The change to /usr/include/linux/socket.h introduced in linux-libc-dev 2.6.26
(from proposed-updates) now breaks the following simple test program:

#include <stddef.h>
#include <linux/socket.h>
#include <linux/netlink.h>

Several autoconf-scripts (notably strace) try to compile this program to
detect "linux/netlink.h" and consequently fail. Please revert the change that
remove the sub-includes from linux/socket.h



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Wed, Nov 11, 2009 at 12:01:00PM +0100, Helge Bahmann wrote:
> The change to /usr/include/linux/socket.h introduced in linux-libc-dev 2.6.26
> (from proposed-updates) now breaks the following simple test program:
>
> #include <stddef.h>
> #include <linux/socket.h>
> #include <linux/netlink.h>
>
> Several autoconf-scripts (notably strace) try to compile this program to
> detect "linux/netlink.h" and consequently fail. Please revert the change that
> remove the sub-includes from linux/socket.h

Thanks for testing!

Does anyone have an alternate fix, or should we just revert this for
now?

--
dann frazier




--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Am Mittwoch, den 11.11.2009, 09:26 -0700 schrieb dann frazier:
> > Several autoconf-scripts (notably strace) try to compile this program to
> > detect "linux/netlink.h" and consequently fail. Please revert the change that
> > remove the sub-includes from linux/socket.h

I can confirm the breakage (patch applied on a Sid system.)

> Does anyone have an alternate fix, or should we just revert this for
> now?

I tested several variations of the patch. Re-adding the "#if
defined ..." line (or: removing that hunk from the patch) gives a
version under which both test cases compile successfully. (The one
mentioned by Helge and the one by the bug submitter.)

Best regards
Manuel




--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Wed, Nov 11, 2009 at 09:26:04AM -0700, dann frazier wrote:
> On Wed, Nov 11, 2009 at 12:01:00PM +0100, Helge Bahmann wrote:
> > The change to /usr/include/linux/socket.h introduced in linux-libc-dev 2.6.26
> > (from proposed-updates) now breaks the following simple test program:
> >
> > #include <stddef.h>
> > #include <linux/socket.h>
> > #include <linux/netlink.h>
> >
> > Several autoconf-scripts (notably strace) try to compile this program to
> > detect "linux/netlink.h" and consequently fail. Please revert the change that
> > remove the sub-includes from linux/socket.h
>
> Thanks for testing!
>
> Does anyone have an alternate fix, or should we just revert this for
> now?

It is probably best to revert this for lenny. However, this change has
already been accepted upstream (I wouldn't have applied it otherwise) so we
also need to come up with a proper fix.

Ben.

--
Ben Hutchings
Lowery's Law:
If it jams, force it. If it breaks, it needed replacing anyway.



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Wed, Nov 11, 2009 at 10:52:09PM +0100, Manuel Prinz wrote:
> Am Mittwoch, den 11.11.2009, 09:26 -0700 schrieb dann frazier:
> > > Several autoconf-scripts (notably strace) try to compile this program to
> > > detect "linux/netlink.h" and consequently fail. Please revert the change that
> > > remove the sub-includes from linux/socket.h
>
> I can confirm the breakage (patch applied on a Sid system.)
>
> > Does anyone have an alternate fix, or should we just revert this for
> > now?
>
> I tested several variations of the patch. Re-adding the "#if
> defined ..." line (or: removing that hunk from the patch) gives a
> version under which both test cases compile successfully. (The one
> mentioned by Helge and the one by the bug submitter.)

Could you test this:

--- a/include/linux/socket.h
+++ b/include/linux/socket.h
@@ -16,7 +16,12 @@ struct __kernel_sockaddr_storage {
/* _SS_MAXSIZE value minus size of ss_family */
} __attribute__ ((aligned(_K_SS_ALIGNSIZE))); /* force desired alignment */

-#ifdef __KERNEL__
+#ifndef __KERNEL__
+
+/* I think this is what you wanted: */
+#include <sys/socket.h>
+
+#else

#include <asm/socket.h> /* arch-dependent defines */
#include <linux/sockios.h> /* the SIOCxxx I/O controls */
--- END ---

Ben.

--
Ben Hutchings
Lowery's Law:
If it jams, force it. If it breaks, it needed replacing anyway.



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

Hi Ben!

Am Mittwoch, den 11.11.2009, 22:34 +0000 schrieb Ben Hutchings:
> Could you test this:
>
> --- a/include/linux/socket.h
> +++ b/include/linux/socket.h
> @@ -16,7 +16,12 @@ struct __kernel_sockaddr_storage {
> /* _SS_MAXSIZE value minus size of ss_family */
> } __attribute__ ((aligned(_K_SS_ALIGNSIZE))); /* force desired alignment */
>
> -#ifdef __KERNEL__
> +#ifndef __KERNEL__
> +
> +/* I think this is what you wanted: */
> +#include <sys/socket.h>
> +
> +#else
>
> #include <asm/socket.h> /* arch-dependent defines */
> #include <linux/sockios.h> /* the SIOCxxx I/O controls */
> --- END ---

This change works for both test cases.

Best regards
Manuel




--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Sat, Nov 14, 2009 at 05:09:02PM +0100, Manuel Prinz wrote:
> Hi Ben!
>
> Am Mittwoch, den 11.11.2009, 22:34 +0000 schrieb Ben Hutchings:
> > Could you test this:
> >
> > --- a/include/linux/socket.h
> > +++ b/include/linux/socket.h
> > @@ -16,7 +16,12 @@ struct __kernel_sockaddr_storage {
> > /* _SS_MAXSIZE value minus size of ss_family */
> > } __attribute__ ((aligned(_K_SS_ALIGNSIZE))); /* force desired alignment */
> >
> > -#ifdef __KERNEL__
> > +#ifndef __KERNEL__
> > +
> > +/* I think this is what you wanted: */
> > +#include <sys/socket.h>
> > +
> > +#else
> >
> > #include <asm/socket.h> /* arch-dependent defines */
> > #include <linux/sockios.h> /* the SIOCxxx I/O controls */
> > --- END ---
>
> This change works for both test cases.

Did you submit this upstream?

Cheers,
Moritz



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

On Tue, 2010-02-09 at 23:21 +0100, Moritz Muehlenhoff wrote:
> On Sat, Nov 14, 2009 at 05:09:02PM +0100, Manuel Prinz wrote:
> > Hi Ben!
> >
> > Am Mittwoch, den 11.11.2009, 22:34 +0000 schrieb Ben Hutchings:
> > > Could you test this:
[...]
> > This change works for both test cases.
>
> Did you submit this upstream?

Not exactly, but I don't think it would be accepted.

David Miller wrote:
> Still, you still essentially have two choices:
>
> 1) Tell userland, sorry you need to include sys/socket.h before
> other "socket stuff" and that means linux/netlink.h in
> particular
>
> 2) Put a !__KERNEL__ sys/socket.h include in there, but that's
> asking for trouble.

(see <http://thread.gmane.org/gmane.linux.network/143380/focus=143389>).

Ben.

--
Ben Hutchings
friends: People who know you well, but like you anyway.