So the last week was much quieter than the preceding ones, which makes
me suspect that one reason -rc7 was bigger than I liked was that
people were gaming the system and had timed some of their pull
requests for just before the release, explaining why -rc7 was big
enough that I didn't actually want to do a final release last week.
Please don't do that.
Anyway. Whatever the reason, this week has been very quiet, which
makes me much more comfortable doing the final 3.9 release, so I guess
the last -rc8 ended up working. Because not only aren't there very
many commits here, even the ones that made it really are tiny and not
pretty obscure and not very interesting.
Also, this obviously means that the merge window is open. I won't be
merging anything today, but if you start sending me your pull requests
(Konrad already sent in his Xen pull request for the 3.10 merge window
a week ago), tomorrow the flood gates start opening..
Linus
---
Borislav Petkov (1):
x86, efi: Fix a build warning
David S. Miller (1):
sparc64: Fix missing put_cpu_var() in tlb_batch_add_one() when
not batching.
Hans Verkuil (1):
[media] cx25821: do not expose broken video output streams
Helge Deller (1):
parisc: disable -mlong-calls compiler option for kernel modules
Jiri Slaby (1):
TTY: fix atime/mtime regression
John David Anglin (3):
parisc: Provide __ucmpdi2 to resolve undefined references in 32
bit builds.
parisc: Change kunmap macro to static inline function
parisc: use spin_lock_irqsave/spin_unlock_irqrestore for PTE updates
Josh Boyer (1):
efi: Check EFI revision in setup_efi_vars
Kishon Vijay Abraham I (1):
ARM: OMAP4: hwmod data: make 'ocp2scp_usb_phy_phy_48m" as the main clock
Linus Torvalds (2):
vm: add no-mmu vm_iomap_memory() stub
Linux 3.9
Linus Walleij (1):
Revert "gpio: pxa: set initcall level to module init"
Mathias Krause (1):
crypto: algif - suppress sending source address information in recvmsg
Matt Fleming (1):
efivars: only check for duplicates on the registered list
Mauro Carvalho Chehab (1):
[media] mb86a20s: Fix estimate_rate setting
Paul E. McKenney (1):
events: Protect access via task_subsys_state_check()
Ralf Baechle (1):
Revert "MIPS: page.h: Provide more readable definition for PAGE_MASK."
Rusty Russell (1):
kernel/hz.bc: ignore.
Will Deacon (1):
parisc: uaccess: fix compiler warnings caused by __put_user casting
Zhao Hongjiang (1):
aio: fix possible invalid memory access when DEBUG is enabled
On 28.04.2013 17:56, Linus Torvalds wrote:
I'm not sure if it is supported, but building with a O= -directory that
was previously used to build a 3.8.9-kernel that was "clean"ed (but not
"mrproper"ed) before building 3.9 results in a build-failure:
make[2]: *** No rule to make target `/xssd/usr_src/ko/leeloo/include/config/hz.h', needed by `kernel/hz.bc'. Stop.
After deleting everything in the KBUILD_OUTPUT-directory, except the
.config-file, the kernel-build worked again.
At least in the past a "clean" seamed to be enough. I had a file with a
2.6.39-name inside, so it seams i have build my kernels this way for the
last few releases.
--
Matthias
On 04/29/2013 01:36 AM, Matthias Schniedermeyer wrote:
> On 28.04.2013 17:56, Linus Torvalds wrote:
>
> I'm not sure if it is supported, but building with a O= -directory that
> was previously used to build a 3.8.9-kernel that was "clean"ed (but not
> "mrproper"ed) before building 3.9 results in a build-failure:
> make[2]: *** No rule to make target `/xssd/usr_src/ko/leeloo/include/config/hz.h', needed by `kernel/hz.bc'. Stop.
>
> After deleting everything in the KBUILD_OUTPUT-directory, except the
> .config-file, the kernel-build worked again.
>
> At least in the past a "clean" seamed to be enough. I had a file with a
> 2.6.39-name inside, so it seams i have build my kernels this way for the
> last few releases.
>
This is a question for the Kbuild people I believe...
-hpa
On 04/29/2013 01:36 AM, Matthias Schniedermeyer wrote:
> On 28.04.2013 17:56, Linus Torvalds wrote:
>
> I'm not sure if it is supported, but building with a O= -directory that
> was previously used to build a 3.8.9-kernel that was "clean"ed (but not
> "mrproper"ed) before building 3.9 results in a build-failure:
> make[2]: *** No rule to make target `/xssd/usr_src/ko/leeloo/include/config/hz.h', needed by `kernel/hz.bc'. Stop.
>
> After deleting everything in the KBUILD_OUTPUT-directory, except the
> .config-file, the kernel-build worked again.
>
> At least in the past a "clean" seamed to be enough. I had a file with a
> 2.6.39-name inside, so it seams i have build my kernels this way for the
> last few releases.
>
This seems to be the fundamental problem here... include/config is used
to track dependencies on configuration, but it looks like there are
events that seem to delete include/config without them being regenerated?
-hpa