Re: 2.6.1-mm5 (compile stats)

From: Adrian Bunk (bunk_at_fs.tum.de)
Date: 01/20/04

  • Next message: Richard B. Johnson: "Re: [1/4] better i386 CPU selection" 
    Date:	Tue, 20 Jan 2004 23:27:00 +0100
To: John Cherry <cherry@osdl.org>

    On Tue, Jan 20, 2004 at 02:15:10PM -0800, John Cherry wrote:
    >...
    > > Regarding allnoconfig:
    > > allnoconfig is a completely pathological case. It says "n" to support
    > > for ISA, MCA and PCI, and neither networking nor any block devices.
    > > Besides, it says "n" to ELF, a.out and other binary formats.
    > > Demanding that allnoconfig should compile (although the resulting kernel
    > > is completely useless) sounds a bit like demanding that no change in the
    > > kernel is allowed to cause regressions in the dbench results...
    > > It is useful to omit a common option like e.g. PCI and check whether the
    > > kernel still compiles, but allnoconfig removes nearly everything and
    > > compiles such a small part of the kernel, that it's hardly useful.
    >
    > I realize that allnoconfig is pathological, but it has caught several
    > config errors. One would never try to boot from such a config. Builds
    > based on allnoconfig have one purpose and that purpose is to validate
    > that defines are not used in cases where they are NOT defined in the
    > configuration. Developers will quite often code a feature or
    > architecture with the config parameters always ON. When the config
    > option is turned OFF, I will find compile errors, undefined variables,
    > and the like. This is actually quite a valuable screen.

    The problem is that allnoconfig turns _everything_ off.

    Cases like e.g. CONFIG_PROC_FS=n are interesting, but allnoconfig
    doesn't really test them since allnoconfig also says "n" to all drivers.

    > If developers feel that this has outlived its usefulness, I'll remove it
    > from the compile regressions. However, all I have received at this
    > point have been requests to put an allnoconfig build into the
    > regressions.

    I'd like to hear from the people requesting it why they consider it
    useful.

    In my personal experience, compiling allyesconfig but with CONFIG_SMP=n
    (which enables BROKEN_ON_SMP drivers), and compiling with gcc 2.95 are
    more interesing (and more realistic) configurations than allnoconfig
    that find many compile errors.

    > John

    cu
    Adrian 

    -- 
       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
  • Next message: Richard B. Johnson: "Re: [1/4] better i386 CPU selection"

    Relevant Pages

    • Re: 2.6.1-mm5 (compile stats)
      ... >> against defconfig, allnoconfig, allyesconfig, and allmodconfig. ... > Demanding that allnoconfig should compile (although the resulting kernel ... One would never try to boot from such a config. ...
      (Linux-Kernel)
    • Re: [PATCH] make miniconfig (take 2)
      ... make miniconfig and make allnoconfig are ... Make allnoconfig is supposed to make a config ... various other bits of the kernel build don't give me an error ...
      (Linux-Kernel)
    • Re: stuck with 2.6.23.14 on x86_64
      ... the lastest that I can compile is 2.6.23.14; ... All .config are the same, ... SCSI controller: ... Right now, on AMD64 machines, I'm forced to patch by hand the kernel, ...
      (Linux-Kernel)
    • Re: [PATCH] make miniconfig (take 2)
      ... Anything fancy based on diff runs into the fact that dependencies can enable ... And I try to avoid perl dependencies in the linux build process. ... > add the ability to use .config and rewrite that when exiting. ... The dominant time factor is running allnoconfig. ...
      (Linux-Kernel)
    • Re: Linux hangs on nVidia nForce2 400 Ultra
      ... According your .config you have compiled all ide-chipsets. ... Have you tried only compile in the ... AMD74XX AMD and nVidia IDE support? ...
      (Linux-Kernel)