Silence missing CONFIG_ symbol noise
PATCH 1/1 Silence missing CONFIG_ symbol noise
For sparc/sparc64, there are CONFIG_ symbols which are undefined. On sparc64, CONFIG_AGP and CONFIG_AGP_INTEL are undefined. On sparc, CONFIG_I2C, CONFIG_I2C_ALGOBIT, CONFIG_AGP, CONFIG_AGP_INTEL, CONFIG_VIDEO_DEV, and CONFIG_ISP1301_OMAP are undefined. This causes errors like "config:1234: trying to assign nonexistent symbol FOO" to be thrown when the kernel config tools are run.
The reason that these CONFIG_ symbols are undefined is because the hardware isn't supported by sparc/sparc64, but other supported hardware resides in the same .config files pulled in by the various driver Kconfig files.
Rather than hacking up the driver Kconfig files, I've created a new arch-specific Kconfig.undefined file, where these (and future) undefined symbols can be added for sparc/sparc64. This Kconfig.undefined file is sourced by arch/{sparc,sparc64}/Kconfig, before the driver Kconfigs are sourced.
The end result is that the undefined CONFIG_ symbol errors are not thrown when the config utilities are run. The errors are not fatal, but they do cause confusion when users are building kernels for sparc/sparc64.
~spot
---
Tom "spot" Callaway: Red Hat Sales Engineer || GPG Fingerprint: 93054260 Fedora Extras Steering Committee Member (RPM Standards and Practices) Aurora Linux Project Leader: http://auroralinux.org
Lemurs, llamas, and sparcs, oh my!
arch/sparc64/Kconfig renamed from arch/sparc64/Kconfig.BAD +2 -0
@@ -579,6 +579,8 @@ config DRM_R128 endmenu +source "arch/sparc64/Kconfig.undefined" + source "drivers/input/Kconfig" source "drivers/i2c/Kconfig"
arch/sparc64/Kconfig.undefined renamed from arch/sparc64/Kconfig.undefined.BAD +10 -0
@@ -0,0 +1,10 @@ +# Little hack to cleanup undefined symbols +# Note: These are not viable config options for this architecture. + +config AGP + bool + depends on BROKEN + +config AGP_INTEL + bool + depends on BROKEN
arch/sparc/Kconfig renamed from arch/sparc/Kconfig.BAD +2 -0
@@ -367,6 +367,8 @@ config UNIX98_PTY_COUNT endmenu +source "arch/sparc/Kconfig.undefined" + source "drivers/input/Kconfig" source "fs/Kconfig"
arch/sparc/Kconfig.undefined renamed from arch/sparc/Kconfig.undefined.BAD +27 -0
@@ -0,0 +1,27 @@ +# Little hack to clean up undefined symbols +# Note: These are not viable config options for this architecture. + +config I2C + bool + depends on BROKEN + +config I2C_ALGOBIT + bool + depends on BROKEN + +config AGP + bool + depends on BROKEN + +config AGP_INTEL + bool + depends on BROKEN + +config VIDEO_DEV + bool + depends on BROKEN + +config ISP1301_OMAP + bool + depends on BROKEN +
Quoted 5 lines · Tom 'spot' Callaway
The end result is that the undefined
CONFIG_symbol errors are not thrown when the config utilities are run. The errors are not fatal, but they do cause confusion when users are building kernels for sparc/sparc64.
Another approch would be to just skip the warnings when reading in a defconfig file. I've never seem them really used.
Sam
David S. Miller
maintainerQuoted 8 lines · On Fri, 4 Mar 2005 23:16:33 +0100 Sam Ravnborg <[email protected]> wrote:
Quoted 5 lines · Tom 'spot' Callaway
The end result is that the undefined
CONFIG_symbol errors are not thrown when the config utilities are run. The errors are not fatal, but they do cause confusion when users are building kernels for sparc/sparc64.Another approch would be to just skip the warnings when reading in a defconfig file. I've never seem them really used.
I don't think I'm going to apply Tom's patch for now.
This really sounds like the defconfig files need to be regenerated. Right?
Quoted 9 lines · David S. Miller
Quoted 6 lines · On Fri, 4 Mar 2005 23:16:33 +0100 Sam Ravnborg <[email protected]> wrote:
Quoted 5 lines · Tom 'spot' Callaway
The end result is that the undefined
CONFIG_symbol errors are not thrown when the config utilities are run. The errors are not fatal, but they do cause confusion when users are building kernels for sparc/sparc64.
Quoted 7 lines
Quoted 2 lines
Another approch would be to just skip the warnings when reading in a defconfig file. I've never seem them really used.
I don't think I'm going to apply Tom's patch for now.
This really sounds like the defconfig files need to be regenerated. Right?
kconfig throws a warning when it see:
select I386_FOO
in sparc context, because the config symbol I386_FOO is not defined. We do not see this with I386, cause people make sure to define all selected CONFIG symbols. But most other architectures suffer from this.
IIRC Tom's patch just cleaned up select usage or similar and regenerating defconfig would not help us here.
But the better solution would be to drop the warning from kconfig. I will do that next time I get around to do some kernel hacking.
Sam
David S. Miller
maintainerQuoted 6 lines · On Mon, 25 Apr 2005 07:02:50 +0200 Sam Ravnborg <[email protected]> wrote:
kconfig throws a warning when it see:
select I386_FOOin sparc context, because the config symbol
I386_FOOis not defined. We do not see this with I386, cause people make sure to define all selected CONFIG symbols. But most other architectures suffer from this.
Thanks for the explanation.
Quoted 2 lines
But the better solution would be to drop the warning from kconfig. I will do that next time I get around to do some kernel hacking.
Ok, thanks again.