Re: [kbuild-devel] kbuild update
On Sun, Sep 16, 2007 at 12:23:43PM -0700, Andrew Morton wrote: > On Sun, 16 Sep 2007 08:58:03 -0400 (EDT) "Robert P. J. Day" <[EMAIL > PROTECTED]> wrote: > > > On Sun, 16 Sep 2007, Sam Ravnborg wrote: > > > > > A summary of what is planned to be submitted in next merge window for > > > kbuild. > > > The shortlog below have additional details but the headlines are: > > ... > > > o add script to find unused kconfig symbols (try it!) > > > > based on my experiences with this, unless you filter carefully, you're > > going to end up with a *whack* of false positives given the number of > > developers who elect to name their local macros starting with a prefix > > of "CONFIG_". good luck dealing with *that*. :-) > > box:/usr/src/linux-2.6.23-rc6> grep -r '^[ ]*#[]*define[ > ]*CONFIG_' . | wc -l > 415 > > bah. They're all bugs - The CONFIG_foo namespace is (should be) reserved in > kernel coding. I get (after a bit more filtering 349 hits of which 182 are outside drivers/ Of the 182 the 25 of them are in arch specific code. I se no good reasons to address this unless we touch code in that area anyway. But avoiding new entries are good. Sam - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, 16 Sep 2007, Sam Ravnborg wrote: > On Sun, Sep 16, 2007 at 08:58:03AM -0400, Robert P. J. Day wrote: > > On Sun, 16 Sep 2007, Sam Ravnborg wrote: > > > > > A summary of what is planned to be submitted in next merge window for > > > kbuild. > > > The shortlog below have additional details but the headlines are: > > ... > > > o add script to find unused kconfig symbols (try it!) > > > > based on my experiences with this, unless you filter carefully, you're > > going to end up with a *whack* of false positives given the number of > > developers who elect to name their local macros starting with a prefix > > of "CONFIG_". good luck dealing with *that*. :-) > This is useful for specific people but not for kernel janitorial > fodder. As you already experience the amount of false positive are > too high. > > But for a developer of a module it is no deal to ignore the false > positives. > > It is for godd reasons not integrated in the build process. right, i agree completely. if it were absolutely reliable in generating *only* actual unused symbols, then that *might* (i stress, *might*) be a viable argument for making it part of the Kbuild system. but, really, if you can whip off a script that does most of the job in a few minutes, there's really no motivation to then spend hours trying to nail those last few harmless bits of irrelevant output. rday -- Robert P. J. Day Linux Consulting, Training and Annoying Kernel Pedantry Waterloo, Ontario, CANADA http://crashcourse.ca - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, Sep 16, 2007 at 08:58:03AM -0400, Robert P. J. Day wrote: > On Sun, 16 Sep 2007, Sam Ravnborg wrote: > > > A summary of what is planned to be submitted in next merge window for > > kbuild. > > The shortlog below have additional details but the headlines are: > ... > > o add script to find unused kconfig symbols (try it!) > > based on my experiences with this, unless you filter carefully, you're > going to end up with a *whack* of false positives given the number of > developers who elect to name their local macros starting with a prefix > of "CONFIG_". good luck dealing with *that*. :-) This is useful for specific people but not for kernel janitorial fodder. As you already experience the amount of false positive are too high. But for a developer of a module it is no deal to ignore the false positives. It is for godd reasons not integrated in the build process. Sam - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, 16 Sep 2007, Oleg Verych wrote: > On Sun, Sep 16, 2007 at 08:58:03AM -0400, Robert P. J. Day wrote: > > On Sun, 16 Sep 2007, Sam Ravnborg wrote: > > > > > A summary of what is planned to be submitted in next merge > > > window for kbuild. The shortlog below have additional details > > > but the headlines are: > > ... > > > o add script to find unused kconfig symbols (try it!) > > > > based on my experiences with this, unless you filter carefully, > > you're going to end up with a *whack* of false positives given the > > number of developers who elect to name their local macros starting > > with a prefix of "CONFIG_". good luck dealing with *that*. :-) > > I saw you patches about that kconfig symbols usage, your > developments of *the* script. I also happened to see > `linux-2.4.X/checkconfig.pl`. Now somebody else proposed something > related to kbuild tree. So, what was your point in doing that, why > your work didn't ended up in `linux/scripts/'? i've never looked at checkconfig.pl, i didn't even know it existed. i just wrote a simple shell script that scanned for the obvious occurrence of "CONFIG_FUBAR" symbols for which there is no corresponding "FUBAR" definition in a Kconfig* file somewhere. i think i can guarantee that i won't *miss* any but, as i've mentioned before, i definitely display a lot of false positives based on developers' less-than-ideal choice of macro names, and i just haven't put any effort into filtering those out since it doesn't seem worth the time. as for adding a check like that to the Kbuild structure itself, i get the feeling that you'd run into the same philosophy as adding a debugger to the kernel would get you. i don't think it's worth over-complicating the Kbuild structure -- i think it's just easier to leave stuff like that as external, cheap, one-off utility scripts. rday -- Robert P. J. Day Linux Consulting, Training and Annoying Kernel Pedantry Waterloo, Ontario, CANADA http://crashcourse.ca - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, Sep 16, 2007 at 08:58:03AM -0400, Robert P. J. Day wrote: > On Sun, 16 Sep 2007, Sam Ravnborg wrote: > > > A summary of what is planned to be submitted in next merge window for > > kbuild. > > The shortlog below have additional details but the headlines are: > ... > > o add script to find unused kconfig symbols (try it!) > > based on my experiences with this, unless you filter carefully, you're > going to end up with a *whack* of false positives given the number of > developers who elect to name their local macros starting with a prefix > of "CONFIG_". good luck dealing with *that*. :-) I saw you patches about that kconfig symbols usage, your developments of *the* script. I also happened to see `linux-2.4.X/checkconfig.pl`. Now somebody else proposed something related to kbuild tree. So, what was your point in doing that, why your work didn't ended up in `linux/scripts/'? > rday _ - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, 16 Sep 2007, Sam Ravnborg wrote: > A summary of what is planned to be submitted in next merge window for kbuild. > The shortlog below have additional details but the headlines are: ... > o add script to find unused kconfig symbols (try it!) based on my experiences with this, unless you filter carefully, you're going to end up with a *whack* of false positives given the number of developers who elect to name their local macros starting with a prefix of "CONFIG_". good luck dealing with *that*. :-) rday -- Robert P. J. Day Linux Consulting, Training and Annoying Kernel Pedantry Waterloo, Ontario, CANADA http://crashcourse.ca - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
[kbuild-devel] kbuild update
A summary of what is planned to be submitted in next merge window for kbuild. The shortlog below have additional details but the headlines are: o Make the kernel buildable on a cygwin box (but not without a few glitches) o detect unterminated device id list o add export_report as a valid top-level make target o add script to find unused kconfig symbols (try it!) o allow a user to specify additional flags on the commandline with CFLAGS, AFLAGS or CPPFLAGS. The flags will be appended to the kernel defiend flags The CFLAGS change will conflict with Andi's x86 patches and I fix it before pushing if Andi's patches are pushed first. Otherwise no conflicts. The vmlinux.lds changes that I have sent to a number of arch maintainers lately are not included. I expect them to be applied by the respective arch maintainers. The --gc-section cleanup from Denys Vlasenko will be included in kbuild after the merge window so we can try it out in -mm before hitting mainline. Jan Beulich kconfig patch will most likely be included. Need to test it first. Marej menuconfig improvement may also be included. Last patch looked good but have not made up my mind if it really help people. Patched referred below are available at git.kernel.org at kbuild.git in an hour or so. Sam Adrian Bunk (2): kbuild: fix export_report.pl kbuild: call export_report from the Makefile Jesper Juhl (3): kbuild: improve scripts/gcc-version.sh output a bit when called without args kbuild: ver_linux fix glibc version print kbuild: scripts/ver_linux : correct printing of binutils version Kees Cook (1): kbuild: make modpost detect unterminated device id lists Mike Frysinger (1): kbuild: clean Modules.symvers in external module dirs Paolo 'Blaisorblade' Giarrusso (1): kbuild: script to check for undefined Kconfig symbols Petr Stetiar (1): kbuild: fix segfault in modpost Ram Pai (1): kbuild: fix perl usage in export_report.pl Randy Dunlap (1): docproc: style & typo cleanups Sam Ravnborg (10): kbuild: Use Elfnn_Half as replacement for Elfnn_Section kbuild: check if we can link gettext not just compile kbuild: fix genksyms Makefile kbuild: __extension__ support in genksyms (fix unknown CRC warning) kbuild: apply genksyms changes kbuild: enable 'make CFLAGS=...' to add additional options to CC kbuild: enable 'make AFLAGS=...' to add additional options to AS ia64: fix sn to add include files using EXTRA_CFLAGS kbuild: enable 'make CPPFLAGS=...' to add additional options to CPP kbuild: fix directory traversal bug Shlomi Fish (1): kconfig: qconf ("make xconfig") Search Dialog Enhancement - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
On Sun, Jun 03, 2007 at 05:04:39PM -0400, Chris Smith wrote: > >What's next? > >... > >-> A few cygwin specific changes may hit kbuild - to allow the kernel to > >be built > > using cygwin on a windoze box. > > This seems vaguely immoral. ;) Especially embedded people like to build on Windoze boxes. Sam - This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
> What's next? > ... > -> A few cygwin specific changes may hit kbuild - to allow the kernel to be > built >using cygwin on a windoze box. This seems vaguely immoral. ;) Thanks, Chris - This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
Re: [kbuild-devel] kbuild update
Hmm, the patch serie should have been chained. Sam - This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel
[kbuild-devel] kbuild update
Short intro to content of kbuild.git at the moment. See http://git.kernel.org/?p=linux/kernel/git/sam/kbuild.git;a=summary Everything are targeted for next merge window. section mismatch kbuild: make better section mismatch reports on i386, arm and mips kbuild: The .paravirtprobe section is obsolete, so modpost doesn't need to handle it kbuild: ignore section mismatch warnings originating from .note section kbuild: refactor code in modpost to improve maintainability kbuild: warn about references from .init.text to .exit.text kbuild: remove hardcoded apic_es7000 from modpost kbuild: refactor code in modpost kbuild: remove hardcoded _logo names from modpost kbuild: suppress modpost warnings for references from the .toc section as used by powerpc kbuild: whitelist references from variables named _timer to .init.text Several changes to either improve reporting - or to ignore false positives. In particular modpost now check for and warn about references from .init.text to .exit.text This check caugth several real bugs - which for the most part are already fixed in upstream. modpost now does a better job finding the symbols using addend - this works for me but had caused troubles in the past. If modpost suddenly bails out with a segmentation violation this may the the culprint. In addition some trivial code refactoring was done to improve maintainability of modpost. kconfig: use POSIX equality test in check-lxdialog.sh Bugfix. scripts: Make cleanfile/cleanpatch warn about long lines Improved so we all adhere to 80 lines (in)sanity kbuild: avoid environment to set variables used by kbuild I had a report that a user tried to do: EXTRA_CFLAGS=FOO make fs/nfs/ This is not supposed to impact the build so make sure it does not. kbuild: add support for reading stdin with gen_init_cpio kbuild: add support for squashing uid/gid in gen_initramfs_list.sh initramfs improvements kbuild: asm-offsets.h is now cleaned with O=.. builds Bugfix kbuild: document cc-fullversion kbuild: New 'cc-fullversion' macro powerpc: Refuse to build 64-bit with GCC-4.2.0 and CONFIG_MODULES Add and use cc-fullversion to obtain exact gcc version See documentation in Documentation/kbuild/makefiles.txt What's next? -> davem has requested to add a new section used for code modifying support. The actual section name to use needs to be sorted out. -> A few cygwin specific changes may hit kbuild - to allow the kernel to be built using cygwin on a windoze box. -> A few trivial kconfig changes are pending. Needs to review them before they are ready. -> A speed-up of headers_check is in the works. Got stalled but will look into it soon. The check goes down to ~3 sec on my not-so-fast machine -> If I get really bored I may start a 'cleaning up all vmlinux.lds files' round. If someone would help out here it would be appreciated. These change will then be merged via the arch-maintaines and not really kbuild stuff.. Sam - This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ ___ kbuild-devel mailing list kbuild-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kbuild-devel