The visibility problems appear to have been fixed but there is
something seriously wrong with the rulebase.

2.4.16, cml2 1.9.5, kbuild 2.4.  Doing a CML2 run from scratch on a
pristine Linus kernel fails.  We have to handle that case or nobody
outside this group will be able to use CML2.  Even using a good CML1
.config has the same problem.

rm .config* config.out* rules.out
make oldconfig
  python2 -O scripts/cmlconfigure.py -t -B 2.4.16 -I config.out  rules.out
  "config.out", line 289: bad token `SCSI_DEBUG_QUEUES' while expecting symbol.
  File load violated these constraints:
  (((FB == n) and X86) implies ((FONT_8x8 == y) and (FONT_8x16 == y)))
  Attempting recovery from invalid configuration:
  Symbol FB forced to n during recovery attempt.
  Symbol X86 forced to n during recovery attempt.
  Symbol FONT_8x8 forced to n during recovery attempt.
  Symbol FONT_8x16 forced to n during recovery attempt.
  Undoing file loads, recovery failed these constraints:
  (SMP implies (X86 or (PPC or (ALPHA_SABLE or (ALPHA_RAWHIDE or (ALPHA_DP264 or 
(ALPHA_WILDFIRE or (ALPHA_TITAN or (ALPHA_GENERIC or (SPARC32 or (MIPS64 or 
ARCH_S390)))))))))))

That is using arch/i386/defconfig, you end up with a config.out that
has all architectures turned off.  Disabling the architecture in error
recovery is too drastic.  BTW, oldconfig has no -D$(ARCHSYMBOL) like
the other *config rules.

To get around the problem above, I commented out the constraint in
drivers/media/video/rules.cml.  Now I see weird defaults and worse, bad
values being written to config.out.

MTRR: MTRR (Memory Type Range Register) support [m] (NEW)?:

How is MTRR defaulting to 'm', it is a bool?  What is worse, pressing
enter takes the incorrect default and writes MTRR=m to config.out.
Running make oldconfig a second time complains about the bad value then
displays m as a default again.

Eric, before releasing a new CML2 please ensure that it works on a
pristine kernel, without copying a .config from a previous kernel.
make oldconfig on a pristine kernel must work first time, no errors, no
warnings.  Anything else is going to cause a backlash against CML2.


_______________________________________________
kbuild-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/kbuild-devel

Reply via email to