Problems with 1.1 AMIX

[Rob Healey]

	After two days of "fun" filled spelunking of AMIX 1.1 I have my
	first whine list. Anybody who knows how to deal with these problems
	PLEASE speak up, either in private e-mail or posting to the group.
	First the whining, then the praise.

Enter whine mode:

	1) Disk partitioning.
	   The 386 R4 install procedure lets you create many UNIX partitions,
	   aka slices, in any slice # and size you want. The 1.1 install
	   scripts seems to limit you to 3 stock UNIX and 1 AmigaDOS
	   slice. I whined earlier about this before the system arrived.
	   It looks like I'll have to create a custom boot set that will
	   allow me to run rdb from floppys to set up all 7 of the slices
	   the way I want them. Now, I'm the type who used iv and a
	   custom boot set on a UNIX PC to satisfy my perverse need to
	   create a disk setup just the way I wanted it. I take it more
	   as a challenge than a bug... To the Amiga folks, you might
	   want to consider that SOME of the AMIX users know EXACTLY
	   what their doing and would like the rope to hang themselves
	   with... A more flexible install partitioning script would
	   be appreciated in future releases. Maybe you can even pull
	   an AT&T "magic mode" a la 3b2 UNIX!

	2) MotherBoard serial driver. Enter face filled with disgust mode:
	   Come on guys, this is an AMIGA for crying out loud. There is
	   zero, zilcho, zippo, nadda excuse for a serial port that drops
	   characters at 9600 baud like it's going out of style. What's
	   worse, the serial driver PANIC'D MY SYSTEM from a USER mode
	   kermit, ckermit 5a(169) alpha to be exact, and I wasn't
	   even running as root! I'll admit I'm spoiled when it comes to
	   serial performance due to where I work but come on people. A
	   thousand lashings with a wet noodle on this one. I hope 2.0
	   has a better serial driver for the motherboard port.

	   In the mean time, is there any way I can increase the buffering
	   or something so that I'm not wasting a telebit T2500? I'd
	   really like to be able to do 19200. My UNIX PC can do 9600
	   full bore with no drop outs on a mid 80's serial driver; I'm
	   disappointed with AMIX's performance... B^(.

	   The MOST frustrating thing is that the system panic didn't
	   do a panic dump on the paging area even though there was
	   enough blocks to do so. How am I supposed to psychoanalyse
	   a panic dump with crash when the kernel doesn't have the common
	   decency to make one? I use crash at work quite a bit so I
	   know what to do once I get the dump. Can I set any magic
	   variables to tell AMIX to dump to swap?

	 3) X needs pty 0 free in order to startup??? HUH? I'd
	    REALLY like to know why this is. I've worked with quite a
	    few X servers and most recently X386. NONE of them require
	    a pty, let alone the FIRST pty, to be free in order to work.
	    I can't wait to hear the reason for this one...

	 4) The berkeley emulation from hell:

	    Sigh. I have a bitch of a time with the 386 S5R4's pathetic
	    emulation of berkeley routines and I see it infected the 68k
	    code as well. I get the feeling the AT&T people who did the
	    Berkeley stuff never went within 2000 miles of a functioning
	    Berkeley system. B^(. I'll keep this short though.

	    1) If /usr/ucb is in your path before /usr/ccs/bin you are SOL.
	       gcc won't work and cc produces executables that loath executing.
	       The fix is easy, as the man says, if it hurts when you do
	       that, then DON'T do that...
	    2) The directory functions in the berkely libs, i.e opendir and
	       co., don't know squat about symlinks or non-S5 file systems.
	       The people doing the ufs patchwork should make sure that the
	       regular lib's dir functions grok S5, ufs, NFS, RFS, etc. and
	       such and maybe fix the berkeley stuff if they have time. This
	       botch is common to all R4's I've seen so it's not C='s fault
	       per say.
	    3) The whole Berkeley emulation package on all R4's really needs
	       to be taken out into a field and shot. Rewriting it from scratch
	       would probably take less time than trying to fix the current
	       mess. B^(. Again, this is not unique to AMIX, all R4's seem
	       to have braindead Berkeley emulation. B^(. Probably AT&T's
	       little way to get you to kick the Berzerkeley habit. B^(.

	 5) gcc gripes.

	    Being a UNIX PC owner myself I was tickled pink that AMIX gcc
	    is basically 3b1 gcc with dwarf tacked on. Will 2.0 have 
	    descriptions that aren't based on 3b1 but totally optimized for
	    AMIX? Seems a shame to retain 3b1 quirks that we don't have to...
	    I am glad to see active support for gcc though, I HATE the AT&T
	    R4 compiler on both AMIX and 386 R4. Thanks to a gentleman from
	    Dell I have a gcc for 386 R4 but it doesn't generate DWARF output
	    so I can't debug with gcc. Which leads me to gdb...

	 6) Will there be a gdb in 2.0? I know Dell ships gcc and gdb with
	    their R4 so a R4 gdb is possible. The conversion to /proc has
	    GOT to be a pain in the butt... I HATE sdb with an intense
	    passion, gdb would be moocho appreciated. Maybe some of the
	    C= guys can milk the Dell boys for info, a couple a kegs ought
	    to do it... B^).

End whine mode
Enter back pat mode.

	15 minutes from box to fully functioning networked box on my company's
	network! YOWSA! I'm impressed! I've NEVER seen a UNIX system come
	up so fast and "just work"(tm)! GREAT JOB C=!!!! After 20 minutes
	I was using rrn and sendmail e-mail over the net. A very pleasant
	experience. Now, if only I could figure out the $%^$%^ mailer system,
	I thought sendmail was terse; YIKES!


	If anyone has comments on any of the above, please send me some email
	or better yet POST. Let's get some REAL AMIX conversations going, not
	like these boring application postings I've been seeing of late.

	I managed to get ckermit 5A(169) Alpha to work. ELM 2.3PL11 is
	bucking it, can't seem to find all the libs due to the /usr/ccs BS.
	I might try merging in the DWARF stuff to gcc 1.39 but I assume the C=
	crew has already done that and might skip straight to getting the R4
	clients and libs compiled. I HATE Open Look, I want my TWM, tvtwm and
	R4 athena widgets dammit!

	So, who all's ported what and what should we port next to our
	favorite 68k box? Let's get something going here!

	By the way, where the heck are all the Virginia U people at? They
	MUST have interesting war stories to tell us greenhorns. B^).

		Happily hacking on my 3000UXD,

			-Rob
-- 

Rob Healey                                          rhealey at digibd.com
Digi International (DigiBoard)
Eden Prairie, MN                                    (612) 943-9020