Subject: Linux-Development Digest #743
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:     Sat, 21 May 94 10:13:06 EDT

Linux-Development Digest #743, Volume #1         Sat, 21 May 94 10:13:06 EDT

Contents:
  Re: Anybody working on BSD dump porting? (David Barr)
  Re: 32-bit Novell desktop OS combines Unix, DOS 7 (Brandon S. Allbery)
  Re: 32-bit Novell desktop OS combines Unix, DOS 7 (Brandon S. Allbery)
  Re: 32-bit Novell desktop OS combines Unix, DOS 7 (Brandon S. Allbery)
  The type of "speed_t" in <linux/types.h> (Hamish Macdonald)
  Re: p-thread (pre-emptive) package: volunteers gladly accepted :) (Herve Soulard)
  Re: Linux Intrntl & Auto Cfg project (Patrick D'Cruze)
  Re: COMMODORE CALLS IT QUITS (Greg Boehlein)
  Re: SIGHUP - Deep Kernal Guts question! (Rob Janssen)
  Using IDE disks where BIOS claims more than 16 heads (Chris Metcalf)
  Re: SIGHUP - Deep Kernal Guts question! (Chet Ramey)
  Re: Cache-optimizing page allocation  (Maarten Boekhold (Who'd you expect??))
  regex, regexp, posix, and linux (Bill Mitchell)
  Re: ncurses weirdness (Greg Robertson)
  Mosaic2.4+Term hangs (Barry Yip kam-wa)
  O_SYNC support (James Orr)

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

From: barr@pop.psu.edu (David Barr)
Crossposted-To: comp.os.linux.help
Subject: Re: Anybody working on BSD dump porting?
Date: 18 May 1994 22:00:06 -0400

In article <1994May18.210039.6822@belvedere.sbay.org>,
David E. Fox <root@belvedere.sbay.org> wrote:
>But what advantages does 'dump' have that 'tar' or 'cpio' do not?

1. Does multiple dump levels
2. Handles sparse files (and handles them correctly)
3. Handles hard links (and handles them completely)
4. Does not affect atimes of backed up files
5. Supports multiple tape volumes (ok, so does GNU tar)
6. No pathname length/filename restrictions
7. Backs up devices/named pipes, etc.

--Dave

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

From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: 32-bit Novell desktop OS combines Unix, DOS 7
Date: Thu, 19 May 1994 22:50:58 GMT

In article <Cq2CvG.GHn@hunan.rastek.com>, jmorris@darkstar.rastek.com (Jim Morris) says:
+---------------
| then the quoted statement *could* be possible.  For the PowerPC and RISC
| markets, they would simply sell a software package that installed and ran
| using the processors native flavor of Unix.  But that would imply that
| Corsair is an application/utility suite, and NOT an operating system.
+------------->8

Given the limitations on what Novell can do to Linux without violating the
GPL or giving away code they don't want to make public (or can't, in the case
of something like Looking Glass --- or Wabi, if that's the MS-Windows
emulation they're using), that makes sense anyway.

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

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

From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: 32-bit Novell desktop OS combines Unix, DOS 7
Date: Thu, 19 May 1994 22:52:09 GMT

In article <1994May19.192201.2639@uk.ac.swan.pyr>, iiitac@uk.ac.swan.pyr (Alan Cox) says:
+---------------
| 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
+------------->8

I suggest you go back and study the latest userfs announcement on c.o.l.a.

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

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

From: bsa@kf8nh.wariat.org (Brandon S. Allbery)
Subject: Re: 32-bit Novell desktop OS combines Unix, DOS 7
Date: Thu, 19 May 1994 22:53:58 GMT

In article <tgmCq2HEn.8F6@netcom.com>, tgm@netcom.com (Thomas G. McWilliams) says:
+---------------
| It is interesting to note that Novell now owns WorkPerfect.
| Would Novell make their WordPerfect available for their Linux?
| Novell's WordPerfect for Linux?
+------------->8

Why bother?  The SCO version works great under the iBCS-2 emulator --- and I
found out recently that you can often get the SCO version of WordPerfect
*cheaper* than the DOS version (the vendors expect to make money on number-of-
users upgrades, not on the base package).

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

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

From: Hamish.Macdonald@bnr.ca (Hamish Macdonald)
Subject: The type of "speed_t" in <linux/types.h>
Date: 19 May 1994 02:02:55 GMT

I'm having problems with some programs dumping core in m68k-linux
(indirectly) as a result of speed_t being defined as "unsigned int" in
<linux/types.h>

Many, many, many applications hard code a "short ospeed" variable (or
an "extern short ospeed").

The Linux libc termcap/termio libraries think that "ospeed" is of type
"speed_t/unsigned int".

On a little-endian machine, things seem to work out fine, since the
applications write a short to &ospeed, and whoever the reader in the
termcap/termio libraries is reads an "unsigned int" from &ospeed, and
the least significant word of the unsigned int is at &ospeed, with the
next word a zero:

   | speed | 000000|
   &ospeed

On a little endian machine, "short speed" concatenated with "short 0"
is the same as "int speed".

On a big-endian machine, the applications which think that they are
dealing with a "short" write a short to &ospeed, but the libraries
read an "unsigned int" from &ospeed, and the *most* significant word of
the unsigned int is at &ospeed.

   | speed | 000000 |
   &ospeed

On a little endian machine, "short speed" concatenated with "short 0"
is the same as "int speed*65536".

This usually results in something seg faulting as something in libc
tries to index a table with the (incorrectly huge) value read from
ospeed.

Since so many applications out there have ospeed hard coded as
"short", the easiest thing to do in my position is to change speed_t
to unsigned short in <linux/types.h>

I wonder how the decision to make speed_t "unsigned int" was arrived
at.  Can anyone shed some light on this?

Thanks,
  Hamish.

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

From: soulard@druuna.inria.fr (Herve Soulard)
Subject: Re: p-thread (pre-emptive) package: volunteers gladly accepted :)
Date: 19 May 1994 08:15:07 GMT
Reply-To: soulard@sor.inria.fr


> Help is gladly accepted--we *will* bother you about it :)

Here is my "help" :

        Chris Provenzano has implemented a pthreads library that is
        based on the POSIX1003.4a Draft 8 pthread standard. It works
        on Linux as well as on *BSD and Sparc and ....
        
        It is available on  sipb.mit.edu in pub/pthreads.


                Herve Soulard.

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

From: pdcruze@iinet.com.au (Patrick D'Cruze)
Subject: Re: Linux Intrntl & Auto Cfg project
Date: 21 May 1994 18:33:20 +0800

cmay@titan.ucs.umass.edu (CHRISTOPHER M MAY) writes:

>Hi, I just saw a solicitation for "volunteer hardware experts"
>to write for the AutoConfiguration Project for Linux International.

>I forget, is LI a "non-profit" organization which is truly working 
>for the betterment of Linux?  Why did they post outside the Linux
>groups _FIRST_?  I keep up with these groups alot, and if it had been
>handled correctly, I might have joined. 

Point of fact.  We didn't post to the comp.sys.ibm.pc.hardware
newsgroups first.  We made a posting to comp.os.linux.announce almost
a full week BEFORE the postings to the ibm newsgroups.

>From what I remember, though,
>it seems that the general linux community was against their autoconfig
>project.  Please inform me, I don't know it all when it comes to linux
>politics.  I don't want to rub anybody the wrong way, but seems like
>a furthering of the fragmentation that we're seeing so much of lately.

We're not attempting to introduce any fragmentation at all or to get
involved in any politics.  Linus and the other kernel developers have not
decided to use the loabale modules within the Linux kernel due to a number
of concerns they have about it.  Hence it is an "unsupported" modification.
They are working on a more robust version though.  This has nothing to do
with autoconfiguration though.  We're just attempting to get a few people
together to work on a "plug and play" package for Linux which should work
for both loadable drivers technologies.  We have not voiced support for
either implementation but rather for both.

As for whether this is a good thing or a bad thing - each person will have
opinions.  I believe that this is a just a little thing which is going
to make Linux easier to setup.  Its not going to reduce the complexity
of the system in any way and is not going to have much of an impact upon
newbie users.  I could be wrong here but I thought the whole point
of using a computer was to unload as much of the work onto the computer
as possible.  Why should I have to go and manually set IRQ's, DMA
addresses, specify which cards I've got, etc when I could get the
computer to do it all?  Its one less task I have to do in maintaining
and configuring my system.

>Can't people just work on drivers and real hardware support, and leave
>the distributions as they are?

Why can't people work on drivers and real hardware support AND distributions?
The 2 are not mutually exclusive.

>--

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

Patrick D'Cruze                                 pdcruze@orac.iinet.com.au
Linux International

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

Crossposted-To: comp.os.linux.misc
From: damin@csuohio.edu (Greg Boehlein)
Subject: Re: COMMODORE CALLS IT QUITS
Date: Thu, 19 May 1994 02:16:07 GMT

Grant Edwards (grante@reddwarf.rosemount.com) wrote:
: Brandon S. Allbery (bsa@kf8nh.wariat.org) wrote:
: : grante@reddwarf.rosemount.com (Grant Edwards) says:
: : +---------------
: : | Nope, there was a dual-mode 8/32 bit wide 6502 compatible cpu (I don't
: : | remember the number, but 68C832 doesn't sound quite right).  
: : +------------->8

: : 65C816, but that was an 8/16 bit microprocessor.  (The Apple IIGS
: : used it; I don't think any other "major manufacturer" machines did.)
: : I don't think there was a 32-bit version.

: Yep, that's the one -- I mis-remembered the bus width.  Did the IIGS
: even run the thing in 16 bit mode?

Yes, the IIGS ran GS/OS, a full 16 bit OS. In fact, there is a "unix"
kernel called GNO that runs under the OS and offers a lot of the
features of a Unix operating system. My GS is jacked up to 14 mhz and
runs GNO with 4 megs of Ram. It's really quite cool! :) I can DL in
the background, and Emacs / Compile programs in the foreground. YOu
can have up to 64 concurrent procesess operating.

Actually, Derek Taubert has just finished up his TCP/IP package for
GNO that should make my machine -REALLY- useful.

The 65C02 and 65816 processors were a REALLY cool architecture when
they were released. Mine is -STILL- serving me well..
    

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

From: rob@pe1chl.ampr.org (Rob Janssen)
Subject: Re: SIGHUP - Deep Kernal Guts question!
Reply-To: pe1chl@rabo.nl
Date: Sat, 21 May 1994 08:57:49 GMT

In <1994May20.181126.9756@light-house.uucp> las@light-house.uucp writes:

>Alan Cox (iiitac@uk.ac.swan.pyr) wrote:

>I am sorry to post this but I have no other way of contacting you Ian.

>Could you please use a valid reply adress in your posts?

>Your site "uk.ac.swan.pyr" is not in the DNS, and many times my replies
>to you have bounced. 

Well, the English prefer to write their addresses backwards, for some
obscure reason.  When you see an address like "uk.xx.yy.zz", and it
bounces, try "zz.yy.xx.uk".

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

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

From: metcalf@CATFISH.LCS.MIT.EDU (Chris Metcalf)
Subject: Using IDE disks where BIOS claims more than 16 heads
Date: 19 May 1994 03:47:47 GMT

I just installed Linux (Slackware 1.2.0) on a client's Pentium box.
The disk was an IDE WDC AC2540H, 515M, with 1048 cylinders, 16 heads, and
63 sectors.  DOS (or BIOS?) apparently doesn't like >1024 cylinders, so
BIOS lies and says the disk is 524 cylinders with 32 heads.  Linux can't
accept this; the head field only has 4 bits, ORed with the drive number,
according to the code in hd.c:hd_out().  

I went ahead and configured LILO to force Linux to accept the real
geometry.  However, this means that fdisk generates warnings about
partition 1, the 32M DOS partition at the beginning of the disk.  It also
apparently means that DOS can't use its MBR/activate mechanism to find
LILO in hda2; it fails with "No operating system" or some such before
finding LILO.  Even with LILO as the MBR, the boot still failed halfway
through until I added "linear" to lilo.config.  

My question is why DOS can handle this BIOS-generated pseudo-geometry,
when Linux can't?  Ideally, I'd like to use the same parameters DOS does,
since I certainly can't get DOS to change.  Also, is it safe for me to
use the 6 meg on cylinders 1025-1047?  (I'm not now, and probably won't;
this is an academic question.)

I've set followups to me, since this is not really a discussion question.
I will summarize back to the group.
-- 
                        Chris Metcalf, MIT Laboratory for Computer Science
                        metcalf@lcs.mit.edu   //   +1 (617) 253-7766

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

From: chet@hela.ins.cwru.edu (Chet Ramey)
Subject: Re: SIGHUP - Deep Kernal Guts question!
Date: 19 May 1994 04:23:22 GMT

In article <1994May13.042516.27841@loreli.ftl.fl.us>,
Sean Puckett <nate@loreli.ftl.fl.us> wrote:
>We're developing a BBS under Linux 1.0, and we've found that 
>many applications, including bash, ksh and perl, do not pass
>on the HUP signal from do_tty_hangup (tty_io.c)'s call to 
>kill_sl (exit.c).

I don't know what version of bash you're running, but bash-1.13
sends SIGHUP to all jobs (and SIGCONT to stopped jobs) if it is
running as a login shell and receives a SIGHUP.  This is what a
Posix shell should do (or at least a shell running on a Posix.1
system).
-- 
"You can watch an actor absolutely sabotage a good script and then read
 reviews like 'Unfortunately, even the impressive talents of Cheech Marin
 could not salvage Anton Chekhov's trite and meandering script.'"
Chet Ramey, Case Western Reserve University     Internet: chet@po.CWRU.Edu

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

Crossposted-To: comp.sys.ibm.pc.misc
Subject: Re: Cache-optimizing page allocation 
From: boekhold@morra.et.tudelft.nl (Maarten Boekhold (Who'd you expect??))
Date: 21 May 94 00:26:19 +0200

Harm Hanemaaijer (hhanemaa@cs.ruu.nl) wrote:

> One thing I did was write an optimizing malloc that allocates are large
> pool of pages and then mmap's pages together for allocation requests,
> keeping track of how many pages have been used for every cache page position
> (there are 64 with a 256K direct-mapped cache and 4K pagesize).

On a similar subject:

I bought a DDJ today (june issue). It has an article about memory management.
It's about writing memory efficient programs (in this case, C and C++). 
In the C++ case, it shows how one can overload the 'new' and 'delete' 
operators, using memory pools with free blocks of fixed size, so each 
class has it's own memory pool, which is optimally filled (no holes), and 
reduce trashing a lot. It mentions some commercial memory managers for 
DOS (libraries which provide some elementary functions) which do this 
kind of things......

Wouldn't it be a good idea to add such functions to the c-library, so 
every programmer can use a standard set of these functions if they wish 
to do so? They are relatively easy to use in classes (see the article, 
pg. 52, src at pg. 96) for an example.

These functions could provide a great speedup in applications that use 
large (and I mean large :-)) amounts of memory (in the article, an 
example of a homogeneous linked list, with sizes from 20,000 to 150,000 
elements). especially in data-sets which are larger then the amount of 
physical memory, when swapping occurs, speedup could be dramatic (the 
article mentions speedups of a factor 40 in some cases.....)

Maarten
boekhold@morra.et.tudelft.nl

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

From: mitchell@mdd.comm.mot.com (Bill Mitchell)
Crossposted-To: comp.unix.programmer
Subject: regex, regexp, posix, and linux
Date: 20 May 1994 23:35:53 -0700
Reply-To: mitchell@mdd.comm.mot.com (Bill Mitchell)

I'm trying to build less-177 for linux.  I previously built it ok
under linux 0.99pl14, but I'm now running a 1.x system and it won't
rebuild.  It now chokes with undefined symbols for _regcmp and _regex.
In trying to track this down, I looked in /usr/include/regexp.h, and
found a note that it's archaic and that I should be using the posix
regex interface instead.  I have a /usr/include/regex.h file, and
tried including <regex.h> instead of <regexp.h>, but got the same
undefined symbols.

I've got a copy of the O'Reilly POSIX Programmer's Guide, but haven't
found any information on the subject of regular expression library
functions in there.

It's nearly midnight, and I'm stumped for now.  If anyone can give
me some help or a pointer to some help, I'd appreciate it.  Email
response if possible, please.

Thanks a lot.


-- 
mitchell@mdd.comm.mot.com (Bill Mitchell)


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

From: grober1@tisdec.tis.tandy.com (Greg Robertson)
Subject: Re: ncurses weirdness
Date: 18 May 1994 12:30:13 -0500

C. van Rij (cvrij@cs.vu.nl) wrote:
: I tried to port some MS-DOS apps. (using a lot of low-level tricks)
: to Linux using ncurses; This offers enough video/keyboard functions,
: but.. it doesn't quite seem to work! (Ok, i might be missing something)
: The idea is: creating 2 windows (w1= newwin(0,0,0,0), same for w2), 
: filling these (wprintw(w1,"%s",string), same for w2), and putting 
: the first on screen (wrefresh(w1)), and after a keystroke, switching
: to w2 (wrefresh(w2)). The first problem occurs with putting w2 on screen:
: the second line from w1 remains on screen! (Also after clear())
: Another problem occurs when going back to w1: it's blank!
: Am I missing something obvious?
: (I hope this is the place to ask, c.o.l.help didn't seem appropriate)
: I tried this with ncurses 1.8.1 and 1.8.5, by the way.

: Cheers,
: Casey Ryder

Casey,

  Use wclear(win); to clear a specific window.  clear() acts upon the 
window called stdscr so if you will use wclear(w2); it should clear 
the second window only.

  Greg
--

+------------------------------+-----------------------------------------+
| Greg Robertson               | Tandy Information Services              |
| Production Control           | Tandy Technology Square, Suite 1431     |
| grober1@tis.tandy.com        | 200 Taylor Street, Fort Worth, TX 76102 |
| Voice: (817) 870-0879        +-----------------------------------------+
| Fax:   (817) 390-2132        | It doesn't hurt to ask!                 | 
+------------------------------+-----------------------------------------+

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

From: g609296@win.or.jp (Barry Yip kam-wa)
Subject: Mosaic2.4+Term hangs
Date: 19 May 94 05:33:14 GMT

Since the name server here is brain dead, one have to give IP address in
order to reach outside such as when using Mosaic or FTP. However
sometimes in Mosaic when I choose a resource it can't find out (probably
can't resolve the domain name address), it just hangs there. This only
happens intermittently (one out of a few times), otherwise I will get
an error saying unable to open the requested resource which I think is
normal and ok. Usually after it hangs, Mosaic won't work even I restart it.
One must exit term and restart again.

On the comm terminal, I get `gethostname: Bad file number'. I guess
there should be better way to handle this situation though don't the
problem is mosaic or term.

The remote side is Sun OS and local side is linux. Both running term
1.15.

BTW, both mosaic and term are great. Thanks to the developers. I probably
owe someone a postcard :-)

--
Barry Yip
g609296@win.or.jp

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

From: orr@progress.COM (James Orr)
Subject: O_SYNC support
Date: 19 May 1994 06:14:59 GMT

I done some looking through the latest source I could find, and
it appears that the O_SYNC flag to the open() call is still not
supported. Therefore, my question is -- is this correct?, and if
so, is anyone working on adding this support. I need this support
to gaurantee file integrity, and if it is not available, I will 
proceed with fsync, but before I do this, I just wanted to be
sure.

Thanks,

-Jim Orr
orr@progress.com
--
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
James Orr - QA Developer             | Progress Software Corporation
orr@bedford.progress.com             | 14 Oak Park
(617) 280-4635                       | Bedford, Mass. 01730

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


** 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
******************************
