PATH.
All of this is well and groovy as long as hard links are supported.
Sadly, this is not the case in general on Windows. So it actually hurts
quite a bit when you have to fall back to copying all of git.exe's
currently 7.5MB 109 times, just for backwards compatibility.
The simple solution would be to install really trivial shell script
wrappers in place of the builtins:
for builtin in $BUILTINS
do
rm git-$builtin.exe
printf '#!/bin/sh\nexec git %s "$@"\n' $builtin > git-builtin
chmod a+x git-builtin
done
This method would work -- even on Windows because Git for Windows ships a
full-fledged Bash. However, the Windows Bash comes at a price: it needs to
spin up a full-fledged POSIX emulation layer everytime it starts.
Therefore, the shell script solution would incur a significant performance
penalty.
The best solution the Git for Windows team could come up with is to extend
the Git wrapper -- that is needed to call Git from cmd.exe anyway, and
that weighs in with a scant 19KB -- to also serve as a drop-in replacement
for the builtins so that the following workaround is satisfactory:
for builtin in $BUILTINS
do
cp git-wrapper.exe git-$builtin.exe
done
This commit allows for this, by extending the module file parsing to
turn builtin command names like `git-log.exe ...` into calls to the main
Git executable: `git.exe log ...`.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This prepares the wrapper for modifications to serve as a drop-in
replacement for the builtins.
This commit's diff is best viewed with the `-w` flag.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
We take care to embed the manifest, too, because we will modify the
wrapper in the next few commits to serve as a drop-in replacement for
the built-ins, i.e. we will want to call the wrapper under names such
as 'git-patch-id.exe', too.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
On Windows, Git is faced by the challenge that it has to set up certain
environment variables before running Git under special circumstances
such as when Git is called directly from cmd.exe (i.e. outside any
Bash environment).
This source code was taken from msysGit's commit 74a198d:
https://github.com/msysgit/msysgit/blob/74a198d/src/git-wrapper/git-wrapper.c
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Since baaf233 (connect: improve check for plink to reduce false
positives, 2015-04-26), t5601 writes out a `plink.exe` for testing that
is actually a shell script. So the assumption that the `.exe` extension
implies that the file is *not* a shell script is now wrong.
The original idea to special-case `.exe` files was probably to
help performance, but since we are in a code path that involves spawning
a new process (which in and of itself is pretty slow on Windows anyway),
we pursue a better idea to improve performance elsewhere: we try to
convert scripts into builtins and to reduce the number of spawned
processes by adding more internal API calls.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Accessing the Windows console through the special CONIN$ / CONOUT$ devices
doesn't work properly for non-ASCII usernames an passwords.
It also doesn't work for terminal emulators that hide the native console
window (such as mintty), and 'TERM=xterm*' is not necessarily a reliable
indicator for such terminals.
The new shell_prompt() function, on the other hand, works fine for both
MSys1 and MSys2, in native console windows as well as mintty, and properly
supports Unicode. It just needs bash on the path (for 'read -s', which is
bash-specific).
On Windows, try to use the shell to read from the terminal. If that fails
with ENOENT (i.e. bash was not found), use CONIN/OUT as fallback.
Note: To test this, create a UTF-8 credential file with non-ASCII chars,
e.g. in git-bash: 'echo url=http://täst.com > cred.txt'. Then in git-cmd,
'git credential fill <cred.txt' works (shell version), while calling git
without the git-wrapper (i.e. 'mingw64\bin\git credential fill <cred.txt')
mangles non-ASCII chars in both console output and input.
Signed-off-by: Karsten Blees <blees@dcon.de>
The `git_terminal_prompt()` function expects the terminal window to be
attached to a Win32 Console. However, this is not the case with terminal
windows other than `cmd.exe`'s, e.g. with MSys2's own `mintty`.
Non-cmd terminals such as `mintty` still have to have a Win32 Console
to be proper console programs, but have to hide the Win32 Console to
be able to provide more flexibility (such as being resizeable not only
vertically but also horizontally). By writing to that Win32 Console,
`git_terminal_prompt()` manages only to send the prompt to nowhere and
to wait for input from a Console to which the user has no access.
This commit introduces a function specifically to support `mintty` -- or
other terminals that are compatible with MSys2's `/dev/tty` emulation. We
use the `TERM` environment variable as an indicator for that: if the value
starts with "xterm" (such as `mintty`'s "xterm_256color"), we prefer to
let `xterm_prompt()` handle the user interaction.
The most prominent user of `git_terminal_prompt()` is certainly
`git-remote-https.exe`. It is an interesting use case because both
`stdin` and `stdout` are redirected when Git calls said executable, yet
it still wants to access the terminal.
When running inside a `mintty`, the terminal is not accessible to the
`git-remote-https.exe` program, though, because it is a MinGW program
and the `mintty` terminal is not backed by a Win32 console.
To solve that problem, we simply call out to the shell -- which is an
*MSys2* program and can therefore access `/dev/tty`.
Helped-by: nalla <nalla@hamal.uberspace.de>
Signed-off-by: Karsten Blees <blees@dcon.de>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
For performance reasons `stdout` is not unbuffered by default. That leads
to problems if after printing to `stdout` a read on `stdin` is performed.
For that reason interactive commands like `git clean -i` do not function
properly anymore if the `stdout` is not flushed by `fflush(stdout)` before
trying to read from `stdin`.
In the case of `git clean -i` all reads on `stdin` were preceded by a
`fflush(stdout)` call.
Signed-off-by: nalla <nalla@hamal.uberspace.de>
HOME initialization was historically duplicated in many different places,
including /etc/profile, launch scripts such as git-bash.vbs and gitk.cmd,
and (although slightly broken) in the git-wrapper.
Even unrelated projects such as GitExtensions and TortoiseGit need to
implement the same logic to be able to call git directly.
Initialize HOME in git's own startup code so that we can eventually retire
all the duplicate initialization code.
Signed-off-by: Karsten Blees <blees@dcon.de>
Lets keep the environment initialization and conversion section as lean as
possible and move recently added tweaks to setup_windows_environment().
This fixes the following potential problems:
* Prevent duplicate TZ variables if both TZ and MSYS2_TZ are set.
* Some of the higher level x* APIs from wrapper.c require a working
getenv(), using e.g. xstrdup() during initialization is dangerous.
* Slashifying the Windows TMP variable may break native Windows programs,
use POSIX TMPDIR instead.
* Properly slashify TMPDIR even if it is already set, and also if we only
have TEMP, but not TMP.
* Reduce complexity from O(n) to O(log n).
Signed-off-by: Karsten Blees <blees@dcon.de>
This developer should really, really have known better. The fact that we
are changing the environment in ways for which the MSVCRT is not
prepared for is bad enough. But then this developer followed the request
to re-enable nedmalloc -- despite the prediction that it would cause an
access violation, predicting it in the same message as the request to
re-enable nedmalloc, no less!
To paper over the issue until the time when this developer finds the
time to re-design the Unicode environment handling from scratch, let's
hope that cURL is the only library we are using that *may* set an
environment variable using MSVCRT's putenv() after we fscked the
environment up.
Note: this commit can serve as no source of pride to anyone, certainly
not yours truly. It is necessary as a quick and pragmatic stop gap,
though, to prevent worse problems.
Note: cURL manages to set the variable CHARSET when nedmalloc is *not*
enabled, without causing an access violation. In that case, it sets it
successfully to the value "cp" + GetACP() (hence it is our choice, too,
cURL may need it, Git does not):
https://github.com/bagder/curl/blob/aa5808b5/lib/easy.c#L157-L162
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Git on native Windows exclusively uses UTF-8 for console output (both with
mintty and native console windows). Gettext uses setlocale() to determine
the output encoding for translated text, however, MSVCRT's setlocale()
doesn't support UTF-8. As a result, translated text is encoded in system
encoding (GetAPC()), and non-ASCII chars are mangled in console output.
Use gettext's bind_textdomain_codeset() to force the encoding to UTF-8 on
native Windows.
In this developers' setup, HAVE_LIBCHARSET_H is apparently defined, but
we *really* want to override the locale_charset() here.
Signed-off-by: Karsten Blees <blees@dcon.de>
A '+' is not a valid part of a filename with Windows file systems (it is
reserved because the '+' operator meant file concatenation back in the
DOS days).
Let's just not use it.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
When Git for Windows was still based on MSys1, we had no gettext, ergo
no msgfmt, either. Therefore, we introduced a small and simple Tcl
script to perform the same task.
However, with MSys2, we no longer need that because we have a proper
msgfmt executable. Plus, the po2msg.sh script somehow manages to hang
when run in parallel.
Two reasons to use real msgfmt.exe instead.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
This happens to shut up t7602 on Windows which would otherwise take
the different line endings for a sign that the merge failed.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
