Fwd: cml2-2.1.3 bug report

[jeff millar <jeff@wa1hco.mv.com>]

I sent this to Eric privately with a lot of data, but maybe someone else can
jump in with an opinion about they way make oldconfig _should_ work.

Several items...

1. Make oldconfig doesn't behave like cml1.  With cml1, we would patch the
kernel or copy in a backup .config file and run make oldconfig.  It would do
its thing for 30 seconds and be done...no (or few) questions asked.  With
CML2, I get an endless series of questions about each possible configuration
item...much like "make config" does.

I just want make oldconfig to get ready for make dep with the minimum of
fuss. It's ok to ask about new configuration items but not a bunch of stuff
already set.

2. Something triggers the building of "everything" modular, like _all_ the
network cards, I2O, IEEE1394, etc.  You implied in an earlier email that
selecting modules leads to this.  But, I just want to build a few modules
because it takes less time and disk space and it easier to audit
/lib/modules/2.4.x/ results.   Besides with bleeding edge kernels, often lots
of drivers don't compile correctly. My original .config has m where I want it
and setting m a bunch more places means I just have to go around and shut off
hundreds of things.