Skip to contentSkip to conversation

Various msysgit patches

PATCH In review Steffen Prohaska 6 participants 37 messages Jun 15, 2016 2 messages not in the archive Kbuild

lore
Share & tools
lore
#0 lore raw

cover Various msysgit patches

Hello,
Here are more patches that are in msysgit but not yet in official git.git. The patches have been discussed in the following two threads:

http://thread.gmane.org/gmane.comp.version-control.msysgit/5373
http://thread.gmane.org/gmane.comp.version-control.msysgit/4876

The first couple of patches are small an obvious. The patches with larger numbers are more complex and might need discussion.

Best,
Steffen

Edward Z. Yang (1):
connect.c: Support PuTTY plink and TortoisePlink as SSH on Windows

Johannes Schindelin (6):
MinGW: GCC >= 4 does not need SNPRINTF_SIZE_CORR anymore Quiet make: do not leave Windows behind
Work around a regression in Windows 7, causing erase_in_line() to
crash sometimes
test-chmtime: work around Windows limitation
winansi: fix compile warnings
Fix warnings in nedmalloc when compiling with GCC 4.4.0

Marius Storm-Olsen (2):
MinGW readdir reimplementation to support d_type Add custom memory allocator to MinGW and MacOS builds

Pat Thoyts (1):
git: browsing paths with spaces when using the start command

Steffen Prohaska (1):
MinGW: Teach Makefile to detect msysgit and apply specific settings

 Makefile                     |   26 +-
 compat/mingw.c               |   59 +
 compat/mingw.h               |   29 +
 compat/nedmalloc/License.txt |   23 +
 compat/nedmalloc/Readme.txt  |  136 +
 compat/nedmalloc/malloc.c.h  | 5752 ++++++++++++++++++++++++++++++++++++++++++
 compat/nedmalloc/nedmalloc.c |  966 +++++++
 compat/nedmalloc/nedmalloc.h |  180 ++
 compat/snprintf.c            |    4 +
 compat/win32.h               |    1 +
 compat/winansi.c             |    3 +-
 connect.c                    |    8 +-
 git-web--browse.sh           |    5 +-
 test-chmtime.c               |    9 +
 14 files changed, 7193 insertions(+), 8 deletions(-)
 create mode 100644 compat/nedmalloc/License.txt
 create mode 100644 compat/nedmalloc/Readme.txt
 create mode 100644 compat/nedmalloc/malloc.c.h
 create mode 100644 compat/nedmalloc/nedmalloc.c
 create mode 100644 compat/nedmalloc/nedmalloc.h
#1 lore raw

PATCH 1/11 MinGW: GCC >= 4 does not need SNPRINTF_SIZE_CORR anymore

From: Johannes Schindelin <[email protected]>

---
 Makefile          |    1 -
 compat/snprintf.c |    4 ++++
 2 files changed, 4 insertions(+), 1 deletions(-)
Makefile +0 -1
index eaae45d..77d8d9c 100644
@@ -845,7 +845,6 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	UNRELIABLE_FSTAT = UnfortunatelyYes
 	OBJECT_CREATION_USES_RENAMES = UnfortunatelyNeedsTo
 	COMPAT_CFLAGS += -D__USE_MINGW_ACCESS -DNOGDI -Icompat -Icompat/regex -Icompat/fnmatch
-	COMPAT_CFLAGS += -DSNPRINTF_SIZE_CORR=1
 	COMPAT_CFLAGS += -DSTRIP_EXTENSION=\".exe\"
 	COMPAT_OBJS += compat/mingw.o compat/fnmatch/fnmatch.o compat/regex/regex.o compat/winansi.o
 	EXTLIBS += -lws2_32
compat/snprintf.c +4 -0
index 357e733..6c0fb05 100644
@@ -6,8 +6,12 @@
  * number of characters to write without the trailing NUL.
  */
 #ifndef SNPRINTF_SIZE_CORR
+#if defined(__MINGW32__) && defined(__GNUC__) && __GNUC__ < 4
+#define SNPRINTF_SIZE_CORR 1
+#else
 #define SNPRINTF_SIZE_CORR 0
 #endif
+#endif
 
 #undef vsnprintf
 int git_vsnprintf(char *str, size_t maxsize, const char *format, va_list ap)
Signature · 1 line
1.6.3.1.54.g99dd
#2 lore raw

PATCH 2/11 Quiet make: do not leave Windows behind

From: Johannes Schindelin <[email protected]>

On Windows, we have to check whether there are scripts which would override .exe files, but this check missed the "quietification". Make now prints 'BUILTIN all' instead of a long chain of 'test || rm' commands.

[spr: added clarification what make will print. ]

---
 Makefile |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
Makefile +1 -1
index 77d8d9c..af08257 100644
@@ -1202,7 +1202,7 @@ SHELL = $(SHELL_PATH)
 
 all:: shell_compatibility_test $(ALL_PROGRAMS) $(BUILT_INS) $(OTHER_PROGRAMS) GIT-BUILD-OPTIONS
 ifneq (,$X)
-	$(foreach p,$(patsubst %$X,%,$(filter %$X,$(ALL_PROGRAMS) $(BUILT_INS) git$X)), test '$p' -ef '$p$X' || $(RM) '$p';)
+	$(QUIET_BUILT_IN)$(foreach p,$(patsubst %$X,%,$(filter %$X,$(ALL_PROGRAMS) $(BUILT_INS) git$X)), test '$p' -ef '$p$X' || $(RM) '$p';)
 endif
 
 all::
Signature · 1 line
1.6.3.1.54.g99dd
#3 lore raw

PATCH 3/11 Work around a regression in Windows 7, causing erase_in_line() to crash sometimes

From: Johannes Schindelin <[email protected]>

The function FillConsoleOutputCharacterA() was pretty content in XP to take a NULL pointer if we did not want to store the number of written columns. In Windows 7, it crashes, but only when called from within Git Bash, not from within cmd.exe. Go figure.

---
 compat/winansi.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
compat/winansi.c +2 -1
index 44dc293..4bee335 100644
@@ -80,6 +80,7 @@ static void set_console_attr(void)
 static void erase_in_line(void)
 {
 	CONSOLE_SCREEN_BUFFER_INFO sbi;
+	long dummy; /* Needed for Windows 7 (or Vista) regression */
 
 	if (!console)
 		return;
@@ -87,7 +88,7 @@ static void erase_in_line(void)
 	GetConsoleScreenBufferInfo(console, &sbi);
 	FillConsoleOutputCharacterA(console, ' ',
 		sbi.dwSize.X - sbi.dwCursorPosition.X, sbi.dwCursorPosition,
-		NULL);
+		&dummy);
 }
 
 
Signature · 1 line
1.6.3.1.54.g99dd
#4 lore raw

PATCH 4/11 test-chmtime: work around Windows limitation

From: Johannes Schindelin <[email protected]>

Windows has problems changing the mtime when the file is write protected, even by the owner of said file.

Add a Windows-only workaround to change the mode if necessary before trying to change the mtime.

---
 test-chmtime.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)
test-chmtime.c +9 -0
index d5358cb..fe476cb 100644
@@ -87,6 +87,15 @@ int main(int argc, const char *argv[])
 			return -1;
 		}
 
+#ifdef WIN32
+		if (!(sb.st_mode & S_IWUSR) &&
+				chmod(argv[i], sb.st_mode | S_IWUSR)) {
+			fprintf(stderr, "Could not make user-writable %s: %s",
+				argv[i], strerror(errno));
+			return -1;
+		}
+#endif
+
 		utb.actime = sb.st_atime;
 		utb.modtime = set_eq ? set_time : sb.st_mtime + set_time;
 
Signature · 1 line
1.6.3.1.54.g99dd
#5 lore raw

PATCH 5/11 winansi: fix compile warnings

From: Johannes Schindelin <[email protected]>

---
 compat/win32.h   |    1 +
 compat/winansi.c |    2 +-
 2 files changed, 2 insertions(+), 1 deletions(-)
compat/win32.h +1 -0
index c26384e..d531130 100644
@@ -1,5 +1,6 @@
 /* common Win32 functions for MinGW and Cygwin */
 #include <windows.h>
+#include <conio.h>
 
 static inline int file_attr_to_st_mode (DWORD attr)
 {
compat/winansi.c +1 -1
index 4bee335..9217c24 100644
@@ -80,7 +80,7 @@ static void set_console_attr(void)
 static void erase_in_line(void)
 {
 	CONSOLE_SCREEN_BUFFER_INFO sbi;
-	long dummy; /* Needed for Windows 7 (or Vista) regression */
+	DWORD dummy; /* Needed for Windows 7 (or Vista) regression */
 
 	if (!console)
 		return;
Signature · 1 line
1.6.3.1.54.g99dd
#6 lore raw

PATCH 6/11 git: browsing paths with spaces when using the start command

From: Pat Thoyts <[email protected]>

msysGit issue 258 tracks a problem opening a browser onto file paths that contain spaces or parentheses when calling the web--browse script. This patch modifies how the start command is called to solve this.

---
 git-web--browse.sh |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)
git-web--browse.sh +4 -1
index 7ed0fad..4f5c740 100755
@@ -161,9 +161,12 @@ case "$browser" in
 		;;
 	esac
 	;;
-    w3m|links|lynx|open|start)
+    w3m|links|lynx|open)
 	eval "$browser_path" "$@"
 	;;
+    start)
+        exec "$browser_path" '"web-browse"' "$@"
+        ;;
     dillo)
 	"$browser_path" "$@" &
 	;;
Signature · 1 line
1.6.3.1.54.g99dd
#7 lore raw

PATCH 7/11 connect.c: Support PuTTY plink and TortoisePlink as SSH on Windows

From: Edward Z. Yang <[email protected]>

OpenSSH uses -p to specify custom ports, while PuTTY plink and TortoisePlink use -P. Git now detects if plink is in GIT_SSH and modify its flags as necessary.

We call plink with -batch, so that it will error out with an error message instead of waiting for user input. As reported in msysGit issue 96, plink wants to interact with the user asking if a host key should be accepted, but this just blocks the terminal, since plink tries to get the answer from stdin. However, stdin is already connected to Git that wants to send input to the remote command.

But we do not pass -batch to TortoisePlink, because TortoisePlink uses a GUI to communicate with the user, and it does not understand -batch.

---
 connect.c |    8 ++++++--
 1 files changed, 6 insertions(+), 2 deletions(-)
connect.c +6 -2
index f6b8ba6..692d476 100644
@@ -602,14 +602,18 @@ struct child_process *git_connect(int fd[2], const char *url_orig,
 		die("command line too long");
 
 	conn->in = conn->out = -1;
-	conn->argv = arg = xcalloc(6, sizeof(*arg));
+	conn->argv = arg = xcalloc(7, sizeof(*arg));
 	if (protocol == PROTO_SSH) {
 		const char *ssh = getenv("GIT_SSH");
+		int putty = ssh && strcasestr(ssh, "plink");
 		if (!ssh) ssh = "ssh";
 
 		*arg++ = ssh;
+		if (putty && !strcasestr(ssh, "tortoiseplink"))
+			*arg++ = "-batch";
 		if (port) {
-			*arg++ = "-p";
+			/* P is for PuTTY, p is for OpenSSH */
+			*arg++ = putty ? "-P" : "-p";
 			*arg++ = port;
 		}
 		*arg++ = host;
Signature · 1 line
1.6.3.1.54.g99dd
#8 lore raw

PATCH 8/11 MinGW readdir reimplementation to support d_type

From: Marius Storm-Olsen <[email protected]>

The original readdir implementation was fast, but didn't support the d_type. This means that git would do additional lstats for each entry, to figure out if the entry was a directory or not. This unneedingly slowed down many operations, since Windows API provides this information directly when walking the directories.

By running this implementation on Moe's repo structure:
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 see the following speedups:
git add .
-------------------
old: 00:00:23(.087)
new: 00:00:21(.512) 1.07x

git status
-------------------
old: 00:00:03(.306)
new: 00:00:01(.684) 1.96x

git clean -dxf
-------------------
old: 00:00:01(.918)
new: 00:00:00(.295) 6.50x

---
 compat/mingw.c |   59 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 compat/mingw.h |   29 +++++++++++++++++++++++++++
 2 files changed, 88 insertions(+), 0 deletions(-)
compat/mingw.c +59 -0
index e190fdd..d85d680 100644
@@ -1171,3 +1171,62 @@ char *getpass(const char *prompt)
 	fputs("\n", stderr);
 	return strbuf_detach(&buf, NULL);
 }
+
+#ifndef NO_MINGW_REPLACE_READDIR
+/* MinGW readdir implementation to avoid extra lstats for Git */
+struct mingw_DIR
+{
+	struct _finddata_t	dd_dta;		/* disk transfer area for this dir */
+	struct mingw_dirent	dd_dir;		/* Our own implementation, including d_type */
+	long			dd_handle;	/* _findnext handle */
+	int			dd_stat; 	/* 0 = next entry to read is first entry, -1 = off the end, positive = 0 based index of next entry */
+	char			dd_name[1]; 	/* given path for dir with search pattern (struct is extended) */
+};
+
+struct dirent *mingw_readdir(DIR *dir)
+{
+	WIN32_FIND_DATAA buf;
+	HANDLE handle;
+	struct mingw_DIR *mdir = (struct mingw_DIR*)dir;
+
+	if (!dir->dd_handle) {
+		errno = EBADF; /* No set_errno for mingw */
+		return NULL;
+	}
+
+	if (dir->dd_handle == (long)INVALID_HANDLE_VALUE && dir->dd_stat == 0)
+	{
+		handle = FindFirstFileA(dir->dd_name, &buf);
+		DWORD lasterr = GetLastError();
+		dir->dd_handle = (long)handle;
+		if (handle == INVALID_HANDLE_VALUE && (lasterr != ERROR_NO_MORE_FILES)) {
+			errno = err_win_to_posix(lasterr);
+			return NULL;
+		}
+	} else if (dir->dd_handle == (long)INVALID_HANDLE_VALUE) {
+		return NULL;
+	} else if (!FindNextFileA((HANDLE)dir->dd_handle, &buf)) {
+		DWORD lasterr = GetLastError();
+		FindClose((HANDLE)dir->dd_handle);
+		dir->dd_handle = (long)INVALID_HANDLE_VALUE;
+		/* POSIX says you shouldn't set errno when readdir can't
+		   find any more files; so, if another error we leave it set. */
+		if (lasterr != ERROR_NO_MORE_FILES)
+			errno = err_win_to_posix(lasterr);
+		return NULL;
+	}
+
+	/* We get here if `buf' contains valid data.  */
+	strcpy(dir->dd_dir.d_name, buf.cFileName);
+	++dir->dd_stat;
+
+	/* Set file type, based on WIN32_FIND_DATA */
+	mdir->dd_dir.d_type = 0;
+	if (buf.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
+		mdir->dd_dir.d_type |= DT_DIR;
+	else
+		mdir->dd_dir.d_type |= DT_REG;
+
+	return (struct dirent*)&dir->dd_dir;
+}
+#endif // !NO_MINGW_REPLACE_READDIR
compat/mingw.h +29 -0
index 4c50f5b..4f7ba4c 100644
@@ -235,3 +235,32 @@ int main(int argc, const char **argv) \
 	return mingw_main(argc, argv); \
 } \
 static int mingw_main(c,v)
+
+#ifndef NO_MINGW_REPLACE_READDIR
+/*
+ * A replacement of readdir, to ensure that it reads the file type at
+ * the same time. This avoid extra unneeded lstats in git on MinGW
+ */
+#undef DT_UNKNOWN
+#undef DT_DIR
+#undef DT_REG
+#undef DT_LNK
+#define DT_UNKNOWN	0
+#define DT_DIR		1
+#define DT_REG		2
+#define DT_LNK		3
+
+struct mingw_dirent
+{
+	long		d_ino;			/* Always zero. */
+	union {
+		unsigned short	d_reclen;	/* Always zero. */
+		unsigned char   d_type;		/* Reimplementation adds this */
+	};
+	unsigned short	d_namlen;		/* Length of name in d_name. */
+	char		d_name[FILENAME_MAX];	/* File name. */
+};
+#define dirent mingw_dirent
+#define readdir(x) mingw_readdir(x)
+struct dirent *mingw_readdir(DIR *dir);
+#endif // !NO_MINGW_REPLACE_READDIR
Signature · 1 line
1.6.3.1.54.g99dd

A message that is not in the archive (sent off-list or not yet received): [email protected]

#10 lore raw

PATCH 10/11 Fix warnings in nedmalloc when compiling with GCC 4.4.0

From: Johannes Schindelin <[email protected]>

---
 compat/nedmalloc/malloc.c.h  |    4 +++-
 compat/nedmalloc/nedmalloc.c |    2 +-
 2 files changed, 4 insertions(+), 2 deletions(-)
compat/nedmalloc/malloc.c.h +3 -1
index 1d9bbe0..678beb8 100644
@@ -1270,7 +1270,9 @@ int mspace_mallopt(int, int);
 /*------------------------------ internal #includes ---------------------- */
 
 #ifdef WIN32
+#ifndef __GNUC__
 #pragma warning( disable : 4146 ) /* no "unsigned" warnings */
+#endif
 #endif /* WIN32 */
 
 #include <stdio.h>       /* for printing in malloc_stats */
@@ -2541,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() (mparams.magic != 0 || init_mparams())
+#define ensure_initialization() if (mparams.magic == 0) init_mparams()
 
 #if !ONLY_MSPACES
 
compat/nedmalloc/nedmalloc.c +1 -1
index a381a7d..60a4093 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif
 #define MSPACES 1
Signature · 1 line
1.6.3.1.54.g99dd
#11 lore raw

PATCH 11/11 MinGW: Teach Makefile to detect msysgit and apply specific settings

This commit changes handling of the msysgit specific settings, so that they can be applied to official git.git. Some msysgit settings differ from the standard MinGW settings. We move them into an ifndef block that is only evaluated if a file THIS_IS_MSYSGIT is present in the parent directory, which is the case for an msysgit working environment. The tag file is unlikely to be present accidentally.

---
 Makefile |   14 ++++++++++++--
 1 files changed, 12 insertions(+), 2 deletions(-)
Makefile +12 -2
index f9657ce..7803ae7 100644
@@ -823,7 +823,6 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	pathsep = ;
 	NO_PREAD = YesPlease
 	NO_OPENSSL = YesPlease
-	NO_CURL = YesPlease
 	NO_SYMLINK_HEAD = YesPlease
 	NO_IPV6 = YesPlease
 	NO_SETENV = YesPlease
@@ -831,7 +830,6 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	NO_STRCASESTR = YesPlease
 	NO_STRLCPY = YesPlease
 	NO_MEMMEM = YesPlease
-	NO_PTHREADS = YesPlease
 	NEEDS_LIBICONV = YesPlease
 	OLD_ICONV = YesPlease
 	NO_C99_FORMAT = YesPlease
@@ -853,6 +851,18 @@ ifneq (,$(findstring MINGW,$(uname_S)))
 	COMPAT_OBJS += compat/mingw.o compat/fnmatch/fnmatch.o compat/regex/regex.o compat/winansi.o
 	EXTLIBS += -lws2_32
 	X = .exe
+ifneq (,$(wildcard ../THIS_IS_MSYSGIT))
+	htmldir=doc/git/html/
+	prefix =
+	INSTALL = /bin/install
+	EXTLIBS += /mingw/lib/libz.a
+	NO_R_TO_GCC_LINKER = YesPlease
+	INTERNAL_QSORT = YesPlease
+	THREADED_DELTA_SEARCH = YesPlease
+else
+	NO_CURL = YesPlease
+	NO_PTHREADS = YesPlease
+endif
 endif
 ifneq (,$(findstring arm,$(uname_M)))
 	ARM_SHA1 = YesPlease
Signature · 1 line
1.6.3.1.54.g99dd
#12 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

In reply to Steffen Prohaska

compat/nedmalloc/malloc.c.h struct malloc_params { · line 2548

Quoted 8 lines · Steffen Prohaska
compat/nedmalloc/malloc.c.h +1 -1
@@ -2541,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() (mparams.magic != 0 || init_mparams())
+#define ensure_initialization() if (mparams.magic == 0) init_mparams()
 
 #if !ONLY_MSPACES

The code after the patch looks more fragile than the original. I know there currently is no code like:

	if (foo())
        	ensure_initialization();
	else
        	warn("oops");

but this change still feels wrong.

What issue is this patch trying to work around? Returned value not being used?

compat/nedmalloc/nedmalloc.c DEALINGS IN THE SOFTWARE. · line 39

Quoted 12 lines
compat/nedmalloc/nedmalloc.c +1 -1
index a381a7d..60a4093 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif

Can somebody enlighten me what this hunk is about, and how it helps GCC 4.4?

There are many "#if[n]def __MINGW32__" remaining in the codebase both inside and outside compat/ area, so it is not that that symbol is somehow special. I cannot even tell which one of the following is closer to the reason behind this change:

(1) "Because tacking '&& !defined(__MINGW32__)' after defined(WIN32) is unnecessary for such and such reasons, it is removed"; or

(2) "Because tacking '&& !defined(__MINGW32__)' after defined(WIN32) is harmful for such and such reasons, it is removed".

Puzzled.

#13 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Hi,

Quoted 14 lines · Junio C Hamano

compat/nedmalloc/malloc.c.h struct malloc_params { · line 2548

Quoted 8 lines · Steffen Prohaska
compat/nedmalloc/malloc.c.h +1 -1
@@ -2541,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() (mparams.magic != 0 || init_mparams())
+#define ensure_initialization() if (mparams.magic == 0) init_mparams()
 
 #if !ONLY_MSPACES

The code after the patch looks more fragile than the original. I know there currently is no code like:

Quoted 6 lines
	if (foo())
        	ensure_initialization();
	else
        	warn("oops");

but this change still feels wrong.

I know, but the whole use of ensure_initialization() feels wrong to me, and I did _not_ want to change the code too much, lest we end up maintaining a proper fork as has happened with libxdiff.

Quoted 2 lines

What issue is this patch trying to work around? Returned value not being used?

I forgot what the GCC warning looked like, and have only text-mode access to the web right now, so I cannot access any Windows machine and try again.

It was a warning, though, that much I remember.

Quoted 7 lines
Quoted 7 lines
compat/nedmalloc/nedmalloc.c +0 -0
index a381a7d..60a4093 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
Quoted 8 lines

compat/nedmalloc/nedmalloc.c DEALINGS IN THE SOFTWARE. · line 39

Quoted 5 lines
compat/nedmalloc/nedmalloc.c +1 -1
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif

Can somebody enlighten me what this hunk is about, and how it helps GCC 4.4?

It helps in that malloc.h is included even if we happen to compile the stuff as a MinGW program. Otherwise necessary function declarations are missing.

Quoted 10 lines

There are many "#if[n]def __MINGW32__" remaining in the codebase both inside and outside compat/ area, so it is not that that symbol is somehow special. I cannot even tell which one of the following is closer to the reason behind this change:

(1) "Because tacking '&& !defined(__MINGW32__)' after defined(WIN32) is unnecessary for such and such reasons, it is removed"; or

(2) "Because tacking '&& !defined(__MINGW32__)' after defined(WIN32) is harmful for such and such reasons, it is removed".

The latter.

Quoted 1 line

Puzzled.

Hopefully less so, now.

Ciao,
Dscho

#14 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 16 lines · Johannes Schindelin
Quoted 7 lines
Quoted 7 lines
compat/nedmalloc/nedmalloc.c +0 -0
index a381a7d..60a4093 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
Quoted 8 lines

compat/nedmalloc/nedmalloc.c DEALINGS IN THE SOFTWARE. · line 39

Quoted 5 lines
compat/nedmalloc/nedmalloc.c +1 -1
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif

Can somebody enlighten me what this hunk is about, and how it helps GCC 4.4?

Quoted 5 lines

It helps in that malloc.h is included even if we happen to compile the stuff as a MinGW program. Otherwise necessary function declarations are missing.
...
Hopefully less so, now.

I wish enough information were in the commit log message from the beginning.

#15 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 3 lines · Junio C Hamano
Quoted 3 lines
Quoted 2 lines

Can somebody enlighten me what this hunk is about, and how it helps GCC 4.4?

Um... It's strange.
I can not connect to http://gcc.gnu.org website
to get GCC 4.4 manual(pdf format) currently.
ping is normal just.

[invain@fedora9 invain]$ ping gcc.gnu.org
PING gcc.gnu.org (209.132.176.174) 56(84) bytes of data.
64 bytes from sourceware.org (209.132.176.174): icmp_seq=1 ttl=40 time=194 ms 64 bytes from sourceware.org (209.132.176.174): icmp_seq=2 ttl=40 time=199 ms ^X64 bytes from sourceware.org (209.132.176.174): icmp_seq=3 ttl=40 time=193 ms 64 bytes from sourceware.org (209.132.176.174): icmp_seq=4 ttl=40 time=203 ms ^C
--- gcc.gnu.org ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3326ms
rtt min/avg/max/mdev = 193.762/198.126/203.978/4.135 ms [invain@fedora9 invain]$
[invain@fedora9 invain]$ nmap gcc.gnu.org
[invain@fedora9 invain]$ nmap -p 80 gcc.gnu.org

Starting Nmap 4.53 ( http://insecure.org ) at 2009-06-02 08:37 KST
Note: Host seems down. If it is really up, but blocking our ping probes, try -PN Nmap done: 1 IP address (0 hosts up) scanned in 2.023 seconds [invain@fedora9 invain]$

Signature · 5 lines
Regards,
GeunSik Lim ( Samsung Electronics )
Blog : http://blog.naver.com/invain/
e-Mail: [email protected]
           [email protected] , [email protected]
#16 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

In reply to Junio C Hamano

Hi,

Quoted 18 lines · Junio C Hamano
Quoted 16 lines · Johannes Schindelin
Quoted 15 lines

compat/nedmalloc/nedmalloc.c DEALINGS IN THE SOFTWARE. · line 39

Quoted 12 lines
compat/nedmalloc/nedmalloc.c +1 -1
index a381a7d..60a4093 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif

Can somebody enlighten me what this hunk is about, and how it helps GCC 4.4?

Quoted 8 lines
Quoted 5 lines

It helps in that malloc.h is included even if we happen to compile the stuff as a MinGW program. Otherwise necessary function declarations are missing.
...
Hopefully less so, now.

I wish enough information were in the commit log message from the beginning.

Okay, how about this in the commit body in addition?

-- snip --
Nedmalloc's source code has a cute #define construct to avoid inserting an if() statement, because that might interact badly with enclosing if() statements. However, there is no danger of that now, but GCC > 4 complains "warning: value computed is not used".

Also, with the version of MinGW's headers in msysGit, we need to include malloc.h lest the compiler complain about an "incompatible implicit declaration of built-in function 'alloca'".
-- snap --

Hmm?
Dscho

#17 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 8 lines · Johannes Schindelin

Okay, how about this in the commit body in addition?

-- snip --
Nedmalloc's source code has a cute #define construct to avoid inserting an if() statement, because that might interact badly with enclosing if() statements. However, there is no danger of that now, but GCC > 4 complains "warning: value computed is not used".

Quoted 6 lines

Also, with the version of MinGW's headers in msysGit, we need to include malloc.h lest the compiler complain about an "incompatible implicit declaration of built-in function 'alloca'".
-- snap --

Hmm?

As to the first one, I think your problem description is clear (the problem is clear without being explained to begin with) but "there is no danger of that now" is somewhere between a lame excuse for not doing it the right way and a bug waiting to happen. If "value computed is not used" is the issue, why doesn't this work?

    #define ensure_initialization() (void)(mparams.magic != 0 || init_mparams())

or at least wrap your fragile "if" in a bog-standard do {} while(0), like...

    #define ensure_initialization() \
    	do { if (mparams.magic == 0) init_mparams(); } while (0)

Then you do not have to make excuses, and instead can just say

Squelch GCC > 4's "value computed not used" warnings from ensure_initialization() macro definition.

As to the second one, I think you lost "even if we happen to" part (that logically corresponds to "... that is why we do not want to limit the inclusion to !__MINGW32__ case") from the description, making it less readable...

#18 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Hi,

Quoted 17 lines · Junio C Hamano
Quoted 8 lines · Johannes Schindelin

Okay, how about this in the commit body in addition?

-- snip --
Nedmalloc's source code has a cute #define construct to avoid inserting an if() statement, because that might interact badly with enclosing if() statements. However, there is no danger of that now, but GCC > 4 complains "warning: value computed is not used".

Quoted 6 lines

Also, with the version of MinGW's headers in msysGit, we need to include malloc.h lest the compiler complain about an "incompatible implicit declaration of built-in function 'alloca'".
-- snap --

Hmm?

Quoted 7 lines

As to the first one, I think your problem description is clear (the problem is clear without being explained to begin with) but "there is no danger of that now" is somewhere between a lame excuse for not doing it the right way and a bug waiting to happen. If "value computed is not used" is the issue, why doesn't this work?

    #define ensure_initialization() (void)(mparams.magic != 0 || init_mparams())

That works.

Quoted 4 lines

As to the second one, I think you lost "even if we happen to" part (that logically corresponds to "... that is why we do not want to limit the inclusion to !__MINGW32__ case") from the description, making it less readable...

You mean "even if we happen to be on Windows"? Fine with me. Want me to resend a fixed patch?

Ciao,
Dscho

#19 lore raw

Re: [PATCH 10/11] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 6 lines · Johannes Schindelin
Quoted 4 lines

As to the second one, I think you lost "even if we happen to" part (that logically corresponds to "... that is why we do not want to limit the inclusion to !__MINGW32__ case") from the description, making it less readable...

You mean "even if we happen to be on Windows"?

I meant this part from your earlier message.

Quoted 3 lines

It helps in that malloc.h is included even if we happen to compile the stuff as a MinGW program. Otherwise necessary function declarations are missing.

compared with the newer "how about" version, which is

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Also, with the version of MinGW's headers in msysGit, we need to include malloc.h lest the compiler complain about an "incompatible implicit declaration of built-in function 'alloca'".

I thought the former explains the change in question

Quoted 2 lines
Quoted 2 lines

Review

Quoted 2 lines
(quoted diff) +1 -1
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)

much more clearly: "If you are compiling for Windows, regardless of MINGW32, you would want this section to apply".

Quoted 1 line

... Want me to resend a fixed patch?

Surely. I think I could come up with something based on the discussion here, but I'd rather not. A patch signed off by either you or Steffen, or both is very much appreciated.

#20 lore raw

PATCH v2 10/11 Fix warnings in nedmalloc when compiling with GCC 4.4.0

Nedmalloc's source code has a cute #define construct to avoid inserting an if() statement, because that might interact badly with enclosing if() statements. However, GCC > 4 complains with a "warning: value computed is not used". So we cast the result to "void".

GCC also does not understand the Visual C++ specific pragmas, so we need to disable them for MinGW.

We need to include malloc.h on Windows even if we happen to compile the stuff as a MinGW program. Otherwise the function declaration of alloca() is missing.

---

Forgot to send.

 compat/nedmalloc/malloc.c.h  |    4 +++-
 compat/nedmalloc/nedmalloc.c |    2 +-
 2 files changed, 4 insertions(+), 2 deletions(-)
compat/nedmalloc/malloc.c.h +3 -1
index bb0f482..b5b1495 100644
@@ -1270,7 +1270,9 @@ int mspace_mallopt(int, int);
 /*------------------------------ internal #includes ---------------------- */
 
 #ifdef WIN32
+#ifndef __GNUC__
 #pragma warning( disable : 4146 ) /* no "unsigned" warnings */
+#endif
 #endif /* WIN32 */
 
 #include <stdio.h>       /* for printing in malloc_stats */
@@ -2541,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() (mparams.magic != 0 || init_mparams())
+#define ensure_initialization() ((void)(mparams.magic == 0 || init_mparams()))
 
 #if !ONLY_MSPACES
 
compat/nedmalloc/nedmalloc.c +1 -1
index 41a3234..d9a17a8 100644
@@ -34,7 +34,7 @@ DEALINGS IN THE SOFTWARE.
 /*#define FULLSANITYCHECKS*/
 
 #include "nedmalloc.h"
-#if defined(WIN32) && !defined(__MINGW32__)
+#if defined(WIN32)
  #include <malloc.h>
 #endif
 #define MSPACES 1
Signature · 1 line
1.6.3.284.g6fecc
#21 lore raw

Re: [PATCH 10/11 v2] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 6 lines · Johannes Schindelin

Nedmalloc's source code has a cute #define construct to avoid inserting an if() statement, because that might interact badly with enclosing if() statements. However, GCC > 4 complains with a "warning: value computed is not used". So we cast the result to "void".

Forgot to send.

Thanks. The description on the pragma is a good thing to add, which you did.

Is Visual C++ happy with casting discarded result to (void)? I'd find it broken if it isn't, but it would not hurt to ask just to make sure.

Steffen, can we move the series, with this patch replaced, to 'next' now? I think it looks good (thanks!) but again it would not hurt to ask just to make sure ;-).

Quoted 27 lines
 compat/nedmalloc/malloc.c.h  |    4 +++-
 compat/nedmalloc/nedmalloc.c |    2 +-
 2 files changed, 4 insertions(+), 2 deletions(-)
compat/nedmalloc/malloc.c.h +3 -1
index bb0f482..b5b1495 100644
@@ -1270,7 +1270,9 @@ int mspace_mallopt(int, int);
 /*------------------------------ internal #includes ---------------------- */
 
 #ifdef WIN32
+#ifndef __GNUC__
 #pragma warning( disable : 4146 ) /* no "unsigned" warnings */
+#endif
 #endif /* WIN32 */
 
 #include <stdio.h>       /* for printing in malloc_stats */
@@ -2541,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() (mparams.magic != 0 || init_mparams())
+#define ensure_initialization() ((void)(mparams.magic == 0 || init_mparams()))
 
 #if !ONLY_MSPACES
 
#22 lore raw

Re: [PATCH 10/11 v2] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 2 lines · Junio C Hamano

Is Visual C++ happy with casting discarded result to (void)? I'd find it broken if it isn't, but it would not hurt to ask just to make sure.

It is OK. Visual C++ is not *that* broken ;)

-- Hannes

#23 lore raw

Re: [PATCH 10/11 v2] Fix warnings in nedmalloc when compiling with GCC 4.4.0

In reply to Junio C Hamano
Quoted 3 lines · Junio C Hamano

Steffen, can we move the series, with this patch replaced, to 'next' now? I think it looks good (thanks!) but again it would not hurt to ask just to make sure ;-).

I agree, it looks good. Thanks.

Steffen

#24 lore raw

Re: [PATCH 10/11 v2] Fix warnings in nedmalloc when compiling with GCC 4.4.0

Quoted 9 lines · Steffen Prohaska
Quoted 4 lines · Junio C Hamano

Steffen, can we move the series, with this patch replaced, to 'next' now? I think it looks good (thanks!) but again it would not hurt to ask just to
make sure ;-).

I agree, it looks good. Thanks.

Please don't advance this series to master just yet: I see severe breakage with the nedmalloc patch on one of my systems, but I ran out of time to investigate further. (And I'm afraid I can continue only next Monday.)

-- Hannes

#25 lore raw

PATCH 1/1 Fix typo in nedmalloc warning fix

---

Quoted 3 lines · Johannes Sixt

Please don't advance this series to master just yet: I see severe breakage with the nedmalloc patch on one of my systems, but I ran out of time to investigate further. (And I'm afraid I can continue only next Monday.)

So, I have the failure on my other system as well, and this is the fix ;)

-- Hannes

 compat/nedmalloc/malloc.c.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
compat/nedmalloc/malloc.c.h +1 -1
index b5b1495..74c42e3 100644
@@ -2543,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() ((void)(mparams.magic == 0 || init_mparams()))
+#define ensure_initialization() ((void)(mparams.magic != 0 || init_mparams()))
 
 #if !ONLY_MSPACES
 
Signature · 1 line
1.6.3.2.1225.g177fa.dirty
#26 lore raw

PATCH 1/1 compat/ has subdirectories: do not omit them in 'make clean'

---

Quoted 1 line · Johannes Sixt

So, I have the failure on my other system as well, and this is the fix ;)

And this was quite helpful to find the breakage.

-- Hannes

 Makefile |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
Makefile +1 -1
index a38dcdd..7002688 100644
@@ -1691,7 +1691,7 @@ distclean: clean
 	$(RM) configure
 
 clean:
-	$(RM) *.o mozilla-sha1/*.o arm/*.o ppc/*.o compat/*.o xdiff/*.o \
+	$(RM) *.o mozilla-sha1/*.o arm/*.o ppc/*.o compat/*.o compat/*/*.o xdiff/*.o \
 		$(LIB_FILE) $(XDIFF_LIB)
 	$(RM) $(ALL_PROGRAMS) $(BUILT_INS) git$X
 	$(RM) $(TEST_PROGRAMS)
Signature · 1 line
1.6.3.2.1225.g177fa.dirty
#27 lore raw

Re: [PATCH] Fix typo in nedmalloc warning fix

In reply to Johannes Sixt

Hi,

Quoted 19 lines · Johannes Sixt

---

Quoted 3 lines · Johannes Sixt

Please don't advance this series to master just yet: I see severe breakage with the nedmalloc patch on one of my systems, but I ran out of time to investigate further. (And I'm afraid I can continue only next Monday.)

So, I have the failure on my other system as well, and this is the fix ;)

-- Hannes

 compat/nedmalloc/malloc.c.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
compat/nedmalloc/malloc.c.h +0 -0
index b5b1495..74c42e3 100644

compat/nedmalloc/malloc.c.h struct malloc_params { · line 2548

Quoted 7 lines
compat/nedmalloc/malloc.c.h +1 -1
@@ -2543,7 +2543,7 @@ struct malloc_params {
 static struct malloc_params mparams;
 
 /* Ensure mparams initialized */
-#define ensure_initialization() ((void)(mparams.magic == 0 ||
 init_mparams()))
+#define ensure_initialization() ((void)(mparams.magic != 0 ||
 init_mparams()))

Aargh!

Thanks for catching this stupid bug,
Dscho

#28 lore raw

Re: [PATCH 05/11] winansi: fix compile warnings

In reply to Steffen Prohaska

compat/win32.h line 6

Quoted 11 lines · Steffen Prohaska
compat/win32.h +1 -0
index c26384e..d531130 100644
@@ -1,5 +1,6 @@
 /* common Win32 functions for MinGW and Cygwin */
 #include <windows.h>
+#include <conio.h>
 
 static inline int file_attr_to_st_mode (DWORD attr)
 {

I assume this is to remove the warning about missing declaration of _getch(). Can we have this patch instead? I don't have conio.h in my oldish MinGW environment.

compat/mingw.c +3 -0
index 52961ee..53053ad 100644
@@ -1211,6 +1211,9 @@ int link(const char *oldpath, const char *newpath)
 	return 0;
 }
 
+/* from conio.h */
+int _getch(void);  // FIXME: really look this up in conio.h!!!!
+
 char *getpass(const char *prompt)
 {
 	struct strbuf buf = STRBUF_INIT;

compat/winansi.c static void set_console_attr(void) · line 86

Quoted 13 lines
compat/winansi.c +1 -1
index 4bee335..9217c24 100644
@@ -80,7 +80,7 @@ static void set_console_attr(void)
 static void erase_in_line(void)
 {
 	CONSOLE_SCREEN_BUFFER_INFO sbi;
-	long dummy; /* Needed for Windows 7 (or Vista) regression */
+	DWORD dummy; /* Needed for Windows 7 (or Vista) regression */
 
 	if (!console)
 		return;

This should really be squashed into 03/11, where the line is introduced.

-- Hannes

#29 lore raw

PATCH v2 3/11 Work around a regression in Windows 7, causing erase_in_line() to crash sometimes

From: Johannes Schindelin <[email protected]>

The function FillConsoleOutputCharacterA() was pretty content in XP to take a NULL pointer if we did not want to store the number of written columns. In Windows 7, it crashes, but only when called from within Git Bash, not from within cmd.exe. Go figure.

---
 compat/winansi.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

I squashed the second chunk of 05/11, as suggested by Hannes.

compat/winansi.c +2 -1
index 44dc293..9217c24 100644
@@ -80,6 +80,7 @@ static void set_console_attr(void)
 static void erase_in_line(void)
 {
 	CONSOLE_SCREEN_BUFFER_INFO sbi;
+	DWORD dummy; /* Needed for Windows 7 (or Vista) regression */
 
 	if (!console)
 		return;
@@ -87,7 +88,7 @@ static void erase_in_line(void)
 	GetConsoleScreenBufferInfo(console, &sbi);
 	FillConsoleOutputCharacterA(console, ' ',
 		sbi.dwSize.X - sbi.dwCursorPosition.X, sbi.dwCursorPosition,
-		NULL);
+		&dummy);
 }
 
 
Signature · 1 line
1.6.3.1.54.g99dd
#30 lore raw

Re: [PATCH 03/11 v2] Work around a regression in Windows 7, causing erase_in_line() to crash sometimes

Hi,

Quoted 6 lines · Steffen Prohaska

From: Johannes Schindelin <[email protected]>

The function FillConsoleOutputCharacterA() was pretty content in XP to take a NULL pointer if we did not want to store the number of written columns. In Windows 7, it crashes, but only when called from within Git Bash, not from within cmd.exe. Go figure.

Did I really have too-long lines in my commit message?

Quoted 8 lines
---
 compat/winansi.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

I squashed the second chunk of 05/11, as suggested by Hannes.

Makes sense, thank you!
Dscho

#31 lore raw

Re: [PATCH 03/11 v2] Work around a regression in Windows 7, causing erase_in_line() to crash sometimes

Quoted 4 lines · Johannes Schindelin

Hi,

On Mon, 1 Jun 2009, Steffen Prohaska wrote:

Quoted 8 lines
Quoted 6 lines

From: Johannes Schindelin <[email protected]>

The function FillConsoleOutputCharacterA() was pretty content in XP to take a NULL pointer if we did not want to store the number of written columns. In Windows 7, it crashes, but only when called from within Git Bash, not from within cmd.exe. Go figure.

Did I really have too-long lines in my commit message?

Yes. See aa3abfc9852a4d4cfaa7f0042102eb56ed2e0daa on 4msysgit's devel. You are author and committer.

Steffen

#32 lore raw

Re: [PATCH 03/11 v2] Work around a regression in Windows 7, causing erase_in_line() to crash sometimes

Hi,

Quoted 14 lines · Steffen Prohaska
Quoted 12 lines · Johannes Schindelin
Quoted 9 lines · Steffen Prohaska

From: Johannes Schindelin <[email protected]>

The function FillConsoleOutputCharacterA() was pretty content in XP to take a NULL
pointer if we did not want to store the number of written columns. In Windows 7,
it crashes, but only when called from within Git Bash, not from within cmd.exe.
Go figure.

Quoted 5 lines
Quoted 1 line

Did I really have too-long lines in my commit message?

Yes. See aa3abfc9852a4d4cfaa7f0042102eb56ed2e0daa on 4msysgit's devel. You are author and committer.

My bad, then!

Ciao,
Dscho

#33 lore raw

Re: [PATCH 05/11] winansi: fix compile warnings

In reply to Johannes Sixt
Quoted 22 lines · Johannes Sixt

compat/win32.h line 6

Quoted 11 lines · Steffen Prohaska
compat/win32.h +1 -0
index c26384e..d531130 100644
@@ -1,5 +1,6 @@
 /* common Win32 functions for MinGW and Cygwin */
 #include <windows.h>
+#include <conio.h>
 
 static inline int file_attr_to_st_mode (DWORD attr)
 {

I assume this is to remove the warning about missing declaration of _getch(). Can we have this patch instead? I don't have conio.h in my oldish MinGW environment.

compat/mingw.c +0 -0
index 52961ee..53053ad 100644

compat/mingw.c int link(const char *oldpath, const char *newpath) · line 1219

Quoted 9 lines
compat/mingw.c +3 -0
@@ -1211,6 +1211,9 @@ int link(const char *oldpath, const char *newpath)
  	return 0;
  }
 
+/* from conio.h */
+int _getch(void);  // FIXME: really look this up in conio.h!!!!
+
  char *getpass(const char *prompt)
  {
  	struct strbuf buf = STRBUF_INIT;

Ok, forget this patch; I'll upgrade my MinGW instead, and let's assume all other mingw.git h4ckrz who aren't using the latest msysgit will upgrade as well.

Nevertheless, I think that the #include <conio.h> is in the wrong file: it should be included from compat/mingw.c.

-- Hannes

#34 lore raw

PATCH v2 5/11 MinGW: fix warning about implicit declaration of _getch()

From: Johannes Schindelin <[email protected]>

conio.h provides the declaration.

---
 compat/mingw.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

This replaces 05/11. The original commit messages was misleading. Apparently, the original 05/11 solved two things. First, it fixed a warning in winansi.c, which has been squashed into 03/11. Second, it fixed a warning about implict decl of _getch(). Including conio.h in mingw.c is sufficient to fix this warning.

compat/mingw.c +1 -0
index e190fdd..12d0c2f 100644
@@ -1,5 +1,6 @@
 #include "../git-compat-util.h"
 #include "win32.h"
+#include <conio.h>
 #include "../strbuf.h"
 
 unsigned int _CRT_fmode = _O_BINARY;
Signature · 1 line
1.6.3.1.54.g99dd
#35 lore raw

Re: [PATCH 04/11] test-chmtime: work around Windows limitation

In reply to Steffen Prohaska
Quoted 7 lines · Steffen Prohaska

From: Johannes Schindelin <[email protected]>

Windows has problems changing the mtime when the file is write protected, even by the owner of said file.

Add a Windows-only workaround to change the mode if necessary before trying to change the mtime.

I'll queue this as-is for now, but shouldn't the code remember the original mode and change it back?

Quoted 28 lines
---
 test-chmtime.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)
test-chmtime.c +9 -0
index d5358cb..fe476cb 100644
@@ -87,6 +87,15 @@ int main(int argc, const char *argv[])
 			return -1;
 		}
 
+#ifdef WIN32
+		if (!(sb.st_mode & S_IWUSR) &&
+				chmod(argv[i], sb.st_mode | S_IWUSR)) {
+			fprintf(stderr, "Could not make user-writable %s: %s",
+				argv[i], strerror(errno));
+			return -1;
+		}
+#endif
+
 		utb.actime = sb.st_atime;
 		utb.modtime = set_eq ? set_time : sb.st_mtime + set_time;
 
Signature · 1 line
1.6.3.1.54.g99dd
#36 lore raw

Re: [PATCH 04/11] test-chmtime: work around Windows limitation

Hi,

Quoted 4 lines · Junio C Hamano
Quoted 2 lines · Steffen Prohaska

From: Johannes Schindelin <[email protected]>

Quoted 8 lines
Quoted 5 lines

Windows has problems changing the mtime when the file is write protected, even by the owner of said file.

Add a Windows-only workaround to change the mode if necessary before trying to change the mtime.

I'll queue this as-is for now, but shouldn't the code remember the original mode and change it back?

Hmm, good point. As I mentioned in another mail, though, I cannot access any Windows machine to fix this issue now, sorry.

Not changing the mode back _might_ paper over another issue: if the files are deleted later, that mode change might mean that the deletion now succeeds, and fixing test-chmtime may make it fail again. Not sure, though.

Ciao,
Dscho

#37 lore raw
In reply to Steffen Prohaska
Quoted 7 lines · Steffen Prohaska

Hello,
Here are more patches that are in msysgit but not yet in official git.git. The patches have been discussed in the following two threads:

http://thread.gmane.org/gmane.comp.version-control.msysgit/5373
http://thread.gmane.org/gmane.comp.version-control.msysgit/4876

Thanks a lot, your work is very much appreciated. Let me add that these patches have been in use for weeks, some even for months now.

Please consider to resend updated 03/11 and 05/11 as I have suggested elsewhere in the thread.

-- Hannes