On Thu, 2006-09-21 at 12:00 +0300, Ismail Donmez wrote:
> This looks way cleaner, I can cook up a patch for this if everyone
> agrees. What do you guys think?
Makes sense.
--
dwmw2
21 Eyl 2006 Per 12:02 tarihinde, David Woodhouse şunları yazmıştı:
> On Thu, 2006-09-21 at 12:00 +0300, Ismail Donmez wrote:
> > This looks way cleaner, I can cook up a patch for this if everyone
> > agrees. What do you guys think?
>
> Makes sense.
Ok here is patch replacing __STRICT_ANSI__ with __extension__ as suggested by
Kyle Moffett:
__STRICT_ANSI__ usage in types.h header results in compile errors for some
userspace packages[1] when used with gcc -ansi flag. With the suggestion of
Kyle Moffett I replace strict ansi checks with __extension__ to tell gcc not
to error or warn on gcc extensions. Compile tested on x86 with 2.6.18.
[1] kopete and kdetv are such userspace applications.
Signed-off-by: Ismail Donmez <[email protected]>
--
Insert some japanese signature here
21 Eyl 2006 Per 15:53 tarihinde, Ismail Donmez şunları yazmıştı:
> 21 Eyl 2006 Per 12:02 tarihinde, David Woodhouse şunları yazmıştı:
> > On Thu, 2006-09-21 at 12:00 +0300, Ismail Donmez wrote:
> > > This looks way cleaner, I can cook up a patch for this if everyone
> > > agrees. What do you guys think?
> >
> > Makes sense.
>
> Ok here is patch replacing __STRICT_ANSI__ with __extension__ as suggested
> by Kyle Moffett:
>
> __STRICT_ANSI__ usage in types.h header results in compile errors for some
> userspace packages[1] when used with gcc -ansi flag. With the suggestion of
> Kyle Moffett I replace strict ansi checks with __extension__ to tell gcc
> not to error or warn on gcc extensions. Compile tested on x86 with 2.6.18.
>
> [1] kopete and kdetv are such userspace applications.
>
> Signed-off-by: Ismail Donmez <[email protected]>
David, is this ok? It would be good to apply this for 2.6.19 so all of KDE
would compile ( all of the parts I tested ) with kernel-headers.
Regards,
ismail
--
They that can give up essential liberty to obtain a little temporary safety
deserve neither liberty nor safety.
-- Benjamin Franklin
On Thu, 2006-09-28 at 10:03 +0300, Ismail Donmez wrote:
> David, is this ok? It would be good to apply this for 2.6.19 so all of KDE
> would compile ( all of the parts I tested ) with kernel-headers.
Looks good to me.
--
dwmw2
28 Eyl 2006 Per 10:06 tarihinde, David Woodhouse şunları yazmıştı:
> On Thu, 2006-09-28 at 10:03 +0300, Ismail Donmez wrote:
> > David, is this ok? It would be good to apply this for 2.6.19 so all of
> > KDE would compile ( all of the parts I tested ) with kernel-headers.
>
> Looks good to me.
Andrew, now that David gave his blessing , can you push this for 2.6.19?
Regards,
ismail
--
They that can give up essential liberty to obtain a little temporary safety
deserve neither liberty nor safety.
-- Benjamin Franklin
On Thu, 28 Sep 2006 10:30:25 +0300
Ismail Donmez <[email protected]> wrote:
> 28 Eyl 2006 Per 10:06 tarihinde, David Woodhouse şunları yazmıştı:
> > On Thu, 2006-09-28 at 10:03 +0300, Ismail Donmez wrote:
> > > David, is this ok? It would be good to apply this for 2.6.19 so all of
> > > KDE would compile ( all of the parts I tested ) with kernel-headers.
> >
> > Looks good to me.
>
> Andrew, now that David gave his blessing , can you push this for 2.6.19?
>
Bisection shows that this patch causes these depmod warnings:
WARNING: "snd_card_disconnect" [sound/usb/usx2y/snd-usb-usx2y.ko] has no CRC!
WARNING: "snd_card_disconnect" [sound/usb/snd-usb-audio.ko] has no CRC!
WARNING: "snd_card_disconnect" [sound/drivers/mpu401/snd-mpu401.ko] has no CRC!
WARNING: "no_llseek" [net/sunrpc/sunrpc.ko] has no CRC!
WARNING: "seq_lseek" [net/sunrpc/sunrpc.ko] has no CRC!
WARNING: "seq_lseek" [net/sctp/sctp.ko] has no CRC!
WARNING: "seq_lseek" [net/netfilter/x_tables.ko] has no CRC!
WARNING: "seq_lseek" [net/netfilter/nfnetlink_queue.ko] has no CRC!
WARNING: "seq_lseek" [net/netfilter/nfnetlink_log.ko] has no CRC!
WARNING: "seq_lseek" [net/ipv4/netfilter/ipt_hashlimit.ko] has no CRC!
WARNING: "seq_lseek" [net/ipv4/netfilter/ipt_CLUSTERIP.ko] has no CRC!
WARNING: "seq_lseek" [net/ipv4/netfilter/ip_conntrack.ko] has no CRC!
WARNING: "generic_file_llseek" [fs/ufs/ufs.ko] has no CRC!
WARNING: "remote_llseek" [fs/smbfs/smbfs.ko] has no CRC!
WARNING: "generic_file_llseek" [fs/ntfs/ntfs.ko] has no CRC!
WARNING: "vfs_llseek" [fs/nfsd/nfsd.ko] has no CRC!
WARNING: "seq_lseek" [fs/nfsd/nfsd.ko] has no CRC!
WARNING: "remote_llseek" [fs/nfs/nfs.ko] has no CRC!
[etc]
I don't know why that would happen.
http://www.zip.com.au/~akpm/linux/patches/stuff/config-sony
From: Ismail Donmez <[email protected]>
__STRICT_ANSI__ usage in types.h header results in compile errors for some
userspace packages[1] when used with gcc -ansi flag. With the suggestion
of Kyle Moffett I replace strict ansi checks with __extension__ to tell gcc
not to error or warn on gcc extensions. Compile tested on x86 with 2.6.18.
[1] kopete and kdetv are such userspace applications.
Signed-off-by: Ismail Donmez <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Cc: Kyle Moffett <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
---
include/asm-arm/types.h | 6 +++---
include/asm-arm26/types.h | 6 +++---
include/asm-cris/types.h | 8 ++++----
include/asm-frv/types.h | 6 +++---
include/asm-h8300/types.h | 6 +++---
include/asm-i386/types.h | 6 +++---
include/asm-m32r/types.h | 6 +++---
include/asm-m68k/types.h | 6 +++---
include/asm-mips/types.h | 12 ++++++------
include/asm-parisc/types.h | 6 +++---
include/asm-powerpc/types.h | 6 +++---
include/asm-s390/types.h | 10 +++++-----
include/asm-sh/types.h | 6 +++---
include/asm-sh64/types.h | 6 +++---
include/asm-v850/types.h | 6 +++---
include/asm-xtensa/types.h | 6 +++---
include/linux/types.h | 18 +++++++++---------
17 files changed, 63 insertions(+), 63 deletions(-)
diff -puN include/asm-arm/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-arm/types.h
--- a/include/asm-arm/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-arm/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-arm26/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-arm26/types.h
--- a/include/asm-arm26/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-arm26/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-cris/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-cris/types.h
--- a/include/asm-cris/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-cris/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
@@ -48,7 +48,7 @@ typedef signed long long s64;
typedef unsigned long long u64;
/* Dma addresses are 32-bits wide, just like our other addresses. */
-
+
typedef u32 dma_addr_t;
typedef u32 dma64_addr_t;
diff -puN include/asm-frv/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-frv/types.h
--- a/include/asm-frv/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-frv/types.h
@@ -30,9 +30,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-h8300/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-h8300/types.h
--- a/include/asm-h8300/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-h8300/types.h
@@ -27,9 +27,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
/*
diff -puN include/asm-i386/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-i386/types.h
--- a/include/asm-i386/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-i386/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-m32r/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-m32r/types.h
--- a/include/asm-m32r/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-m32r/types.h
@@ -23,9 +23,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-m68k/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-m68k/types.h
--- a/include/asm-m68k/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-m68k/types.h
@@ -27,9 +27,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-mips/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-mips/types.h
--- a/include/asm-mips/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-mips/types.h
@@ -34,9 +34,9 @@ typedef unsigned long __u64;
#else
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif
@@ -69,9 +69,9 @@ typedef unsigned long u64;
#else
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long s64;
-typedef unsigned long long u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long s64;
+__extension__ typedef unsigned long long u64;
#endif
#endif
diff -puN include/asm-parisc/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-parisc/types.h
--- a/include/asm-parisc/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-parisc/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-powerpc/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-powerpc/types.h
--- a/include/asm-powerpc/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-powerpc/types.h
@@ -40,9 +40,9 @@ typedef unsigned int __u32;
typedef __signed__ long __s64;
typedef unsigned long __u64;
#else
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __powerpc64__ */
diff -puN include/asm-s390/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-s390/types.h
--- a/include/asm-s390/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-s390/types.h
@@ -28,9 +28,9 @@ typedef __signed__ int __s32;
typedef unsigned int __u32;
#ifndef __s390x__
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#else /* __s390x__ */
typedef __signed__ long __s64;
@@ -38,9 +38,9 @@ typedef unsigned long __u64;
#endif
/* A address type so that arithmetic can be done on it & it can be upgraded to
- 64 bit when necessary
+ 64 bit when necessary
*/
-typedef unsigned long addr_t;
+typedef unsigned long addr_t;
typedef __signed__ long saddr_t;
#endif /* __ASSEMBLY__ */
diff -puN include/asm-sh/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-sh/types.h
--- a/include/asm-sh/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-sh/types.h
@@ -19,9 +19,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-sh64/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-sh64/types.h
--- a/include/asm-sh64/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-sh64/types.h
@@ -30,9 +30,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* __ASSEMBLY__ */
diff -puN include/asm-v850/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-v850/types.h
--- a/include/asm-v850/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-v850/types.h
@@ -27,9 +27,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
#endif /* !__ASSEMBLY__ */
diff -puN include/asm-xtensa/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/asm-xtensa/types.h
--- a/include/asm-xtensa/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/asm-xtensa/types.h
@@ -29,9 +29,9 @@ typedef unsigned short __u16;
typedef __signed__ int __s32;
typedef unsigned int __u32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __signed__ long long __s64;
-typedef unsigned long long __u64;
+#if defined(__GNUC__)
+__extension__ typedef __signed__ long long __s64;
+__extension__ typedef unsigned long long __u64;
#endif
/*
diff -puN include/linux/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__ include/linux/types.h
--- a/include/linux/types.h~fix-compile-errors-for-64-bit-types-in-headers-with-__strict_ansi__
+++ a/include/linux/types.h
@@ -54,8 +54,8 @@ typedef __kernel_uid_t uid_t;
typedef __kernel_gid_t gid_t;
#endif /* __KERNEL__ */
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __kernel_loff_t loff_t;
+#if defined(__GNUC__)
+__extension__ typedef __kernel_loff_t loff_t;
#endif
/*
@@ -120,10 +120,10 @@ typedef __u8 uint8_t;
typedef __u16 uint16_t;
typedef __u32 uint32_t;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __u64 uint64_t;
-typedef __u64 u_int64_t;
-typedef __s64 int64_t;
+#if defined(__GNUC__)
+__extension__ typedef __u64 uint64_t;
+__extension__ typedef __u64 u_int64_t;
+__extension__ typedef __s64 int64_t;
#endif
/* this is a special 64bit data type that is 8-byte aligned */
@@ -172,9 +172,9 @@ typedef __u16 __bitwise __le16;
typedef __u16 __bitwise __be16;
typedef __u32 __bitwise __le32;
typedef __u32 __bitwise __be32;
-#if defined(__GNUC__) && !defined(__STRICT_ANSI__)
-typedef __u64 __bitwise __le64;
-typedef __u64 __bitwise __be64;
+#if defined(__GNUC__)
+__extension__ typedef __u64 __bitwise __le64;
+__extension__ typedef __u64 __bitwise __be64;
#endif
#ifdef __KERNEL__
_
On Oct 01, 2006, at 00:53:43, Andrew Morton wrote:
> On Thu, 28 Sep 2006 10:30:25 +0300
> Ismail Donmez <[email protected]> wrote:
>
>> 28 Eyl 2006 Per 10:06 tarihinde, David Woodhouse şunları
>> yazmıştı:
>>> On Thu, 2006-09-28 at 10:03 +0300, Ismail Donmez wrote:
>>> David, is this ok? It would be good to apply this for 2.6.19 so
>>> all of KDE would compile ( all of the parts I tested ) with
>>> kernel-headers.
>>>
>>> Looks good to me.
>>
>> Andrew, now that David gave his blessing , can you push this for
>> 2.6.19?
>>
>
> Bisection shows that this patch causes these depmod warnings:
>
> WARNING: "snd_card_disconnect" [sound/usb/usx2y/snd-usb-usx2y.ko]
> has no CRC!
> [etc]
>
> I don't know why that would happen.
>
> From: Ismail Donmez <[email protected]>
>
> __STRICT_ANSI__ usage in types.h header results in compile errors
> for some userspace packages[1] when used with gcc -ansi flag. With
> the suggestion of Kyle Moffett I replace strict ansi checks with
> __extension__ to tell gcc not to error or warn on gcc extensions.
> Compile tested on x86 with 2.6.18.
Best guess: Depmod does some kind of funny type-based expansion and
hashing of the symbols which doesn't understand the "__extension__"
keyword. Probably the simplest thing to do is to add "-
D__extension__=" to the depmod preprocessing flags. Alternatively
you could teach depmod to completely ignore the __extension__ keyword
when it shows up in the sources, but the former seems like it would
be much simpler.
Just thinking about it we probably also need to educate sparse about
__extension__ too. Perhaps somebody could also add an sparse flag to
make it warn about nonportable constructs in exported header files.
I'd submit a patch but my knowledge of kernel makefiles and depmod is
somewhere between zero and none, exclusive.
Cheers,
Kyle Moffett
On Sunday 01 October 2006 08:20, Kyle Moffett wrote:
> On Oct 01, 2006, at 00:53:43, Andrew Morton wrote:
[...]
> > Bisection shows that this patch causes these depmod warnings:
> >
> > WARNING: "snd_card_disconnect" [sound/usb/usx2y/snd-usb-usx2y.ko]
> > has no CRC!
> > [etc]
> >
> > I don't know why that would happen.
> >
> > From: Ismail Donmez <[email protected]>
> >
> > __STRICT_ANSI__ usage in types.h header results in compile errors
> > for some userspace packages[1] when used with gcc -ansi flag. With
> > the suggestion of Kyle Moffett I replace strict ansi checks with
> > __extension__ to tell gcc not to error or warn on gcc extensions.
> > Compile tested on x86 with 2.6.18.
>
> Best guess: Depmod does some kind of funny type-based expansion and
> hashing of the symbols which doesn't understand the "__extension__"
> keyword. Probably the simplest thing to do is to add "-
> D__extension__=" to the depmod preprocessing flags. Alternatively
> you could teach depmod to completely ignore the __extension__ keyword
> when it shows up in the sources, but the former seems like it would
> be much simpler.
>
> Just thinking about it we probably also need to educate sparse about
> __extension__ too. Perhaps somebody could also add an sparse flag to
> make it warn about nonportable constructs in exported header files.
>
> I'd submit a patch but my knowledge of kernel makefiles and depmod is
> somewhere between zero and none, exclusive.
Thanks, I will have a look at it.
Regards,
ismail
On Sun, Oct 01, 2006 at 10:34:56AM +0300, Ismail Donmez wrote:
> On Sunday 01 October 2006 08:20, Kyle Moffett wrote:
> > On Oct 01, 2006, at 00:53:43, Andrew Morton wrote:
> [...]
> > > Bisection shows that this patch causes these depmod warnings:
> > >
> > > WARNING: "snd_card_disconnect" [sound/usb/usx2y/snd-usb-usx2y.ko]
> > > has no CRC!
> > > [etc]
> > >
> > > I don't know why that would happen.
> > >
> > > From: Ismail Donmez <[email protected]>
> > >
> > > __STRICT_ANSI__ usage in types.h header results in compile errors
> > > for some userspace packages[1] when used with gcc -ansi flag. With
> > > the suggestion of Kyle Moffett I replace strict ansi checks with
> > > __extension__ to tell gcc not to error or warn on gcc extensions.
> > > Compile tested on x86 with 2.6.18.
> >
> > Best guess: Depmod does some kind of funny type-based expansion and
> > hashing of the symbols which doesn't understand the "__extension__"
> > keyword. Probably the simplest thing to do is to add "-
> > D__extension__=" to the depmod preprocessing flags. Alternatively
> > you could teach depmod to completely ignore the __extension__ keyword
> > when it shows up in the sources, but the former seems like it would
> > be much simpler.
> >
> > Just thinking about it we probably also need to educate sparse about
> > __extension__ too. Perhaps somebody could also add an sparse flag to
> > make it warn about nonportable constructs in exported header files.
> >
> > I'd submit a patch but my knowledge of kernel makefiles and depmod is
> > somewhere between zero and none, exclusive.
>
> Thanks, I will have a look at it.
I assume you will same errors from the in-kernel modpost.
If you do not do so then there is some inconsistency between depmod
and modpost that ougth to be fixed.
I'm not up to the task atm - busy with my day job and travelling a lot
Sam
On Sunday 01 October 2006 12:14, Sam Ravnborg wrote:
[...]
> > Thanks, I will have a look at it.
>
> I assume you will same errors from the in-kernel modpost.
> If you do not do so then there is some inconsistency between depmod
> and modpost that ougth to be fixed.
>
I didn't see any errors when I did the build, I will try to reproduce and fix
the issue.
Regards,
ismail
01 Eki 2006 Paz 12:14 tarihinde, Sam Ravnborg şunları yazmıştı:
> On Sun, Oct 01, 2006 at 10:34:56AM +0300, Ismail Donmez wrote:
> > On Sunday 01 October 2006 08:20, Kyle Moffett wrote:
[...]
> > > Just thinking about it we probably also need to educate sparse about
> > > __extension__ too. Perhaps somebody could also add an sparse flag to
> > > make it warn about nonportable constructs in exported header files.
> > >
> > > I'd submit a patch but my knowledge of kernel makefiles and depmod is
> > > somewhere between zero and none, exclusive.
> >
> > Thanks, I will have a look at it.
>
> I assume you will same errors from the in-kernel modpost.
> If you do not do so then there is some inconsistency between depmod
> and modpost that ougth to be fixed.
The problem shows itself in the modpost, somehow __extension__ clause seems to
foobar module CRC. I am not yet successfull on making modpost ignore
__extension__ .
Any ideas appreciated.
Regards,
ismail
I'm away from my regular work systems at the moment, but I have a
couple ideas:
On Oct 05, 2006, at 04:16:11, Ismail Donmez wrote:
> 01 Eki 2006 Paz 12:14 tarihinde, Sam Ravnborg şunları yazmıştı:
>> On Sun, Oct 01, 2006 at 10:34:56AM +0300, Ismail Donmez wrote:
>>> On Sunday 01 October 2006 08:20, Kyle Moffett wrote:
> [...]
>>>> Just thinking about it we probably also need to educate sparse
>>>> about
>>>> __extension__ too. Perhaps somebody could also add an sparse
>>>> flag to
>>>> make it warn about nonportable constructs in exported header files.
>>>>
>>>> I'd submit a patch but my knowledge of kernel makefiles and
>>>> depmod is
>>>> somewhere between zero and none, exclusive.
>>>
>>> Thanks, I will have a look at it.
>>
>> I assume you will same errors from the in-kernel modpost.
>> If you do not do so then there is some inconsistency between depmod
>> and modpost that ougth to be fixed.
>
> The problem shows itself in the modpost, somehow __extension__
> clause seems to
> foobar module CRC. I am not yet successfull on making modpost ignore
> __extension__ .
Since GCC doesn't actually _do_ anything with __extension__ in the
object output files, it must be a C-file-based parser problem, so you
should probably be able to build one of the problematic modules and
diff the resultant .cmd files, the .mod files, and the .mod.c files.
Sorry I can't be of more help.
Cheers,
Kyle Moffett
On Thu, 2006-10-05 at 11:16 +0300, Ismail Donmez wrote:
> The problem shows itself in the modpost, somehow __extension__ clause seems to
> foobar module CRC. I am not yet successfull on making modpost ignore
> __extension__ .
>
> Any ideas appreciated.
Something like this (and build with GENERATE_PARSER=1) _ought_ to do it,
but doesn't work:
diff --git a/scripts/genksyms/keywords.gperf b/scripts/genksyms/keywords.gperf
index c75e0c8..1c31f38 100644
--- a/scripts/genksyms/keywords.gperf
+++ b/scripts/genksyms/keywords.gperf
@@ -11,6 +11,7 @@ __attribute, ATTRIBUTE_KEYW
__attribute__, ATTRIBUTE_KEYW
__const, CONST_KEYW
__const__, CONST_KEYW
+__extension__,EXTENSION_KEYW
__inline, INLINE_KEYW
__inline__, INLINE_KEYW
__signed, SIGNED_KEYW
diff --git a/scripts/genksyms/parse.y b/scripts/genksyms/parse.y
index ca04c94..66ae413 100644
--- a/scripts/genksyms/parse.y
+++ b/scripts/genksyms/parse.y
@@ -60,6 +60,7 @@ remove_list(struct string_list **pb, str
%token CONST_KEYW
%token DOUBLE_KEYW
%token ENUM_KEYW
+%token EXTENSION_KEYW
%token EXTERN_KEYW
%token FLOAT_KEYW
%token INLINE_KEYW
@@ -269,7 +270,7 @@ cvar_qualifier_seq:
cvar_qualifier:
CONST_KEYW | VOLATILE_KEYW | ATTRIBUTE_PHRASE
- | RESTRICT_KEYW
+ | RESTRICT_KEYW | EXTENSION_KEYW
{ /* restrict has no effect in prototypes so ignore it */
remove_node($1);
$$ = $1;
--
dwmw2
On Oct 06, 2006, at 04:26:22, David Woodhouse wrote:
> On Thu, 2006-10-05 at 11:16 +0300, Ismail Donmez wrote:
>> The problem shows itself in the modpost, somehow __extension__
>> clause seems to foobar module CRC. I am not yet successfull on
>> making modpost ignore
>> __extension__ .
>>
>> Any ideas appreciated.
>
> Something like this (and build with GENERATE_PARSER=1) _ought_ to
> do it,
> but doesn't work:
>
> @@ -269,7 +270,7 @@ cvar_qualifier_seq:
>
> cvar_qualifier:
> CONST_KEYW | VOLATILE_KEYW | ATTRIBUTE_PHRASE
> - | RESTRICT_KEYW
> + | RESTRICT_KEYW | EXTENSION_KEYW
> { /* restrict has no effect in prototypes so ignore it */
> remove_node($1);
> $$ = $1;
Well it's actually technically not a cvar_qualifier; it's a pseudo-
arbitrarily attached keyword that can be stuck on any number of GCC-
only constructs, like nested code: "__extension__ ({ foo(); 1; })"
for example.
Probably the simplest thing to do is actually to just convert it into
a nonexistent or whitespace token, or if there's a preprocessing step
just #define it to the empty string.
Cheers,
Kyle Moffett