Problems (as usual)

[Konrad Kokoszkiewicz <draco@piwo.bl.pg.gda.pl>]

Hello,

some time ago I've changed gas to version 2.5.1. This immediately started
to make me troubles with compiling assembler modules for the MiNT kernel;
namely, it apparently refuses to compile them with -m68000 option. Since
the modules contain 68030/68040/68882 stuff, the new gas used to complain
much about it, saying "wrong architecture" and tended to abort the
compilation.

I've fixed this by changing assembly options (only for gas, nothing for
the C compiler) to -m68030 -m68881. However, what I discovered today, MiNT
kernels compiled such way for 68000 (make clean; make mint.prg) refuse to
work on 68000 machines... I tried it on two of them, one 1040STE, 4 meg
RAM, 170 MB IDE disk, the other 1040STFM, 4 meg RAM, no hard drive.  Both
TOS 2.06. The kernel locks up while booting, I've even tracked the problem
down to find that it hangs in init_intr(). The same kernel works extermely
fine on my Falcon030, 68882, 14 meg RAM, 1.5 gig IDE. 

Since the kernel works on 68030 machine, I suspect the problem may be in
one of assembler modules compiled with -m68030 -m68881. Especially, it may
be intr.spp which contains the scheduler and generally VBL stuff. Does
anyone have a clue on it? Does gas 2.5.1 _really_ "optimize" stuff while
compiling, as someone here suggested once? The lockup looks exactly like
the kernel had some interrupt problems on ST, but I won't have a time
until tomorrow afternoon to do further research on it.

By the way, I've installed such "non-working" 68000 kernel on my Falcon
for test purposes. It obviously worked fine, but I also noticed, that mp2
player runs more smoothly when the 68000 kernel is installed in
opposition to a 68030 kernel. Namely, the known problem with mp2 player
(and Background player for DVSM tracks) suddenly becoming silent due to an
AES lock when that last is left alone for some time, seems to be much less
annoying. It looks like it occurs e.g. four times per hour with 68030
kernel or once/twice a day with 68000 kernel. Could someone verify that?

Thx,

Konrad M.Kokoszkiewicz
mail:draco@bl.pg.gda.pl
http://www.orient.uw.edu.pl/~conradus/

** Quem Iuppiter vult perdere, dementat prius.
*******************************************************
** Kogo Jowisz chce zgubic, temu wpierw rozum odbiera.