At 03:24 PM 12/8/99 -0600, you wrote:
>On Wed, 08 Dec 1999, you wrote:
>> I did a nice clean install of 6.1 (2.2.13-7mdk) to my new K7. Worked like
>> a champ. After it was up and (mostly) configured, I clicked on Updates and
>> noted an update to kernel 2.2.13-22mdk. I said, "SURE!" It downloaded. I
>> did 'make depend'. Fine. I then did 'make bzImage'. It got all the way
>> to the bind phase and then died with an unresolved symbol
>> _buggy_fxsr_alignment. This was in (as I recall, and machine is not
>> available right now), a bugs.h file? It seems to have something to do with
>> a misaligned field in a data structure when doing fast floating
>> save/restore. So, I turn off that option in the config and repeat. This
>> time it binds okay, but the system panics on boot. Any ideas?
>
> Prob'ly just more questions. I don't/didn't know you were
>supposed to do anything with a 'ready made kernel' other than
>install the rpm and edit lilo.conf, run lilo and boot it ? I did a
>few that way, -22mdk was one, no problems.
>
> I compiled one from source, I believe it was -7mdk with no
>errors, edited lilo.conf, ran lilo and booted to it. I did this
>one, 'cause being a newbie, I thought I ought'a try one. Beginner's
>luck! A few I tried after that, no dice .... errors during 'make'
>usually. Being a newbie I retraced my steps, where did i go wrong?
>Then it dawned on me. Besides being a newbie, I'm an obsessive
>o/c'r. I have a fixation on runnin pc100 ram at 135mhz cas2
>timings (2-3-3, precharge on). Altho I can run anything in Windows
>clocked to the hilt (p3-450 @ 608), and most everything in Linux,
>if I just reduce my ram timing to 3-3-3, compiles work, no problems.
>
> I know it's a hard pill to swallow, specially if you built the
>system yourself, and even worse if you paid Dell to do it for you
>but ..... Well, a lot of these kinds of problems I read about on
>lists and NG's, I jus' havt'a suspect it's hardware problems. O/c'd
>or not, and specially if there's a CuMine/i8whatever or a K7/athlon
>motherboard involved. And, of course, anything o/c'd.
>
> I just havt'a believe that the newer, faster systems, 'specially
>since both the Cu/i8xx and Athlon/mobo/voltages systems have ram
>issues, that might be a good place to start lookin at. See if
>setting your ram to 3-3-3, precharge disabled, won't get you thru a
>kernel compile (from source).
thanks for the tips, but i don't see how it can possibly be hardware. a
stock 7mdk kernel builds from source and runs fine. a stock 22mdk kernel
fails with a compiler assertion (NOTE: NOT a SEGV!) the assertion is due
to something that makes complete sense.
