kbuild updates
PATCH 1/1 kbuild updates
Hi Linus.
kbuild updates as accumulated over the last few months. All patches has been in -mm in one or several versions.
Most noteworthy:
1) -Wundef added to CFLAGS. This is the cause of several new warnings, which for the most part has been fixed for now.
2) "PREEMPT" in UTS_VERSION. So we complain when dealing with modules compiled for a wrong kernel
3) Introduced Kbuild.include (the cause of most of
the changes lines in top-level Makefile).
It killed a lot of duplicate code
4) Introduce debug_kallsyms to better debug situations where kallsyms fails the consistency check
5) Added support for building rpm tarballs of source
Some of this was leftovers from my old bitkeeper tree, and authorship is not correct due to wron cogito usage. The changelog message though attributes the correct author.
Please pull from:
www.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.git
Sam
Diffstat:
Documentation/kbuild/makefiles.txt | 6 + Makefile | 160 ++++++++++++++---------------------- arch/m68knommu/Makefile | 2 arch/mips/Makefile | 2 drivers/block/Kconfig | 42 --------- include/asm-generic/vmlinux.lds.h | 9 ++ init/Kconfig | 22 ++++ init/Makefile | 3 scripts/Kbuild.include | 96 +++++++++++++++++++++ scripts/Lindent | 2 scripts/Makefile.build | 7 + scripts/Makefile.clean | 14 +-- scripts/Makefile.host | 3 scripts/Makefile.lib | 99 ---------------------- scripts/Makefile.modinst | 2 scripts/Makefile.modpost | 1 scripts/conmakehash.c | 2 scripts/kallsyms.c | 6 - scripts/kconfig/lkc.h | 2 scripts/kconfig/menu.c | 4 scripts/kconfig/zconf.tab.c_shipped | 8 - scripts/kconfig/zconf.y | 8 - scripts/kernel-doc | 8 + scripts/lxdialog/dialog.h | 2 scripts/lxdialog/inputbox.c | 4 scripts/mkcompile_h | 12 +- scripts/mod/modpost.c | 9 +- scripts/mod/sumversion.c | 8 - scripts/package/Makefile | 24 ++++- scripts/package/builddeb | 56 +++++++++++- scripts/package/buildtar | 111 ++++++++++++++++++++++++ scripts/package/mkspec | 9 ++ scripts/reference_discarded.pl | 1 scripts/reference_init.pl | 1 scripts/setlocalversion | 56 ++++++++++++ usr/Kconfig | 46 ++++++++++ usr/Makefile | 2 37 files changed, 555 insertions(+), 294 deletions(-)
Coywolf Qi Hunt:
kbuild: make help binrpm-pkg fix
Fabio Massimo Di Nitto:
kbuild: modpost needs to cope with new glibc elf header on sparc
Greg Edwards:
kbuild: add ia64 support to rpm Makefile target
Ian Campbell:
kbuild: allow cscope to index multiple architectures
J.A. Magallon:
kbuild: signed char fixes for scripts
Jan-Benedict Glaw:
kbuild: create tarballs
Jeff Mahoney:
Lindent: ignore .indent.pro
Jesper Juhl:
kallsyms: clarify KALLSYMS_ALL help text
Karl Hegbloom:
kbuild: make 'cscope -q' play well with cscope.el
Keenan Pepper:
kbuild: signed/unsigned char fix for make menuconfig
Matthias Urlichs:
kbuild: obey HOSTLOADLIBES_programname for single-file compilation
Olaf Hering:
kbuild: add -Wundef to global CFLAGS
Paolo 'Blaisorblade' Giarrusso:
kbuild: describe Kbuild pitfall
kconfig: trivial cleanup
Randy Dunlap:
scripts/kernel-doc: don't use uninitialized SRCTREE
Ryan Anderson:
kbuild: automatically append a short string to the version based upon the git commit
Sam Ravnborg:
kbuild: Fix build as root then user
buildcheck: reduce DEBUG_INFO noise from reference* scripts
kbuild: Avoid inconsistent kallsyms data
kbuild: "PREEMPT" in UTS_VERSION
kbuild: Add target debug_kallsyms
kbuild: fix buildcheck
kbuild: Don't fail if include/asm symlink exists
uml: Make deb-pkg build target build a Debian-style user-mode-linux package
uml: Restore proper descriptions in make deb-pkg target
kbuild: Fix bug in make deb-pkg when using seperate source and output directories
kbuild: fix make O=... build
kbuild: drop -Wundef from HOSTCFLAGS for now
kbuild: drop descend - converting existing users
kbuild: introduce Kbuild.include
kbuild: fix make O=...
kbuild: define clean before including kbuild file
kbuild: KBUILD_VERBOSE was exported twice
kbuild: pass less variables to second make invocation when using make O=...
kbuild: silence mystery message
kbuild: fix building external modules
kbuild: fix make TAGS (for emacs use)
kconfig: move initramfs options to General Setup
Tom Rini:
kbuild: When checking depmod version, redirect stderr
Yum Rayan:
kbuild: restrain output of "make help" to 80 columns
Quoted 10 lines · Sam Ravnborg
Hi Linus.
kbuild updates as accumulated over the last few months. All patches has been in -mm in one or several versions.
Most noteworthy:
1) -Wundef added to CFLAGS. This is the cause of several new warnings, which for the most part has been fixed for now.
2) "PREEMPT" inUTS_VERSION. So we complain when dealing with modules compiled for a wrong kernel
How is this different from the preempt module vermagic?
~$ modinfo agpgart | grep vermagic
vermagic: 2.6.13 preempt gcc-4.0
Quoted 7 lines
3) Introduced Kbuild.include (the cause of most of the changes lines in top-level
Makefile).
It killed a lot of duplicate code
4) Introducedebug_kallsymsto better debug situations where kallsyms fails the consistency check
5) Added support for building rpm tarballs of source
Quoted 8 lines
Some of this was leftovers from my old bitkeeper tree, and authorship is not correct due to wron cogito usage. The changelog message though attributes the correct author.
Please pull from:
www.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.gitSam
[snipped diffstat]
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 12 lines · Alistair John Strachan
Quoted 10 lines · Sam Ravnborg
Hi Linus.
kbuild updates as accumulated over the last few months. All patches has been in -mm in one or several versions.
Most noteworthy:
1) -Wundef added to CFLAGS. This is the cause of several new warnings, which for the most part has been fixed for now.
2) "PREEMPT" inUTS_VERSION. So we complain when dealing with modules compiled for a wrong kernel
Quoted 4 lines
How is this different from the preempt module vermagic?
~$ modinfo agpgart | grep vermagic
vermagic: 2.6.13 preempt gcc-4.0
My bad. Adding PREEMT to UTS_VERSION makes it visible in uname -a.
Sam
Quoted 13 lines · Sam Ravnborg
Quoted 12 lines · Alistair John Strachan
Quoted 10 lines · Sam Ravnborg
Hi Linus.
kbuild updates as accumulated over the last few months. All patches has been in -mm in one or several versions.
Most noteworthy:
1) -Wundef added to CFLAGS. This is the cause of several new warnings, which for the most part has been fixed for now.
2) "PREEMPT" inUTS_VERSION. So we complain when dealing with modules compiled for a wrong kernel
Quoted 7 lines
Quoted 4 lines
How is this different from the preempt module vermagic?
~$ modinfo agpgart | grep vermagic
vermagic: 2.6.13 preempt gcc-4.0My bad. Adding PREEMT to
UTS_VERSIONmakes it visible in uname -a.
I see. I can understand adding an extraversion for SMP and experimental patches (like Ingo's RT work), but why is it useful to differentiate (by name) between preempt and non-preempt kernels? Do distributors wish to package both in parallel?
Signature · 6 lines
Cheers, Alistair. 'No sense being pessimistic, it probably wouldn't work anyway.' Third year Computer Science undergraduate. 1F2 55 South Clerk Street, Edinburgh, UK.
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
How is this different from the preempt module vermagic?
~$ modinfo agpgart | grep vermagic
vermagic: 2.6.13 preempt gcc-4.0
Quoted 7 lines
Quoted 2 lines
My bad. Adding PREEMT to
UTS_VERSIONmakes it visible in uname -a.I see. I can understand adding an extraversion for SMP and experimental patches (like Ingo's RT work), but why is it useful to differentiate (by name) between preempt and non-preempt kernels? Do distributors wish to package both in parallel?
PREEMPT is in UTS_VERSION for the same reason as being present in vermagic. It is one of these options being so fundamentally that we like to both test for it and to know it when runinng uname.
Sam
Quoted 19 lines · Alistair John Strachan
Quoted 18 lines · Sam Ravnborg
Quoted 16 lines · Alistair John Strachan
Quoted 10 lines · Sam Ravnborg
Hi Linus.
kbuild updates as accumulated over the last few months. All patches has been in -mm in one or several versions.
Most noteworthy:
1) -Wundef added to CFLAGS. This is the cause of several new warnings, which for the most part has been fixed for now.
2) "PREEMPT" inUTS_VERSION. So we complain when dealing with modules compiled for a wrong kernelHow is this different from the preempt module vermagic?
~$ modinfo agpgart | grep vermagic
vermagic: 2.6.13 preempt gcc-4.0
Quoted 7 lines
Quoted 2 lines
My bad. Adding PREEMT to
UTS_VERSIONmakes it visible in uname -a.I see. I can understand adding an extraversion for SMP and experimental patches (like Ingo's RT work), but why is it useful to differentiate (by name) between preempt and non-preempt kernels? Do distributors wish to package both in parallel?
I created the patch so that it would show up in oops reports and elsewhere and avoid the inevitable question "was this a preempt kernel?"
Signature · 1 line
Mathematics is the supreme nostalgia of our time.