mingw: ensure temporary file handles are not inherited by child processes

When the index is locked and child processes inherit the handle to
said lock and the parent process wants to remove the lock before the
child process exits, on Windows there is a problem: it won't work
because files cannot be deleted if a process holds a handle on them.
The symptom:

    Rename from 'xxx/.git/index.lock' to 'xxx/.git/index' failed.
    Should I try again? (y/n)

Spawning child processes with bInheritHandles==FALSE would not work
because no file handles would be inherited, not even the hStdXxx
handles in STARTUPINFO (stdin/stdout/stderr).

Opening every file with O_NOINHERIT does not work, either, as e.g.
git-upload-pack expects inherited file handles.

This leaves us with the only way out: creating temp files with the
O_NOINHERIT flag. This flag is Windows-specific, however. For our
purposes, it is equivalent to O_CLOEXEC (which does not exist on
Windows), so let's just open temporary files with the O_CLOEXEC flag and
map that flag to O_NOINHERIT on Windows.

As Eric Wong pointed out, we need to be careful to handle the case where
the Linux headers used to compile Git support O_CLOEXEC but the Linux
kernel used to run Git does not: it returns an EINVAL.

This fixes the test that we just introduced to demonstrate the problem.

Signed-off-by: Ben Wijen <[email protected]>
Signed-off-by: Johannes Schindelin <[email protected]>
Signed-off-by: Junio C Hamano <[email protected]>

Author: bwijen

compat/mingw.h

@@ -67,6 +67,10 @@ typedef int pid_t;
 #define F_SETFD 2
 #define FD_CLOEXEC 0x1
 
+#if !defined O_CLOEXEC && defined O_NOINHERIT
+#define O_CLOEXEC	O_NOINHERIT
+#endif
+
 #ifndef EAFNOSUPPORT
 #define EAFNOSUPPORT WSAEAFNOSUPPORT
 #endif

git-compat-util.h

@@ -650,6 +650,10 @@ void *gitmemmem(const void *haystack, size_t haystacklen,
 #define getpagesize() sysconf(_SC_PAGESIZE)
 #endif
 
+#ifndef O_CLOEXEC
+#define O_CLOEXEC 0
+#endif
+
 #ifdef FREAD_READS_DIRECTORIES
 #ifdef fopen
 #undef fopen

lockfile.h

@@ -55,6 +55,10 @@
  *   * calling `fdopen_lock_file()` to get a `FILE` pointer for the
  *     open file and writing to the file using stdio.
  *
+ *   Note that the file descriptor returned by hold_lock_file_for_update()
+ *   is marked O_CLOEXEC, so the new contents must be written by the
+ *   current process, not a spawned one.
+ *
  * When finished writing, the caller can:
  *
  * * Close the file descriptor and rename the lockfile to its final

t/t6026-merge-attr.sh

@@ -181,7 +181,7 @@ test_expect_success 'up-to-date merge without common ancestor' '
 	)
 '
 
-test_expect_success !MINGW 'custom merge does not lock index' '
+test_expect_success 'custom merge does not lock index' '
 	git reset --hard anchor &&
 	write_script sleep-one-second.sh <<-\EOF &&
 		sleep 1 &

tempfile.c

@@ -120,7 +120,12 @@ int create_tempfile(struct tempfile *tempfile, const char *path)
 	prepare_tempfile_object(tempfile);
 
 	strbuf_add_absolute_path(&tempfile->filename, path);
-	tempfile->fd = open(tempfile->filename.buf, O_RDWR | O_CREAT | O_EXCL, 0666);
+	tempfile->fd = open(tempfile->filename.buf,
+			    O_RDWR | O_CREAT | O_EXCL | O_CLOEXEC, 0666);
+	if (O_CLOEXEC && tempfile->fd < 0 && errno == EINVAL)
+		/* Try again w/o O_CLOEXEC: the kernel might not support it */
+		tempfile->fd = open(tempfile->filename.buf,
+				    O_RDWR | O_CREAT | O_EXCL, 0666);
 	if (tempfile->fd < 0) {
 		strbuf_reset(&tempfile->filename);
 		return -1;

tempfile.h

@@ -33,6 +33,10 @@
  *   * calling `fdopen_tempfile()` to get a `FILE` pointer for the
  *     open file and writing to the file using stdio.
  *
+ *   Note that the file descriptor returned by create_tempfile()
+ *   is marked O_CLOEXEC, so the new contents must be written by
+ *   the current process, not any spawned one.
+ *
  * When finished writing, the caller can:
  *
  * * Close the file descriptor and remove the temporary file by