I'll try those on a known-good kernel tree (2.6.23 with Geert's quilt patches^^ Is that correct?
Yup, that's the latest git tree with quilt that I had floating around on my powerbook. In case you didn't notice, I've been quite inactive on the kernel front in the last year :-) The last I had successfully run on the Falcon was 2.6.19-m68k, even. After losing one SCSI filesystem to kernel hacking, I decided not to push it any further. The same 2.2.23 that boots on ARAnym does not boot even as far as atafb init on the Falcon.
didn't boot on real hardware but does on Aranym, fwiw). 2.6.25rc9 vanilla still has the byte ordering problem on IDE, is that what your ide bug patch is about?2.6.25 indeed still needs 141-ide.diff. 2.6.26-rc8 doesn't need it anymore, at least not on ARAnyM.
ARAnyM is pretty good in that respect. But I'll definitely try on the hardware itself.
Duh - cc1: error: unrecognized option `-fno-stack-protector' again. Can we have something like KBUILD_CFLAGS := $(subst -fno-stack-protector,,$(KBUILD_CFLAGS)) -g in arch/m68k/Makefile, dependent on ancient cross compiler versions?
It would be nice if you could try 2.6.26-rc8 on real hardware, as I use ARAnyM for most testing these days, except for the casual boot on real Amiga hardware. atari_defconfig would be a good one to try, followed by multi_defconfig, or (for the brave) multi_defconfig modified to have all drivers built-in.
I hear you ... will go and find a serial nullmodem cable somewhere to debug the general boot problems.
Speaking of IDE bug: the IDE lock_release bug is indeed gone for good even when testing on real hardware.Good! I'll add it.
Fine. I'll try to get my remaining patches updated to apply to 2.6.26 over the next weeks.
Michael