Subject: Linux-Development Digest #739
From: Digestifier <Linux-Development-Request@senator-bedfellow.MIT.EDU>
To: Linux-Development@senator-bedfellow.MIT.EDU
Reply-To: Linux-Development@senator-bedfellow.MIT.EDU
Date:     Fri, 20 May 94 14:13:10 EDT

Linux-Development Digest #739, Volume #1         Fri, 20 May 94 14:13:10 EDT

Contents:
  Re: Distributions considered harmful (Alan Cox)
  source of strptime? (Dirk-Jan Koopman)
  Bug report (?): Gnu C++ 2.5.8 with Linux 1.0.8 (Posix) or 1.1.0 (Remco v. Engelen)
  Re: Distributions considered harmful (Brandon S. Allbery)
  Re: 32-bit Novell desktop OS combines Unix, DOS 7 (Richard Krehbiel)
  Re: Anybody working on BSD dump porting? (Brandon S. Allbery)
  Re: Terminator-Power Question (Rob Janssen)
  Re: Modem hangup? (Rob Janssen)
  Re: Anybody working on BSD dump porting? (Rob Janssen)
  Re: TCP lockups through Cisco box (Rob Janssen)
  Re: LOGISCAN V0.0.1 released (CHRISTOPHER M MAY)
  Re: Cache-optimizing page allocation (Wayne Schlitt)
  Re: In defence of variety (was Re: Distributions considered harmful) (Daniel Quinlan)
  Re: 32-bit Novell desktop OS combines Unix, DOS 7 (Laurent Chavey)
  Re: net3 tcp window sizes (NOT SEQUENCE NUMBERS!), Please read (Bradley E. Smith)
  Re: X Programming Example Wanted ... (kempsonc@logica.co.uk)
  Binary compatibility (lilo)
  Help with GDB (Doug Lenz)
  mmap locks heavily (Davor Jadrijevic)
  Re: Z-Note Ethernet device driver status (Donald Becker)
  Re: SIGHUP - Deep Kernal Guts question! (Sean Puckett)

----------------------------------------------------------------------------

From: iiitac@uk.ac.swan.pyr (Alan Cox)
Subject: Re: Distributions considered harmful
Date: Tue, 17 May 1994 22:47:29 GMT

In article <1994May17.221401.22928@oxvaxd> u92drb@ecs.ox.ac.uk (Daniel Barlow) writes:
>This one will run and run!
Russ Nelson's last great observation produced Net2Debuggeda nd NET3...

>We could start with a merger between SLS and the kernel.  I don't know
>how much of their module stuff has been incorporated into the standard
>kernel, but given that I had to recompile it the other day to get
>parallel printer support, probably not much. 

I have stuff for adding modules support to network drivers that has
been contributed by Bjorn Eckwall (sp ?) of the DE600 driver fame. Until we
have nice working module drivers I'm not going to start modularising the
protocol layers.

Alan


------------------------------

From: djk@dirku.demon.co.uk (Dirk-Jan Koopman)
Subject: source of strptime?
Date: Fri, 20 May 1994 10:25:06 GMT


Does anyone here know of a source of the source of 'strptime', the
routine provided by sunos for converting ascii dates into a 'struct
tm'.

Dirk

------------------------------

From: rcbare@rwa.urc.tue.nl (Remco v. Engelen)
Crossposted-To: comp.os.linux.misc
Subject: Bug report (?): Gnu C++ 2.5.8 with Linux 1.0.8 (Posix) or 1.1.0
Date: 20 May 1994 12:10:41 +0200


Hello,

I think I've found a bug in either the Linux floating point emulation code
or the Gnu C++ compiler for Linux. When the running the following piece
of code I get a Floating Exception:

// BUG.cc : expose bug in floating point handling when using class
//          parameter

class some_class { public: ~some_class(); };

some_class::~some_class() { } // empty destructor

float func(some_class n) { return 1; }

int main()
{ some_class p;

  float f=func(p);
        f=func(p);
} 

// End of BUG.cc


The problem is clearly the combination of a float return argument and a
parameter of a class with an (empty) destructor. If either is removed
(i.e. the type is changed to int or void, or the parameter is given
another type) the bug disappears. The exception is raised on the second
call to the function, although I think any reference to the variable p
will raise the exception.

I've compiled this with version 2.5.8 of the GNU C++ Compiler, both on
a straight out of the box Slackware system (version 1.0.8 (Posix)) and a
slightly modified Linux system version 1.1.0. Both gave the sazxme
exception at the same place.

Unfortunately I have no expertise in either Linux, C++ or GNU compilers
to give a more accurate or precise description of the bug, but I hope
this will be enough to be able to replicate it. You can always ask for
more info when necessary.

Remco van Engelen,
rcbare@urc.tue.nl
flufje@stack.urc.tue.nl

[Note: this posting may have appeared before, but I haven't seen it
       so I'll just post it this one more time...]

------------------------------

From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: Distributions considered harmful
Date: Wed, 18 May 1994 00:15:05 GMT

In article <2rbcg5$pj1@news.tamu.edu>, drs0587@net.tamu.edu says:
+---------------
| On a Sun there is no operating system distribution competition, and
| as a result, you see almost no movement toward including significant
| features/tools that we have come to expect in any decent linux 
| distribution. Why does Solaris still not have emacs, mosaic, archie,
| ftptool, mtools, perl, tcl/tk, gnu tar, tn3270, gzip, and patch,
+------------->8

Your argument here falls flat with one counter-question:  Why doesn't SCO
provide emacs, Mosaic, archie, ftptool, mtools, perl, tcl/tk, ...?  After all,
even if they don't count Linux as competition, there's Interactive/SunSoft,
Novell, Esix, etc. (although SCO doesn't seem to consider *them* to be
competition, either, at least in print...).  (Actually, recent ODT may have
Mosaic.)

++Brandon
-- 
Brandon S. Allbery         kf8nh@kf8nh.ampr.org          bsa@kf8nh.wariat.org
The FUDs at Microsoft are shouting "Kill The Wabi!"

------------------------------

From: richk@netcom6.netcom.com (Richard Krehbiel)
Subject: Re: 32-bit Novell desktop OS combines Unix, DOS 7
Date: Fri, 20 May 1994 11:07:34 GMT

In article <1994May19.192201.2639@uk.ac.swan.pyr> iiitac@uk.ac.swan.pyr (Alan Cox) writes:

> In article <2rfvq5INNrlp@ope001.iao.ford.com> ekimmina@pms709.pms.ford.com (Eric Kimminau) writes:
> >version of Stacker. Lets take a really wild stab at something and hope
> >that Novell has Linux/Corsair running on a stacked partition. Sounds
> >like something really cool to me!
> 
> This is why I find the whole thing so improbable in some ways. There is no
> way Novell could add Stacker support to the Linux kernel environment without
> in effect releasing a GPL'd stacker equivalent. Given that stacker are
> fairly protective of their code and ideas this is unlikely. Also the GPL
> prevents patented items being GPL'd and the patent used to lock them down

Maybe Novell has decided to try to prove that the GPL is
unenforceable.  Some one would have to actually sue Novell to enforce
the GPL.  Who would do that?  Linus?  FSF?  You?  Novell has a lot
more money for legal defense, and a bit more experience in the
courtroom as well...
-- 
Richard Krehbiel                                richk@netcom.com
Picture a clever one-liner here...

------------------------------

Crossposted-To: comp.os.linux.help
From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: Anybody working on BSD dump porting?
Date: Fri, 20 May 1994 11:24:48 GMT

In article <2rhdpn$r7r@bosnia.pop.psu.edu>, barr@pop.psu.edu (David Barr) says:
+---------------
| And you're not with tar?  (forgetting Emacs's tar-mode)
+------------->8

Henry Spencer's tarx works fine for recovering damaged GNUtar archives...

++Brandon
-- 
Brandon S. Allbery         kf8nh@kf8nh.ampr.org          bsa@kf8nh.wariat.org
The FUDs at Microsoft are shouting "Kill The Wabi!"

------------------------------

From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: Terminator-Power Question
Reply-To: pe1chl@rabo.nl
Date: Tue, 17 May 1994 22:26:56 GMT

In <CpwswH.Gz9@seneca.ix.de> hm@seneca.ix.de (Harald Milz) writes:

>Rob Janssen (rob@pe1chl.ampr.org) wrote:

>: > Interfered in what way?
>: > Did the tape drive short-circuit the terminator power?

>: > (if it is not that, I really cannot imagine how supplying terminator
>: > power can interfere with anything)

>No? Imagine you connect two external devices with TERMPWR enabled. You never
>ever get equal voltages from both, resulting in a high current flowing
>from the higher voltage source to the lower. There is normally no current
>limitation with TERMPWR terminals except for a blow-fuse. I once blew the
>1.5A fuse on my AHA1542B when connecting an external Viper 150S with TERMPWR
>enabled. If this does not happen, it's mere coincidence. This does not 
>interfere with the signal lines, however. In most cases (where blow fuses 
>are supplied), you'll probably never notice it until you must rely on the
>TERMPWR of the blown device.

I thought the supply of terminator power was supposed to be done via
a (shottky) diode, so that no reverse current flows into the device.

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

------------------------------

From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: Modem hangup?
Reply-To: pe1chl@rabo.nl
Date: Tue, 17 May 1994 22:34:43 GMT

In <2ravc0$ve@potrzebie.ee.pdx.edu> bairds@ee.pdx.edu (Scarrow) writes:

>I'm developing some software for a small company (actually it's been in use
>for a while now, but we're having a few problems) involving automated modem
>transactions, and I've had some difficulty getting Linux to drop the DTR
>line.  I've tried setting the line flags, setting the output baud rate to
>zero, etc.  I noticed that kermit seems to drop it sometimes, so I'm going
>to go take a look at that source, but was just gonna pop the question here
>to see if anyone else has had problems with this.

Setting B0 works fine.  Some other UNIX systems require close/open of
the device.  Make sure the HUPCL flag is set.
Linux is not so picky.

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

------------------------------

Crossposted-To: comp.os.linux.help
From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: Anybody working on BSD dump porting?
Reply-To: pe1chl@rabo.nl
Date: Tue, 17 May 1994 22:39:02 GMT

In <2rad8i$bqu@serra.unipi.it> gio@di.unipi.it (Gioacchino La_vecchia) writes:


>I'm trying to port bsd dump to linux.
>Many header files have not corrispondence on linux systems.

>Nobody working on it?

Because of the many filesystems that Linux supports, trying to port
something like "dump" is not simple, and IMHO not a good idea.

You are better off making backups with "tar" or "cpio".

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

------------------------------

Crossposted-To: comp.os.linux.help
From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: TCP lockups through Cisco box
Reply-To: pe1chl@rabo.nl
Date: Tue, 17 May 1994 22:45:42 GMT

In <2raopb$bjh@news.u.washington.edu> erik@hilo.phys.washington.edu (Erik Olson) writes:

>Okay, haven't seen anyone post about this yet, so I figured I'd spew.
>I use my Linux system over the phone through a Cisco terminal server
>that telnets to its hosts.  One major problem I have is that at random
>the session will completely freeze and the only way to start it back again
>is to send a BREAK signal, giving me back the cisco prompt, at which point I
>hit return to resume the session.  Text then blasts by and the session resumes.

>The criteria:

>  * Happens on other tcp servers, not just telnet.  Therefore, I think it's
>something in the kernel-level code, not telnetd.

>  * Only happens on the Linux box.  Therefore, I think it's a Linux-specific
>oddity.

>  * Happens on dialin AND locally-connected lines.  Therefore, it's not
>a modem handshake problem.

>  * Only happens on Linux boxes, and has been happening since pl14.  I am
>using 1.1.12 now.

>  * Oh, and it happens on our Departmental cisco box, which has a SLIGHTLY
>earlier software revision than the campus-wide cisco box, which doesn't
>exhibit the problem.

>Someone say they know what this means, please!

I have seen this as well on a Cisco 2503 at work.
As I could not understand what this breaking back to the cisco prompt and
then entering the session would do at the TCP protocol level, I have assumed
it is a bug in the Cisco.  The fact that a newer release of the software
does not exhibit this seems to make that even more likely.

Rob
-- 
=========================================================================
| Rob Janssen                | AMPRnet:   rob@pe1chl.ampr.org           |
| e-mail: pe1chl@rabo.nl     | AX.25 BBS: PE1CHL@PI8UTR.#UTR.NLD.EU     |
=========================================================================

------------------------------

From: cmay@titan.ucs.umass.edu (CHRISTOPHER M MAY)
Crossposted-To: comp.os.linux.misc
Subject: Re: LOGISCAN V0.0.1 released
Date: 18 May 1994 02:06:14 GMT

Well, the file made its way into the ALPHA/scanner directory on tsx-11.
Look for logiscan.tar.z, I'll let you all know how it works, but 
probably not until after finals are over.
--

-Chris May, Computer Science, University of MA, Amherst
-       Technical Assistant, P.C. Maintenance Lab


------------------------------

From: wayne@backbone.uucp (Wayne Schlitt)
Crossposted-To: comp.sys.ibm.pc.misc
Subject: Re: Cache-optimizing page allocation
Date: Fri, 20 May 1994 14:18:18 GMT
Reply-To: wayne@cse.unl.edu

In article <Cq1zBq.FJ3@cs.ruu.nl> hhanemaa@cs.ruu.nl (Harm Hanemaaijer) writes:
> > [ 2-way set associative out performs direct mapped caches ]
> >
> >Some studies have shown that it's possibel to greatly reduce this
> >factor (on a physically mapped cache) by moving the pages that
> >trashes.
> 
> I have been thinking how nice it would be to do this in Linux...
> It would be very beneficial (even with proper write-back caches).

I agree.

> [ wrote an optimizing malloc that helped ]
> 
> I think should not be too difficult to implement something like this at
> the kernel level with a simple per-process hit count array. Dynamically
> detecting trashing pages sounds rather more difficult though.

Dynamically detecting trashing pages sounds real difficult.  I can't
see any way to do it because the OS can't get any information about
cache collisions.


Linux could do something like your malloc modifications.  If you need
a new page for a process, select the physical page that has the fewest
cache conflicts with other pages in the working set.  This would probably
require an array with counts of the number of pages mapped to each
slot and increment/decrement the counts every time you add/remove a
page.  You could probably assume that the cache is 256K.  Caches which
are smaller will just have the counts spread across several entries,
caches which are larger will have some entries doubled up.  It still
might give you a fairly large performance increase for not much
coding.


-wayne

-- 
Vg'f _nznmvat_ ubj znal crbcyr jvyy eha fghss guebhtu ebg13 vs gur grkg
ybbxf tneoyrq, vfa'g vg?

------------------------------

From: quinlan@pollux.cs.bucknell.edu (Daniel Quinlan)
Subject: Re: In defence of variety (was Re: Distributions considered harmful)
Date: 18 May 1994 02:30:38 GMT
Reply-To: quinlan@spectrum.cs.bucknell.edu


Brandon Allbery wrote:

(about iBCS binary "standard paths")

>> [...] (I brought this up once on FSSTND and was bluntly told that
>> that was unimportant.  I haven't bothered to mention it since; it
>> may not be important to them, but in my environment it's essential.)

Daniel Quinlan responded :

> If you haven't bothered to bring something up a second time, then
> don't blame it on FSSTND.  In my FSSTND archives, I have a *single*
> message from you (buried in over 6MB of other messages), mentioning
> iBCS standard paths -- dated Dec 19.  And now, May is almost over.
> The first public version wasn't even released until February.  You
> are being somewhat unrealistic if a single message about it is going
> to start a discussion.

I should add that I am more than willing to discuss something that
concerns you, Brandon.  You've already given quite a bit to the
discussion, IMO, and it is probably worth some amount of attention.

In other words, pouting over a five month old message isn't going to
get anyone anywhere.

Dan
--
Daniel Quinlan  <quinlan@spectrum.cs.bucknell.edu>

------------------------------

From: chavey@bambam.cis.udel.edu (Laurent Chavey)
Subject: Re: 32-bit Novell desktop OS combines Unix, DOS 7
Date: 20 May 1994 13:56:39 GMT

>|> we all know that Linus himself says that it would be virtually impossible,
>|> or at least very difficult, to port the Linux kernel to another processor.
>
>That is definitely wrong, because the kernel is currently beeing ported
>to 680x0 machines (ATARI,AMIGA) and running there and also there are people
>doing ports to the DEC ALPHA and PowerPC.
>
I would also add, that some of the people working at novell are
not born from the last rain. Some may in fact know very well what they
are doing, and how one can port the kernel.
Now, the fact that they used Linux as a base, show that they were not
interested in the base (usual os stuff), but were more interested in
adding all the functionality that will make Linux a good competitor
against NT. Do not forget novell is on our side, TO NOT LET 
MICROSOFT NAD BILL GATE OWN THE COMPUTER WORLD. Lets all fight to
not have another giant dictate the computer future.


------------------------------

From: brad@bradley.bradley.edu (Bradley E. Smith)
Subject: Re: net3 tcp window sizes (NOT SEQUENCE NUMBERS!), Please read
Date: 20 May 1994 11:19:34 -0500

no soon then I thought it was working, it hung again.....oh well
I really could use some help.
-- 
Bradley Smith                    brad@bradley.edu ---  309-677-2337
Network & Technical Services @ Bradley University, Peoria, IL

"It's amazing how much scrap metal you get from 4 cans of beer"

------------------------------

From: kempsonc@logica.co.uk
Subject: Re: X Programming Example Wanted ...
Date: Fri, 20 May 1994 14:57:15 GMT

From charlie Fri May 20 14:55:25 1994
To: rgeens@wins.uia.ac.be
Subject: Re: X Programming Example Wanted ...
Status: R

> |> I have just started some X development on my new
> |> Linux workstation and I have come across the following problem:
> |>
> |>    Many of the Xt Calls I am used to are no longer available
> |>    eg. XtVaAppInitialize ...  Where are these?  I note that
> |>    XtAppInitialize is available though.
> |>

> What do you mean with does not exist ?
> They can both be used without a problem...
> I agree that it is hard (or even impossible to find a manpage, 
> but that doesn't mean they don'r exist anymore....

OK, point taken.  I have just figured out that I was not linking
with all the necessary libraries - hence the call was not being
resolved.  Odd that the manpages aren't included in the X distribution 
though.  Any idea where I can get hold of them?

Charlie

==========================================================================
Charlie Kempson                       INTERNET: charlie@mpef4.logica.co.uk
                                                kempsonc@logica.co.uk
Logica Space & Communnications Ltd.,
68 Newman St.,
LONDON W1A 4SE                        TEL: 071-637-9111 ext. 4192

==========================================================================


------------------------------

From: lilo@slip-1-52 (lilo)
Subject: Binary compatibility
Date: 18 May 1994 03:22:16 GMT

On Mon, 16 May 1994 20:51:40 GMT, Warner Losh (imp@boulder.parcplace.com)
wrote in comp.os.linux.admin, under the subject, "Re: Software not available
for Linux":

> I'll take this out of context....  I think that as good as Linux is,
> it has fallen down on the binary compatibility issue with previous
> releases and that failure may cause people to think twice before
> supporting linux with their product.

Warner,

I saw your original posting on the activists list, and I'd like to talk
about this.  It's an important issue.

One of the largest vendors of systems software for mainframes, though
certainly not the most user-friendly, is IBM.  As someone who worked as a
systems person at installations of IBM systems for over a decade, I think my
experience will be useful.

As new versions of CICS (one of IBM's transaction processing monitors) came
out, it frequently happened that application vendors' products were broken
by the changes.  Vendors cheerfully provided new releases, as a rule.  It
was simply a cost of doing business.

Um, and let's not forget the many versions of DOS, and all the systems
software which is broken when major changes occur in DOS, or a new version
of Windows comes out.  Or programs which are broken by new features such as
disk compression.

Well, I think I've argued successfully that it is *possible* for us to
change our binary formats on a relatively frequent basis.  But that doesn't
address why it might actually be *advantageous* to do so, at least for the
time being.  Let me outline my reasoning.

One of the real advantages Linux has over other operating systems,
particularly other Unix systems, is that our architecture is not wedded so
strongly to the past!  Granted, in certain areas we have to be--source
compatibility (as much as possible) with BSD and System V, general system
call compatibility with other Unix systems and so on.  But the internals of
Linux are not as coated with the muck of trying to maintain years of
downward compatibility with obsolete techniques as, say, Windows or DOS have
become.

Linux is not mature technology.  It probably won't be for some time.  And
that's one of its real advantages.  When something goes wrong we can fix it,
without being burdened by a lot of historical baggage.  And usually
maintaining a fair amount of source compatibility.  I just don't think that
binary compatibility problems every six months or so are that high a price,
considering what we gain in flexibility.

Anyway, there you have one person's opinion.  :)  I'd like to hear other
opinions, and I think we can only profit from more discussion!


lilo

------------------------------

Crossposted-To: comp.os.linux.help
From: lenz@ssd.comm.mot.com (Doug Lenz)
Subject: Help with GDB
Date: Wed, 18 May 1994 02:04:35 GMT

Linux Guru's:

I just migrated to Linux (Slackware 1.2.0 Distribution) and
I have one small problem...

Simple example:

$ gcc -g hello.c -o hello
$ gdb hello
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.12 (i486-unknown-linux), Copyright 1994 Free Software Foundation, Inc...
(gdb) list
1       #include <stdio.h>
2
3       main()
4       {
5               printf("hello\n");
6       }

** works fine **

$ make
gcc -Wall  -g -c workbone.c
gcc -Wall  -g -c hardware.c
gcc -g -o workbone workbone.o hardware.o -s
groff -man workbone.1 | compress >workbone.1.Z
[doug]> gdb workbone
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.12 (i486-unknown-linux), Copyright 1994 Free Software Foundation, Inc...
(no debugging symbols found)...
(gdb) list
No symbol table is loaded.  Use the "file" command.
(gdb) 

** this don't **

I hope somebody else has seen this before...

Thanks for any help you can give me (and I promise not to bother again).

Doug
-- 
=============================================================
| Douglas Lenz   |   Friends don't let friends use MS-DOS   |
=============================================================
|           internet :   lenz@ssd.comm.mot.com              |
=============================================================

-- 
=============================================================
| Douglas Lenz   |   Friends don't let friends use MS-DOS   |
=============================================================
|           internet :   lenz@ssd.comm.mot.com              |
=============================================================

------------------------------

From: davj@ds5000.irb.hr (Davor Jadrijevic)
Subject: mmap locks heavily
Date: 20 May 1994 13:36:00 GMT


HI ALL KERNEL DEVELOPERS - BUG REPORT

If I mmap a file as user, although opening and mmaping read-only, 
mmap locks that file heavily
so even root cannot write such mmaped file. Linux 1.1.11

Best regards, Davor.
--

------------------------------

From: becker@cesdis (Donald Becker)
Subject: Re: Z-Note Ethernet device driver status
Date: 17 May 1994 23:53:57 -0400

In article <1994May17.202049.7873@midway.uchicago.edu>,
 <pegettin@midway.uchicago.edu> wrote:
>  I am considering putting Linux on my Z-Note 425Lnc and see no problems
>(yet) except that I want to be able to use the built-in ethernet adaptor.
>The latest Ethernet How-To said that a driver for the adaptor (which is
>based on the Intel 82593 chip) should be done by mid-1994.  I was 
>wondering if anyone knows what the status of the driver is?  Thanks.

My Z-Note/Thinkpad ethernet device driver (a i82593 driver),  is
called 'znet.c' and is available from
        cesdis.gsfc.nasa.gov:/pub/linux/drivers/

The hooks to include it are already in the Linux 1.0 kernel source.
It would have been included with 1.0, but alpha testing wasn't done in
time for the 1.0 code freeze.

   Please note my new affiliation -- I'm no longer 'becker@super.org'.
-- 
Donald Becker                            becker@cesdis1.gsfc.nasa.gov
USRA Center of Excellence in Space Data and Information Sciences.
Code 930.5,  Goddard Space Flight Center,  Greenbelt, MD.  20771

------------------------------

From: nate@loreli.ftl.fl.us (Sean Puckett)
Subject: Re: SIGHUP - Deep Kernal Guts question!
Date: Fri, 20 May 1994 12:31:48 GMT

In the scrolls of comp.os.linux.development, David Monro (DM) enscribed,
DM> Does this explain why if I kill an xterm/rxvt running less, the less
DM> process goes beserk and eats all my CPU? If so, this is bad, and could
DM> cause real havoc on a multiuser machine (user logs out, process goes
DM> beserk, nobody can kill it till root comes along...)

Yes.  I get things like this a lot.  Very annoying.

But this gentleman has a different view:

From: iiitac@uk.ac.swan.pyr (Alan Cox) at Swansea University College
AC> NOPE: It's in the specs for good reasons... kernels should not set policy.
AC> If you want people to get the boot then the shells you choose to issue them
AC> with should follow the policy. In addition shells can do cleverer things
AC> such as killing other process groups they have created. Fix your shell.

If it IS a shell thing, then all of the shells I've used under Linux
are defective.  And I used sh, bash, and ksh all of which came with the SLS
distribution.  In David's case, less is also broken, because it ignores the
EIO value on its tty input, which is returned when there is no longer a
terminal attached and you attempt to read a key.

If the kernel is going to support Posix, then all of its subprograms should,
too, by all accurately checking for EIO, and forking instead of spawning so
the process ID can be logged, and so that all children can be killed in the
event of a SIGHUP that is not passed on.

But this never happens.  It is impractical.  The official shells supplied
with Slackware don't behave properly, according to Posix.  I can set up,
using bash, a whole herd of applications that came with SLS, then hang up.
WHAM, my CPU is completely firewalled processing infinite requests for input
on a terminal that simply isn't there anymore.

Are all these applications defective?  Is the shell that spawned them
defective?  Should every application be required to know about terminals,
hanging up, and other nitty gritty details?  By your interpretation of
Posix, yes they should.  Maybe you're right. But does it matter if you're
right, if the question is wrong?

Linux is a practical system in many ways, but with this signal problem we have
revealed a point where something does not make sense, the point where the
shield of "Posix Compliance" has prevented real people in real situations from
getting satisfactory results.

 - * -

Based on the answers I've seen to this issue so far, It seems that there
in reality is NO good reason why the kernel can't send a SIGHUP to processes
that will lose their tty input on a hangup.  I'm going to patch the kernel
and see what breaks.

I'll report on what happens, even if it blows my position out of the water.
I'm far more interested in solutions than in being right.  If it doesn't
work, or if something really important breaks, I'll have to come up with
an uglier solution -- which I will also share.


--
                       ..:: nate@loreli.ftl.fl.us ::..
             ..:: Sean Puckett - Albino Frog Software, Inc. ::..
          ..:: The Right Reverend Aural Hardly, MSK, BoC, FCoC ::..

------------------------------


** FOR YOUR REFERENCE **

The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:

    Internet: Linux-Development-Request@NEWS-DIGESTS.MIT.EDU

You can send mail to the entire list (and comp.os.linux.development) via:

    Internet: Linux-Development@NEWS-DIGESTS.MIT.EDU

Linux may be obtained via one of these FTP sites:
    nic.funet.fi				pub/OS/Linux
    tsx-11.mit.edu				pub/linux
    sunsite.unc.edu				pub/Linux

End of Linux-Development Digest
******************************
