As you can see, I'm back to the old "once or twice a week" kernel
version cycle. The reason is the existence of some silly bugs in pl3
that don't always show up, but can lock up your machine or result in
other weird errors (parse errors in gcc etc). It depends on your
hardware setup - specifically what kinds of interrupts your motherboard
sends, so I (and many others) actually never saw it.
0.99pl4 is a small diff (1300 lines of diff - and it sounds more than it
actually is), and contains only bug-fixes. The changes are:
- include/asm/irq.h: fix the irq handling bug by adding a missing
initialization of a segment register.
- console redirection bug fixed
- Makefile fixes
- support for more than 64 tasks - it was there before, but the GDT
overflowed with NR_TASKS more than about 120..
- ptrace() - the bogus ESRCH bug is corrected (fix by Branko Lankester?
Hope I haven't messed up the attributions *again*).
- flock() support by adding two new ioctl() values - hlu.
- seagate.c patches (Eric Youngdale) to fix the problem with some disks
off the ST-0x controller.
- tcp/ip volatile fixes (sorry - lost the attribution)
- remount patches by almesber
I'm getting happier about the kernel again - but I have got several
comments on the line "I'll wait for 1.0", so I'm starting to think I
won't get 0.99 debugged due to lack of tester interest. I guess I
should have thrown out 1.0 totally unexpectedly instead of warning about
it. I can only repeat my plea for people trying it out: if you don't
try it out (or don't report problems to me), you may find that 1.0 won't
work for you... Also, if you find that pl4 doesn't fix a bug that you
already reported, please take the time to report it again by mailing me.
Otherwise I'll probably assume that I found them all.
And, in case you can't remember, pl4 is naturally available both as full
source and diffs against pl3 at nic.funet.fi: pub/OS/Linux/PEOPLE/Linus,
and will also probably show up at the other major sites within a day or