Can the Yggdrasil CD REALLY be this bad?!?

Can the Yggdrasil CD REALLY be this bad?!?

Post by Jordan K. Hubba » Mon, 21 Jun 1993 12:09:17




some of my problems and asking for advice, but apparently they fall on
deaf ears.  Time to go public I guess.

First off, with my 486/33 EISA system using an Adaptec 1742A
controller in enhanced mode and using the Yggdrasil Beta Test Release
CD, here were my rather unpleasant experiences:

1.  The kernel supplied falls over very frequently with SCSI errors,
    where the controller times out and hangs completely during the install.
    Perhaps it works on the more common IDE drives, in which case this is not
    a problem for the majority, but it was certainly a cause for concern
    to me.  4 install tries were necessary to get it to run all the way
    through.  This was, in any case, by far the least of my problems.

2.  Once I got it reasonably well installed,  I tried to regenerate a
    kernel.  This doesn't work, due to the fact that the build environment
    is broken and ALWAYS tries to select all possible configuration options
    as 'on'.  I did a 'make config' until I was blue in the face and no
    combination seemed to work.  If you turn off the options you don't
    want (say, a disk controller board you don't have), then the link will
    fall over wanting symbols from that controller's driver (or NFS, or the
    minix filesystem, or almost whatever it was you said 'no' to).  If you
    decide to say 'the heck with it then, give me all the options!' the
    generated kernel will be "too big" and fail to load!  I spent a couple of
    hours trying to code around it before giving up in disgust and going on
    to option #3, needless to say, the "documentation" was no help at all.

3.  Trying to patch the kernel up to the latest revision in hopes that the
    build system might be fixed, I followed the instructions to the letter
    but found that the kernel patches didn't apply again.  This time I
    examined the patch file more closely and discovered that it was in a
    completely non-standard, and therefore unusable, format (take a look at
    it).  The supplied instructions for using it were obviously never
    attempted by the author!

    Because I know a fair amount about the internals of patch, I was able
    to construct Index lines that enabled a good bit of it to apply.
    Needless to say, the new patched kernel tree (even with careful hand
    application of failing patches) doesn't even come close to compiling.
    I think some necessary patches to the include files are simply missing.

4.  Some of the supplied utilities, like emacs, don't work at all - failing
    with weird errors when you ask, for example, for info.  I'd think that
    broken things would simply be left off the distribution rather than left
    around to confuse and annoy.  More disturbing is the implication that they
    were never even run tested, since the errors are very obvious and easy to
    reproduce.

Basically, I'm greatly annoyed and disturbed by the quality of this
product and I have to wonder if this is the best anyone can do.

Surely, among all the people doing Linux, there must be SOMEBODY
willing to do a more professional job?  We are talking money here, and
a product people are paying for - I sure would be happy to see market
forces taking their proper course and offering a better mousetrap!

                                        Jordan

--

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Gilbert Nar » Mon, 21 Jun 1993 16:37:01



[stuff excerpted]
Quote:> 1.  The kernel supplied falls over very frequently with SCSI errors,
>     where the controller times out and hangs completely during the install.

        SCSI timeouts for the adaptec + CDROM are still being investigated
with 0.99pl10. In the meantime, reduce the bus time clock through your
bios setup (if your bios allows).

Quote:> 2.  Once I got it reasonably well installed,  I tried to regenerate a
>     kernel.  This doesn't work, due to the fact that the build environment
>     is broken and ALWAYS tries to select all possible configuration options
>     as 'on'.  I did a 'make config' until I was blue in the face and no

        Try removing /usr/src/linux/.config by hand. Then run make config.

Quote:> Basically, I'm greatly annoyed and disturbed by the quality of this
> product and I have to wonder if this is the best anyone can do.

        Welcome to the frontier :-)
--

Migrant Computing Services | (415)664-1032 (voice)
1032 Irving Street, #435   |-----------------
San Francisco, 94122       | Save the Universe: Stop Entropy Now!

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Ed Ca » Mon, 21 Jun 1993 18:26:35


:       SCSI timeouts for the adaptec + CDROM are still being investigated
: with 0.99pl10. In the meantime, reduce the bus time clock through your
: bios setup (if your bios allows).

Can you give me a little more detail about this?  I'm getting occasional
SCSI HD timeouts with an Adaptec 1542B + Maxtor 1.2GB Panther drive.

BTW, in order to get my SCSI HD to work reliably, I had to replace my
70ns SIMMS with 80ns SIMMS (they were mixed 80's and 70's before).  Don't
know why - maybe my 486/25 isn't set up right.  It was doing wierd stuff like
trashing the FS on the HD, etc. - but Linux wasn't complaining about memory
parity problems!  Is this turned off or something?  MSDOS would complain
(halt the machine, etc.), and that's how I finally clued into what was going
on.
--

"It is your resistance to 'what is' that causes your suffering."   -- Buddha

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Bill Rieme » Mon, 21 Jun 1993 21:34:12


Nntp-Posting-Host: whisker.lotus.ie

        Jordan> 3.  Trying to patch the kernel up to the latest revision in
        Jordan>     hopes that the build system might be fixed, I followed
        Jordan>     the instructions to the letter but found that the kernel
        Jordan>     patches didn't apply again.  This time I examined the
        Jordan>     patch file more closely and discovered that it was in a
        Jordan>     completely non-standard, and therefore unusable, format
        Jordan>     (take a look at it).  The supplied instructions for using
        Jordan>     it were obviously never attempted by the author!

To be perfectly fair, this was really a problem with SLS1.0.  Towards the
end of my mail order days with SLS1.0, I distrubuted a patch that would
convert SLS's non-standard kernel into a standard kernel.  After that
Peter came out with SLS1.01 which seemes to have a completely standard
kernel, with extra patches for the user to apply.  Still, you have to use
a special undocumented make option, I think it was "mrproper"  to clean
things first; just clean leaves too many things around.  However, this is
true of any kernel that has already been compiled before you try patching.

Still, Yggdrasil has been around long enough that they should either be
using a later version of SLS, provide a patch to standardize the kernel,
or be making there own distribution without this problem.

        Jordan> 4.  Some of the supplied utilities, like emacs, don't work
        Jordan>     at all - failing with weird errors when you ask, for
        Jordan>     example, for info.  I'd think that broken things would
        Jordan>     simply be left off the distribution rather than left around
        Jordan>     to confuse and annoy.  More disturbing is the implication
        Jordan>     that they were never even run tested, since the errors are
        Jordan>     very obvious and easy to reproduce.

Sounds like either paths.el is missing, or you are starting emacs with the
-q option, but not loading this file in you .emacs file.  Something that
has long confused me is why SLS does not include x11emacs.  With a short
script you have the non-x11 version, and the size difference between
x11emacs and just emacs is really neglishable.

 emacs:

   #! /bin/sh
   emacs -nw $*

Anyways, that aside, if I where to pay good money for the Yggdrasil cd,
I would expect emacs-19, x11emacs-19, and lemacs all to run properly
directly from the CD.  (With cursor keys and backspace defaults in all
versions.)

                                    Bill

--
"Yeti!  Saw them in the London Underground twenty years ago.  Ghosts!
A headless woman used to walk through my bedroom at midnight.  Mermaids?
Grandpa was rescued from the Marie Celeste by one.  Vampires?  I always
wondered where my dad went to at night.  Telepathy?  Right now you're
thinking that I'm talking crap.  So what can you tell me that I won't
believe in?" - Andrew Hunt, "CAT'S CRADLE: WITCH MARK"

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Yggdrasil Computi » Mon, 21 Jun 1993 21:47:24


        I am going to address Jordan's specific complaints in the rest
of this posting, but first I want to provide a some background.

        Shortly after the LGX alpha release started shipping, Jordan
Hubbard announced his plans to do his own 386bsd/Linux CDROM, which I
assume is still in the works.  I am not accusing Jordan of
deliberately lying in order to clear a path for his product, but I do
think readers should read his complaints with an awareness of this
potential conflict of interest.  Also, for the sake of full
disclosure, I should mention that in my college days I worked for
Jordan for about three months (he was the full time sysadmin for whom
I and three other students worked).  I have had friendly contact with
Jordan since then, and I believe that Jordan is basically a good guy.

        Also in the way of background, it is important to realize that
the current release of LGX is a beta release.  The whole point of a
beta release is that adventurous users are supposed to find, report and,
occasionally even fix bugs.  If you're not so adventurous, wait for
the production release.  There are already hundreds of advance orders
from resellers for the production release.  And, yes, I did use the
environment that became the beta release quite extensively--almost
exclusively--for months before I cut the CD.

        Now I will address Jordan's statements.


| some of my problems and asking for advice, but apparently they fall on
| deaf ears.  Time to go public I guess.

        On 16 June 1993, I got a letter from Jordan that began, "I
actually wrote this quasi-review for someone else, but it occurred to
me in hindsight that you might also be interested to know what I
thought of it all."  The only complaint which was common between the
letter and the posting was Jordan's trouble building the kernel.

        It looked to me like Jordan had basically reviewed the beta
release and set it aside, so it looked like a responding to the letter
was optional or at least not urgent.  So, I set Jordan's letter aside
to work on more pressing matters.  With the 2000th beta copy shipping on
Monday, there are a lot of more pressing matters, such as seeing to it
that things are fixed for the production release, and answering the
several phone calls, faxes or email messages that come in each business
hour.

| 1.  The kernel supplied falls over very frequently with SCSI errors,
|    where the controller times out and hangs completely during the
|    install.

        Since Jordan didn't identify what kind of SCSI controller, CDROM
drive or hard disk drive he is using, I cannot respond specifically
to this problem, which was not mentioned in Jordan's original letter.
However, there are now open bugs against every single Linux SCSI
driver, although many of these reported bugs are difficult to
reproduce and may in some cases be due to IRQ conflicts or other
hardware configuration problems.  Once Yggdrasil grows a little
more, I intend to hire someone to work on these SCSI problems, which
are presumably common to all Linux distributions, by the way.

| 2.  Once I got it reasonably well installed,  I tried to regenerate a
|     kernel.  This doesn't work [...]

        If you install the beta release, you definitely can build a
new kernel without errors by typing:

                        cd /usr/src/linux
                        make

        I just reinstalled the beta release on a hard disk here and
did a kernel build with those two commands just to see this again.

| 3.  [The supplied kernel patches for upgrading the sources from
|     0.99.4 to 0.99.7A don't work.]

        This is true.  When people have called about this, I've
offered to mail them a floppy with the 0.99.7A+ kernel source tree and
helped them get newer kernel sources by FTP.

| 4.  Some of the supplied utilities, like emacs, don't work at all - failing
|     with weird errors when you ask, for example, for info.  I'd think that
|     broken things would simply be left off the distribution rather than left
|     around to confuse and annoy.

        Emacs runs fine both in character mode and as a window under
X.  What is true is that somehow some of the byte-compiled lisp files
on the CD were corrupted.  If you use emacs simply as a text editor,
you generally won't notice this, but you will if you live in emacs.  I
did not use the non-editor functions of emacs extensively enough to
notice this problem, but the one person who initially reported it was
able to fix the problem by doing something like "M-X
byte-recompile-directory" in the emacs lisp directory.  (Note: I think
he also had to load the common lisp stuff before recompiling that
particular file.)  The production release won't have these problems
since the emacs tree has long since been replaced by emacs-19 and
Lucid emacs.

--
Adam J. Richter                             Yggdrasil Computing, Incorporated
409 Evelyn Ave., Apt. 312, Albany CA 94706  PO Box 8418, Berkeley CA 94707-8418
(510) 528-3209                              (510) 526-7531, fax: (510) 528-8508


 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Peter MacDona » Mon, 21 Jun 1993 23:44:35



>Nntp-Posting-Host: whisker.lotus.ie


>    Jordan> 3.  Trying to patch the kernel up to the latest revision in
>    Jordan>     hopes that the build system might be fixed, I followed
>    Jordan>     the instructions to the letter but found that the kernel
>    Jordan>     patches didn't apply again.  This time I examined the
>    Jordan>     patch file more closely and discovered that it was in a
>    Jordan>     completely non-standard, and therefore unusable, format
>    Jordan>     (take a look at it).  The supplied instructions for using
>    Jordan>     it were obviously never attempted by the author!

>To be perfectly fair, this was really a problem with SLS1.0.  Towards the
>end of my mail order days with SLS1.0, I distrubuted a patch that would
>convert SLS's non-standard kernel into a standard kernel.  After that

I just knew that this one would come back to bite me :-)
For the record, only one release of SLS ever included a prepatched
kernel.  All others have included the untouched distribution
(literally, since /linux is a link to /usr/src/linux).
The reason for that one dist   was that there were so many patches,
and I had to twiddle some conflicts to get them all to patch.

Anyways Bill, I don't think that this is the problem here.  It sounds
to me like the author is saying the format of the patch files are
incorrect.  Maybe this is because Linus use unified diffs.  More
info would be needed.

...

Quote:>-q option, but not loading this file in you .emacs file.  Something that
>has long confused me is why SLS does not include x11emacs.  With a short

SLS has had x11emacs since version 1.02.  Ditto for the SLS CDROM.  
In fact, it only stopped having it for a short time after the X11
upgrade to DLL jumps.

Peter

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Gilbert Nar » Tue, 22 Jun 1993 02:27:58




> :  SCSI timeouts for the adaptec + CDROM are still being investigated
> : with 0.99pl10. In the meantime, reduce the bus time clock through your
> : bios setup (if your bios allows).

> Can you give me a little more detail about this?  I'm getting occasional
> SCSI HD timeouts with an Adaptec 1542B + Maxtor 1.2GB Panther drive.

        This is partly the setup I have (with a Texel CD-ROM on the SCSI).
No, I can't give you more details mainly because I'm still gathering them.
I've put printk's, status checks, and collision detection variables in
the scsi drivers.  Unfortunately, it is a painstakingly slow process since
I have to be judicious which detection mechansim is activated, else I run
smack in the middle of Heisenberg's Principal.  Stay tuned....
--

Migrant Computing Services | (415)664-1032 (voice)
1032 Irving Street, #435   |-----------------
San Francisco, 94122       | Save the Universe: Stop Entropy Now!
 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Ruediger Hels » Tue, 22 Jun 1993 03:45:03



|> [stuff excerpted]
|> > 1.  The kernel supplied falls over very frequently with SCSI errors,
|> >     where the controller times out and hangs completely during the install.
|>   SCSI timeouts for the adaptec + CDROM are still being investigated
|> with 0.99pl10. In the meantime, reduce the bus time clock through your
|> bios setup (if your bios allows).

Recently we did some investigations into this and discovered that many SCSI
devices simply don't release the SCSI bus while they are busy. They think
they own the bus and when they seek (up to one second for CDROM drives)
they keep the SCSI Bus Busy line active. We tested this with the
AHA1542C and hooked a voltmeter to the line (#36). The Philips CD
writer CDD 521 holds the bus active during the verify command (up to
half an hour)! No wonder the other devices are getting SCSI timeouts.

Some time ago on alt.cdrom NEC drives were also reported not to release
the bus. It seems pretty common. Nowadays they try to sell you the worst
garbage.


 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Ed Ca » Tue, 22 Jun 1993 04:53:30




: >
: > :        SCSI timeouts for the adaptec + CDROM are still being investigated
: > : with 0.99pl10. In the meantime, reduce the bus time clock through your
: > : bios setup (if your bios allows).

: > Can you give me a little more detail about this?  I'm getting occasional
: > SCSI HD timeouts with an Adaptec 1542B + Maxtor 1.2GB Panther drive.

:       This is partly the setup I have (with a Texel CD-ROM on the SCSI).
: No, I can't give you more details mainly because I'm still gathering them.
: I've put printk's, status checks, and collision detection variables in
: the scsi drivers.  Unfortunately, it is a painstakingly slow process since
: I have to be judicious which detection mechansim is activated, else I run
: smack in the middle of Heisenberg's Principal.  Stay tuned....

Hmmm...I've noticed that if I'm running fsck on a large partition (200+ MB)
that I will get several SCSI timeout messages, then the machine will hang -
no disk activity light, etc.  Note that when I boot from floppy the machine
doesn't hang.  Also note that the swap partition is on the SCSI drive.

For the brain dead among us <grin> it seems that there is some sort of
interaction between disk activity between partitions, and that when I
get a SCSI timeout on the fsck'd partition, and the kernel wants to
swap, it does * things.  Hmmm...I've compiled the kernel with debug
on in scsi.c, but the stuff scrolls by too fast for me to see, and I
can't have the machine down for very long, since it's handling a full
news feed, and my feed site doesn't like for the news to back up too
far.
--

"It is your resistance to 'what is' that causes your suffering."   -- Buddha

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Gilbert Nar » Tue, 22 Jun 1993 05:34:14


[rm stuff]

Quote:> Recently we did some investigations into this and discovered that many SCSI
> devices simply don't release the SCSI bus while they are busy. They think
> they own the bus and when they seek (up to one second for CDROM drives)
> they keep the SCSI Bus Busy line active. We tested this with the
> AHA1542C and hooked a voltmeter to the line (#36). The Philips CD
> writer CDD 521 holds the bus active during the verify command (up to
> half an hour)! No wonder the other devices are getting SCSI timeouts.

> Some time ago on alt.cdrom NEC drives were also reported not to release
> the bus. It seems pretty common. Nowadays they try to sell you the worst
> garbage.

        Yipes! I was afraid it would be a hardware problem.  Now things
are starting to make sense. For example, when a processes is copying a
large file from a CD to HD, the only way to prevent the timeouts (in my
setup) was to wrap the read and write calls with both sleep() and sync(),
like this

        for(;;) {
                getsome = read(cdfd, bigbuffer, sizeof(bigbuffer));
                if ( !getsome ) break;
                if ( getsome == -1 ) ....deal with err....
                sleep(1); write(hdfd, bigbuffer, getsome);
                sync(); sleep(1);
        }

        I thought the drivers could do something similar to cope with
this hardward problem (by using additional forms of flow control), but
now I'm not so sure that this will work for the general case (that
'verify' command taking up ~ 1/2 hour is a killer).
--

Migrant Computing Services | (415)664-1032 (voice)
1032 Irving Street, #435   |-----------------
San Francisco, 94122       | Save the Universe: Stop Entropy Now!

 
 
 

Can the Yggdrasil CD REALLY be this bad?!?

Post by Warner Lo » Tue, 22 Jun 1993 06:55:49



Quote:Computing) writes:
>I am not accusing Jordan of
>deliberately lying in order to clear a path for his product, but I do
>think readers should read his complaints with an awareness of this
>potential conflict of interest.

This is getting close to a personal attack.  Yuck.  We don't need this
stuff in col.

I seriously doubt that he's trying to trash anybody's reputations
here.  In fact, he sent me mail about the same problems that he
posted, but since I didn't have the hardware he had, I was clueless as
to what the problem was.

I understand that you and he are potential competitors, but please
keep petty bickering like this out of col.  The last thing we need is
a war of word over whose SLS distribution is the best.  Peter deserves
better than that for his wonderful integration work.

At Xhibition, I heard nothing but good things about Yggdrasil while I
was running the OB/OI demos on the Linux machine that we had in our
booth.

Warner
--

I've almost finished my brute force solution to subtlety.

 
 
 

1. It's not bad canned meat...

Thanks for your time.  We at I.S.P.O.T. (The International Society
for the Preservation Of Trees (URL http://www.local-touch.com/ISPOT/)) feel
that electronic messages are great for getting the word out while preserving
our friend the tree.  Anyway, please stop by our home page!

Thanks again!

SusieQ

2. Add extended attributes to ext2/3

3. Dual 1600x1200 LCD digital displays

4. Help on installing Yggdrasil summer 94 CD from SB16 CD-ROM

5. Makefile for .so

6. Won't detect and mount CD-ROM (Yggdrasil, cd-dependent)

7. find hangs or dumps core

8. Can't install Yggdrasil summer 94 CD-ROM with SCSI-1 CD-ROM

9. I am the most useless dicklicker that you can find on usenet , really.

10. I must be really bored if I am responding to this spam

11. How secure am I really??

12. Am I really being hosted by a 386