[Prev][Next][Index][Thread]

RE: Sphere

My 1.5 cents:

Wow. I took a look at your website... and you're even more ambitious than I
am... I just want to rule the world.

>> BERNARDINI TURRENS, RICARDO wrote:
>> I've built many basic kernels reading about 15% of the documentation
>> and the examples. I'm not a genius either.
>
>You are, in my book...
>

Have you tried playing? Make a little kernel. Not your
mammoth-new-operating-system-that-will-revolutionize-the-world-and-make-Tux-
hide-in-shame-and-Bill-buy-you-out -- just a toy kernel. Hello world+?

>I tried to build the thing under BeOS (A 30 minute download on dialup)
>and this is the error I got running "make":

BACKUPS!!! Download once, back it up to some removeable media, even X
floppies... And make a friend with a fast connection.

>gcc -c -o cpuid.o -MD -DHAVE_CONFIG_H  -DOSKIT_X86 -DOSKIT_X86_PC -I.
>-I../kern/x86 -I../kern/x86/pc -I../kern/x86/dos -I../kern  -I-
>-I../oskit/c -I.. -I.. -nostdinc -Wall -fno-strict-aliasing -O2 -g
>../kern/x86/cpuid.c
>In file included from
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c:25:
>/boot/home/Sourcecode/oskit-20010214/oskit/c/string.h:48: warning:
>conflicting types for built-in function `memset'
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c: In function
>`get_cache_config':
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c:44: Invalid `asm'
>statement:
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c:44: fixed or
>forbidden register 3 (bx) was spilled for class BREG.
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c: In function
>`cpuid':
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c:96: Invalid `asm'
>statement:
>/boot/home/Sourcecode/oskit-20010214/kern/x86/cpuid.c:96: fixed or
>forbidden register 3 (bx) was spilled for class BREG.
>make[1]: *** [cpuid.o] Error 1
>rm pc_asm.symc.o pc_asm.symc
>make[1]: Leaving directory `/boot/home/Sourcecode/oskit-20010214/kern'
>make: *** [kern/all.MAKE] Error 2
>
>om
>
>I don't know how to do inline assembly, much less 32 bit inline assembly
>with this screwed up (GNU) compiler... =\

Considering the number of people who do not find the compiler screwed up,
don't blame it immediately.

Now, I don't pretend to know what the problem here is -- I know I compiled
the OSKit itself without a hitch -- I just ran out of space when compiling
example kernels and haven't had a chance to go back after repartitioning my
drive (that's this weekend or next...).

You have an error -- track it down. Find out what it's conflicting with. If
you're running BeOS, that could be the problem right there -- I don't
remember off the top of my head if BeOS is a permissible development base...
and even if it is, RTFM -- who knows what special things need to be done for
BeOS? I don't... do you?

>I tried to compile the compiler several months ago but ran into
>unresolvable problems with the makefile.
>
>> I've modified sample kernels and booted them without any problem.
>
>Do you understand it enough to design a non-trivial kernel to use it? ;)

That's what takes the time. You have to start with the trivial, modify the
samples, try things. Can you boot a sample kernel? Get that working before
you worry about any of the complex stuff.

>> And looking at the list's archive I've suceeded at building the BSD
>> boot adapter which configure fails to configure on its own.
>
>I barely even know what you are talking about... I can't even get ISP
>dialup working under BSD.

WinModem, I'm guessing. If it's a configuration thing, read up on BSD and
Linux -- there are a couple of very good books discussing the Kernel design
for BSD and Linux available, you might want to look at those, too (just as
an aside).

>> So I don't understand what is your problem using OSKit, its
>> documentation and the mailing list.
>
>See above.

What was supposed to be seen?

>> BTW once again in this discussion it must be said that OSKit is a set
>> of tools for developing an operating system, not an OS itself so there
>> is no abstract stuff in its documentation.
>
>Gronk?
>Oskit is bursting at the seams with abstractions!
>A whole boatload of them are listed in the table of contents alone; let
>me point them out to you.

Okay, I cut out your definitione, 'cause they bored me to tears. Not really,
just trying to save space.

OSKit provides CONCRETE IMPLEMENTATIONS of various abstractions. All the
abstraction does is serve as a definition of the theory BEHIND the
components -- but the components ARE there. So saying that because
"Component Object Model" is in the manual means that the OSKit is
abstraction based is a complete oversimplification -- and quite frankly,
verging on being untrue. COM is a known abstraction -- that the OSKit uses.
Does the manual mention abstract ideas? Yes. But the OSKit is a concrete
system, not some theoretical topic.

>Really, you should be very thankful for many of these. Without them you
>would be left with a nothing but a memory dump and machine language. ;)
>
>	Heck without abstractions you would be left speechless trying to
>demonstrate how useful your pointed rock can be when used in conjunction
>with a split stick and leather lashings... You could hardly even begin
>to think about building a house or a city as that would require a
>manipulation of sense impressions that can only be accomplished in the
>abstract. The point that Oskit is "just a toolkit" should be a point of
>*SHAME* not pride!

I'm very glad that the OSKit is a toolkit -- otherwise it wouldn't be very
useful to me!

You seem to be blending the idea of a programming abstraction with a
communications abstraction. And abstractions are useful in both. But the
OSKit provides something behind the abstraction -- both the pointed
rock-tipped stick and the "spear."

>	Let me tell you a little sekrit: If you want OSkit to catch
>on and be
>used as widely as GRUB you must invent a more useful abstraction than
>the "kit", You will find the areas of functionality such as the stuff
>you use to manage executable files, (Chapts 21, 30), Instead of calling
>it "runtime linker and loader", call it "Flexi-load(tm)", and then
>package and sell these tools individualy. (allowing the devel to make
>his own toolkit and focus on the stuff he wants and leave the rest)

Package and sell? Right. Sorry, this struck me as a bit... inappropriate.

The OSKit, from what I recall, is intended to aid in operating systems
RESEARCH and DESIGN -- the people who do that for work (or fun, there are
some of us). We know what a runtime linker and loader is. We'd be the ones
to write something useful... like "Flexi-Loader(r) (tm) (c)".

>	The strange thing is that once you give these things names
>in your mind
>those names become *forms* in the platonic sense, As you begin to
>explore and understand what the names MEAN, and start working to make
>more perfect examples of the form you begin to think of changes and
>enhancements in design and organization that would never have occoured
>to you when they languished as relatively nameless components of the
>vague, amorphous OSKit.

*blink* You took entirely too much Psych in college.

OSKit isn't designed to hold your hand. Work a bit here, folks.


>////////////////////////////////////////////////
>>>>Getting code to the point of having a bootable entry point is a
>monumental task. <<<
>
>Amen.
>
>>>OSKit has already done this for you.<<<
>
>How; show it to me...

He's talking in general, starting from scratch. The moment you compile a
kernel with OSKit, it has a bootable entry point. Unless, of course, you're
changing the boot loader -- in which case, no guarantees. :)

>>>> This, in itself, is worth it's bits in gold. Cheers to the
>OSKit developers!<<<
>
>How do I collect?

PLAY! And switch to Linux, get off your little graphical interface.


>>>>Everybody in the course was at least able to get the examples to boot.
>
>They must have had excelent instruction.

Read a bit. A lot, actually. If you have a problem, ASK A QUESTION ABOUT IT.
Don't just come here and gripe.

>>>> A tip for you. Get the Mandrake 8.0 release and install it and
>the GRUB bootloader (if you are doing stand-alone development).<<<<

Yay Mandrake. :)

>My drive isn't anywhere near that large... I only have about a gigabyte
>for weird OSes, Mandrake requires all 3gb; currently BeOS is installed
>and functioning OK. I've been moving my platters from machine to machine
>for the last five years. That way I don't have to re-build my DOS system
>each time I upgrade.. ;) It used to be teamed with an older 850 but I
>retired that one... (that thing is tough to get spun up these days). I
>have a 60 giger which I am saving for an OS that deserves it. I thought
>BeOS was the one but it crashed when I gave it a workout... Right now
>the 60 gb drive is mounted in a relic of a case (a classic full-Tower)
>that I am building up as a compile farm. If that machine ever goes
>on-line again, it will run BSD...

Okay: Upgrade to Windows 2000. I refused to give my DOS prompt up. Sure, I
can't run everything DOS-based, but the stability and compatibility is worth
is. If you're stuck on a Command Line fetish, get Linux, and don't install
X. It's so much more fun, and powerful, than DOS could ever be. And if you
have some programs, write equivalents on Linux -- you'll learn Linux
programming fast, that way. And you can certainly install Mandrake to a
gig -- do a custom install, skip everything unnecessary -- ie, X is the
first thing that comes to mind.

And a drive won't die if you install something on it that you later
delete -- TRUST ME. You have a 60gig drive -- set it up tonight! Start using
it tomorrow! There's no use in saving a drive for an "OS that deserves it."
Just install, keep important stuff backed up, and reinstall if you make a
mistake.

>Last time I had Mandrake 6.2 running, and recompiled the kernel the
>flaming INITd keept giving me this error message because I had chosen
>not to install VFAT (not desired for a classic FAT16 partition). It
>really irked me that I had no idea how to go about fixing it. I felt
>there was no reason to get higher-level stuff working when there was no
>way to fix even minor crap like this...

So did you figure out what caused that problem? Like the VFAT driver being
expected by the kernel, or mentioned in an init script?

>If I try linux again it'll be either Slackerware or Debian.

Mandrake is an easier install, and will get you up and running faster.

>>>> This will allow you to compile the OSKit in native mode, then
>load the kernels directly as multiboot images. <<<
>
>I never worried about boot loaders after I got lilo to work almost
>right... (the booter for MINIX is vastly supperior). I hardly know a
>thing about this "multiboot" and am not sure I really want to... =\

Multiboot -- able to boot boot images/kernels in multiple formats -- ie, a
Linux kernel, DOS, OSKit, BSD, etc.

And getting Grub working isn't much different than getting LILO working --
what with the X-based and menu-based LILO config makers, LILO is a piece of
cake.

>--
>NUKE NIGERIA!!!!  (419)

I MUST have missed something...

>http://users.erols.com/alangrimes/  <my website.
>Any usage of this e-mail account is subject to the terms and conditions
>specified on my website.

Ah, a disclaimer...

Mr. Grimes -- if you are having trouble, ASK. You'll get help a lot more
easily if you ask questions directed at solving each problem you encounter
than you will be continuing to send generic "OSKit is dreadful" messages to
the mailing list. No one can understand everything immediately -- but you'll
get a lot better help if you ASK before you complain. Assume things do work,
even if they don't make sense initially. Be assured, if something didn't
work, someone else will find it, too.

Good luck on your projects.

Andrew
References: