Skip to contentSkip to conversation

replace O_SYNC with O_FSYNC

PATCH In review m a 8 participants 19 messages Jun 20, 2009 – Jun 24, 2009

lore
Share & tools
lore

m a

author
#0 lore raw

PATCH 1/1 replace O_SYNC with O_FSYNC

This patch replaces O_SYNC with O_FSYNC. These two flags do the same thing, but only O_FSYNC is available in Mac OS 10.3 and under. It only replaces O_SYNC if it doesn't exist. This patch allows the file block-raw-posix.c to compile on Mac OS 10.3. This is my first time submitting a patch, so there might have been a few mistakes made.

block-raw-posix (updated).c renamed from block-raw-posix.c +6 -0
@@ -73,6 +73,12 @@
 #define DEBUG_BLOCK_PRINT(formatCstr, args...)
 #endif
 
+// O_SYNC isn't available on Mac OS 10.3 and under
+// O_SYNC and O_FSYNC do the same thing
+#ifndef O_SYNC
+#define O_SYNC O_FSYNC
+#endif
+
 /* OS X does not have O_DSYNC */
 #ifndef O_DSYNC
 #define O_DSYNC O_SYNC
#1 lore raw

Hi,

Quoted 1 line · m a

This patch replaces O_SYNC with O_FSYNC. These two flags do the same thing, but only O_FSYNC is available in Mac OS 10.3 and under. It only replaces O_SYNC if it doesn't exist. This patch allows the file block-raw-posix.c to compile on Mac OS 10.3. This is my first time submitting a patch, so there might have been a few mistakes made.

Please include a Signed-off-by line and use C-style comments instead of C++ style comments. Otherwise, looks good.

Regards,

Anthony Liguori

Quoted 19 lines
block-raw-posix (updated).c Fri Jun 19 22:01:07 2009 renamed from block-raw-posix.c Wed May 20 16:46:58 2009 +6 -0
@@ -73,6 +73,12 @@
 #define DEBUG_BLOCK_PRINT(formatCstr, args...)
 #endif
 
+// O_SYNC isn't available on Mac OS 10.3 and under
+// O_SYNC and O_FSYNC do the same thing
+#ifndef O_SYNC
+#define O_SYNC O_FSYNC
+#endif
+
 /* OS X does not have O_DSYNC */
 #ifndef O_DSYNC
 #define O_DSYNC O_SYNC
#2 lore raw
Quoted 1 line · Anthony Liguori

Please include a Signed-off-by line and use C-style comments instead of C++ style comments. Otherwise, looks good.

What do we have against C++ style comments?

Signature · 2 lines
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.
#3 lore raw
Quoted 6 lines · Anthony Liguori
Quoted 3 lines

Please include a Signed-off-by line and use C-style comments instead
of C++ style comments. Otherwise, looks good.

What do we have against C++ style comments?

They are not valid in C code :P

François.

#4 lore raw
Quoted 11 lines · François Revol
Quoted 8 lines · Anthony Liguori
Quoted 5 lines

Please include a Signed-off-by line and use C-style comments instead
of C++ style comments. Otherwise, looks good.

What do we have against C++ style comments?

They are not valid in C code :P

I think C99 allows them. Too bad date-related jokes are forbidden.

Signature · 2 lines
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.
#5 lore raw
In reply to François Revol
Quoted 8 lines · François Revol
Quoted 6 lines · Anthony Liguori
Quoted 3 lines

Please include a Signed-off-by line and use C-style comments instead
of C++ style comments. Otherwise, looks good.

What do we have against C++ style comments?

They are not valid in C code :P

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Of course I still forbid them in my code too - they look wrong :-)

-- Jamie

#6 lore raw

Am 21.06.2009 um 01:25 schrieb Jamie Lokier:

Quoted 7 lines
Quoted 6 lines · François Revol
Quoted 5 lines · Anthony Liguori
Quoted 2 lines

Please [...] use C-style comments instead
of C++ style comments. [...]

What do we have against C++ style comments?

Quoted 8 lines
Quoted 1 line

They are not valid in C code :P

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Gladly even ten-year-old GCC 2.95.x supports 'em. But even a recent GCC does not when compiled with -ansi, which is why system headers still mustn't use them.

Andreas

Quoted 3 lines

Of course I still forbid them in my code too - they look wrong :-)

-- Jamie

#7 lore raw
Quoted 17 lines · Andreas Färber

Am 21.06.2009 um 01:25 schrieb Jamie Lokier:

Quoted 14 lines
Quoted 7 lines · François Revol
Quoted 5 lines · Anthony Liguori
Quoted 2 lines

Please [...] use C-style comments instead
of C++ style comments. [...]

What do we have against C++ style comments?

They are not valid in C code :P

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

Quoted 6 lines
Quoted 2 lines

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Gladly even ten-year-old GCC 2.95.x supports 'em. But even a recent GCC does not when compiled with -ansi, which is why system headers still mustn't use them.

This is wrong too. They can be used in system headers.

"gcc -ansi -pedantic -Wall -Werror" accepts C99 comments in system headers.

It just complains about them in your own code, since you asked it to. :-)

-- Jamie

#8 lore raw

Am 24.06.2009 um 20:25 schrieb Jamie Lokier:

Quoted 18 lines
Quoted 17 lines · Andreas Färber

Am 21.06.2009 um 01:25 schrieb Jamie Lokier:

Quoted 14 lines
Quoted 7 lines · François Revol
Quoted 5 lines · Anthony Liguori
Quoted 2 lines

Please [...] use C-style comments instead
of C++ style comments. [...]

What do we have against C++ style comments?

They are not valid in C code :P

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Gladly even ten-year-old GCC 2.95.x supports 'em. But even a recent GCC does not when compiled with -ansi, which is why system headers still mustn't use them.

This is wrong too. They can be used in system headers.

No, that is wrong. freetype2 definitely complained when Haiku had C99 comments in its POSIX system headers. ;)

Andreas

Quoted 6 lines

"gcc -ansi -pedantic -Wall -Werror" accepts C99 comments in system headers.

It just complains about them in your own code, since you asked it to. :-)

-- Jamie

#9 lore raw
Quoted 26 lines · Andreas Färber

Am 24.06.2009 um 20:25 schrieb Jamie Lokier:

Quoted 23 lines
Quoted 22 lines · Andreas Färber

Am 21.06.2009 um 01:25 schrieb Jamie Lokier:

Quoted 18 lines
Quoted 9 lines · François Revol
Quoted 7 lines · Anthony Liguori
Quoted 3 lines

Please [...] use C-style comments instead
of C++ style comments. [...]

What do we have against C++ style comments?

They are not valid in C code :P

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

Gladly even ten-year-old GCC 2.95.x supports 'em. But even a recent GCC does not when compiled with -ansi, which is why system headers still mustn't use them.

This is wrong too.  They can be used in system headers.

No, that is wrong. freetype2 definitely complained when Haiku had C99 comments in its POSIX system headers. ;)

#pragma GCC system_header

Best regards,
Filip Navara

#10 lore raw
Quoted 31 lines · Filip Navara
Quoted 30 lines · Andreas Färber

Am 24.06.2009 um 20:25 schrieb Jamie Lokier:

Quoted 27 lines
Quoted 25 lines · Andreas Färber

Am 21.06.2009 um 01:25 schrieb Jamie Lokier:

Quoted 11 lines
Quoted 9 lines · François Revol
Quoted 7 lines · Anthony Liguori
Quoted 3 lines

Please [...] use C-style comments instead
of C++ style comments. [...]

What do we have against C++ style comments?

They are not valid in C code :P

Quoted 7 lines

Wrong.

Valid in GNU C for about 14 years, and ANSI/ISO C for about 10 years...

I'll be surprised if any C compiler used on any open source operating system, or any major commercial C compiler, doesn't accept them.

Gladly even ten-year-old GCC 2.95.x supports 'em. But even a recent GCC does not when compiled with -ansi, which is why system headers still mustn't use them.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

This is wrong too.  They can be used in system headers.

No, that is wrong. freetype2 definitely complained when Haiku had C99 comments in its POSIX system headers. ;)

#pragma GCC system_header

GCC has a notion of certain directories being system header directories.

So you don't need #pragma for headers in those, and if freetype2 was installed in them, that should have worked.

But if you install those headers in other places, then it doesn't treat them as system headers and becomes more strict about the language.

You can add more system header directories, but of course (nearly) everyone uses -Idir to add directories as that's well known and portable.

-- Jamie

#11 lore raw
In reply to Avi Kivity
Quoted 4 lines · Avi Kivity
Quoted 1 line · Anthony Liguori

Please include a Signed-off-by line and use C-style comments instead of C++ style comments. Otherwise, looks good.

What do we have against C++ style comments?

Just consistency:

anthony@squirrel:~/git/qemu$ grep '//' *.c | wc -l
163
anthony@squirrel:~/git/qemu$ grep '/\*' *.c | wc -l
5642

Nothing personal. I think the kernel has a policy of C89 comments only.

Regards,

Anthony Liguori

#12 lore raw
Quoted 4 lines · Anthony Liguori
Quoted 3 lines · Avi Kivity
Quoted 1 line · Anthony Liguori

Please include a Signed-off-by line and use C-style comments instead of C++ style comments. Otherwise, looks good.

Quoted 9 lines
Quoted 1 line

What do we have against C++ style comments?

Just consistency:

anthony@squirrel:~/git/qemu$ grep '//' *.c | wc -l
163
anthony@squirrel:~/git/qemu$ grep '/\*' *.c | wc -l
5642

Well, consistency trumps convenience.

Quoted 1 line

Nothing personal. I think the kernel has a policy of C89 comments only.

This isn't the kernel, we have our own coding style. But if we don't want C++ comments, that's fine.

Signature · 1 line
error compiling committee.c: too many arguments to function
#13 lore raw
In reply to m a
Quoted 5 lines · m a

This patch replaces O_SYNC with O_FSYNC. These two flags do the same thing, but only O_FSYNC is available in Mac OS 10.3 and under. It only replaces O_SYNC if it doesn't exist. This patch allows the file block-raw-posix.c to compile on Mac OS 10.3. This is my first time submitting a patch, so there might have been a few mistakes made.

But O_SYNC is a standard posix flag, while O_FSYNC appears to be a BSD extension. Also the actual code uses O_DSYNC anyway, which also is in Posix but not actually natively supported by some OSes, e.g. Linux (but still provided in libc there).

block-raw-posix (updated).c line 82

Quoted 13 lines
block-raw-posix (updated).c renamed from block-raw-posix.c +6 -0
@@ -73,6 +73,12 @@
 
+// O_SYNC isn't available on Mac OS 10.3 and under
+// O_SYNC and O_FSYNC do the same thing
+#ifndef O_SYNC
+#define O_SYNC O_FSYNC
+#endif
+
 /* OS X does not have O_DSYNC */
 #ifndef O_DSYNC
 #define O_DSYNC O_SYNC

So if the code here is correct and Darwin is the only supported OS where O_DSYNC is missing we could just replace the O_SYNC in the last line with O_FSYNC.

#14 lore raw
Quoted 11 lines · Christoph Hellwig
Quoted 5 lines · m a

This patch replaces O_SYNC with O_FSYNC. These two flags do the same thing, but only O_FSYNC is available in Mac OS 10.3 and under. It only replaces O_SYNC if it doesn't exist. This patch allows the file block-raw-posix.c to compile on Mac OS 10.3. This is my first time submitting a patch, so there might have been a few mistakes made.

But O_SYNC is a standard posix flag, while O_FSYNC appears to be a BSD extension. Also the actual code uses O_DSYNC anyway, which also is in Posix but not actually natively supported by some OSes, e.g. Linux (but still provided in libc there).

If O_FSYNC and O_SYNC do the same thing, and O_SYNC is used anywhere, there's no harm in this for portability:

    #if !defined(O_SYNC) && defined(O_FSYNC)
    #define O_SYNC O_FSYNC
    #endif

The patch assumes O_FSYNC is defined if O_SYNC isn't, which is wrong.

Quoted 7 lines
Quoted 3 lines

/* OS X does not have O_DSYNC */
#ifndef O_DSYNC
#define O_DSYNC O_SYNC

So if the code here is correct and Darwin is the only supported OS where O_DSYNC is missing we could just replace the O_SYNC in the last line with O_FSYNC.

I agree, though the comment might be misleading, if there's another supported OS without O_DSYNC.

-- Jamie

#15 lore raw
Quoted 9 lines · Jamie Lokier

If O_FSYNC and O_SYNC do the same thing, and O_SYNC is used anywhere, there's no harm in this for portability:

    #if !defined(O_SYNC) && defined(O_FSYNC)
    #define O_SYNC O_FSYNC
    #endif

The patch assumes O_FSYNC is defined if O_SYNC isn't, which is wrong.

For what this patch is, it really isn't that important IMHO.

This is just a work around for a broken platform.

Regards,

Anthony Liguori

author
#16 lore raw
Quoted 8 lines · Anthony Liguori
Quoted 7 lines · Jamie Lokier

If O_FSYNC and O_SYNC do the same thing, and O_SYNC is used anywhere, there's no harm in this for portability:

    #if !defined(O_SYNC) && defined(O_FSYNC)
    #define O_SYNC O_FSYNC
    #endif
Quoted 7 lines
Quoted 2 lines

The patch assumes O_FSYNC is defined if O_SYNC isn't, which is wrong.

For what this patch is, it really isn't that important IMHO.

This is just a work around for a broken platform.

Even though you are correct, did you have to add the 'broken platform' remark?

#17 lore raw
Quoted 2 lines · G 3 wrote:

Even though you are correct, did you have to add the 'broken platform' remark?

Well, O_SYNC is POSIX so anything which claims to be a unix and doesn't have O_SYNC is broken :-)

-- Jamie

author
#18 lore raw
In reply to Jamie Lokier
Quoted 26 lines · Jamie Lokier
Quoted 11 lines · Christoph Hellwig
Quoted 5 lines · m a

This patch replaces O_SYNC with O_FSYNC. These two flags do the same thing, but only O_FSYNC is available in Mac OS 10.3 and under. It only replaces O_SYNC if it doesn't exist. This patch allows the file block-raw-posix.c to compile on Mac OS 10.3. This is my first time submitting a patch, so there might have been a few mistakes made.

But O_SYNC is a standard posix flag, while O_FSYNC appears to be a BSD extension. Also the actual code uses O_DSYNC anyway, which also is in Posix but not actually natively supported by some OSes, e.g. Linux (but still provided in libc there).

If O_FSYNC and O_SYNC do the same thing, and O_SYNC is used anywhere, there's no harm in this for portability:

    #if !defined(O_SYNC) && defined(O_FSYNC)
    #define O_SYNC O_FSYNC
    #endif

The patch assumes O_FSYNC is defined if O_SYNC isn't, which is wrong.

Quoted 4 lines
Quoted 3 lines

/* OS X does not have O_DSYNC */
#ifndef O_DSYNC
#define O_DSYNC O_SYNC

Quoted 10 lines
Quoted 3 lines

So if the code here is correct and Darwin is the only supported OS where O_DSYNC is missing we could just replace the O_SYNC in the last line with O_FSYNC.

I agree, though the comment might be misleading, if there's another supported OS without O_DSYNC.

-- Jamie

Your logic does make sense. I guess I should have done this instead:

#ifndef O_SYNC
	#ifdef O_FSYNC
		#define O_SYNC 	O_FSYNC
	
	#else
		#define O_SYNC 	0
	
	#endif

#endif

Defining O_SYNC as zero is ok because this flag is 'ORed' with other flags.