Linux-Development-Sys Digest #94, Volume #8 Tue, 22 Aug 00 14:13:19 EDT
Contents:
Using block devices from user space. (liran)
Re: crash or adb? other kernel debug tools ([EMAIL PROTECTED])
Re: purify and memory managers (Szabolcs Csetey)
kenel remains the same... ("Christos Karayiannis")
Re: how to insert my protocol into the linux protocol stack? (Grant Edwards)
Re: kenel remains the same... (Stefaan A Eeckels)
Re: purify and memory managers ([EMAIL PROTECTED])
Performance tuning tools ([EMAIL PROTECTED])
serial port access with read() delays ([EMAIL PROTECTED])
Message queues (IPC) under Linux Redhat 6.0 (Laurent Pourchet)
Re: Adding a system call dynamically (Kasper Dupont)
Re: Performance tuning tools (Andi Kleen)
Re: how to insert my protocol into the linux protocol stack? (Andi Kleen)
Re: kenel remains the same... (Kaz Kylheku)
Re: all threads in a process share the same pid? (Michael Wojcik)
Re: how to insert my protocol into the linux protocol stack? (Grant Edwards)
Re: Camera Application software run on Linux PC (David Konerding)
Re: how to see large prinouts from the kernel (Brian Horton)
Re: Signal to a process (Peter)
Re: "Best" x86 Linux C/C++ compiler?? ("H.W. Stockman")
NT/HFS-Style Multiple "Resources" In A Single File (Questioner)
----------------------------------------------------------------------------
From: liran <[EMAIL PROTECTED]>
Subject: Using block devices from user space.
Date: Tue, 22 Aug 2000 05:00:33 -0700
Hi,
I am writing a storage manager in user space, I need to
manage several disks ,
the question is it possible to make I/O on a block device
from user space without using the
buffer cache or at least with guarantee that the buffer
cache will always go down staight to the
physical device.
Thanks Liran.
* Sent from AltaVista http://www.altavista.com Where you can also find related Web
Pages, Images, Audios, Videos, News, and Shopping. Smart is Beautiful
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: crash or adb? other kernel debug tools
Date: Tue, 22 Aug 2000 12:05:33 GMT
Hi,
We from IBM have released IBM Dynamic Probes an automated kernel
debugger.
Follow this link to know more about it and to download it:
http://oss.software.ibm.com/developerworks/opensource/linux/projects/dprobes
Hope this helps
thanx
Subodh
In article <[EMAIL PROTECTED]>,
Brian Horton <[EMAIL PROTECTED]> wrote:
> What tools exist to debug linux kernel code, other than liberal
> printk's? Is there a version of adb or crash for linux? I would prefer
> to have something that doesn't cause me to rebuild the kernel (I'm
just
> building and installing my own device driver and kernel module), but
> will consider anything!
>
> thx.bri.
>
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: Szabolcs Csetey <[EMAIL PROTECTED]>
Subject: Re: purify and memory managers
Date: Tue, 22 Aug 2000 13:04:55 GMT
"Rob Love" <[EMAIL PROTECTED]> wrote:
> I have written a fairly big program in C++ that does a bunch
> of dynamic memory allocation. I think that I clean up after myself
> pretty well, but I would like to double check.
I've found yamd the most powerful: http://www3.hmc.edu/~neldredge/yamd/
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: "Christos Karayiannis" <[EMAIL PROTECTED]>
Subject: kenel remains the same...
Date: Tue, 22 Aug 2000 16:29:04 +0300
I am trying to change Linux code at the following way (which is mine)
>From /usr/src/linux/net/ipv4 I add code inside udp.c and then I go to
linux directory and make:
make dep ; make clean ; make zImage;
I also do cp /usr/src/linux/arch/i386/boot/zImage
/boot/vmlinuz-2.0.36.-0.7
and the kernel keeps working the old way. Why nothing has changed?
Christos
------------------------------
From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: how to insert my protocol into the linux protocol stack?
Date: Tue, 22 Aug 2000 14:10:55 GMT
In article <[EMAIL PROTECTED]>, Andi Kleen wrote:
>Josef Moellers <[EMAIL PROTECTED]> writes:
>> [EMAIL PROTECTED] wrote:
>>
>> > im trying to implement a data link protocol and i wonder how i
>> > can enter this protocol in the kernel protocol stack?? any
>> > books or internet sources will be welcome
>>
>> UTSL (Use the source, Luke)! You may want to look at the linux
>> sources, e.g. the TCP code.
>
>It is probably easier to read packet(7) instead and implement
>the protocol in user space.
It's actually pretty simple to do it in a driver.
You just call dev_add_pack() with a pointer to a struct
containing the Etherenet protocol number and a pointer to a
callback routine to be called when packets of that type are
received.
--
Grant Edwards grante Yow! "DARK SHADOWS"
at is on!! Hey, I think
visi.com the VAMPIRE forgot his
UMBRELLA!!
------------------------------
From: [EMAIL PROTECTED] (Stefaan A Eeckels)
Subject: Re: kenel remains the same...
Date: Tue, 22 Aug 2000 16:15:35 +0200
In article <[EMAIL PROTECTED]>,
"Christos Karayiannis" <[EMAIL PROTECTED]> writes:
> I am trying to change Linux code at the following way (which is mine)
> From /usr/src/linux/net/ipv4 I add code inside udp.c and then I go to
> linux directory and make:
> make dep ; make clean ; make zImage;
> I also do cp /usr/src/linux/arch/i386/boot/zImage
> /boot/vmlinuz-2.0.36.-0.7
> and the kernel keeps working the old way. Why nothing has changed?
You need to add the new kernel to the lilo cofiguration
(/etc/lilo.conf) and run "lilo" in order to make the
kernel bootable from your LILO prompt.
Take care,
--
Stefaan
--
This was a thread between ignorant people until I jumped in.
-- Richard Kulisz in g.m.d.
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: purify and memory managers
Date: 22 Aug 2000 10:09:40 -0400
Look at njamd (not just another memory debugger). It is easier to use
than any of them.
------------------------------
From: [EMAIL PROTECTED]
Subject: Performance tuning tools
Date: Tue, 22 Aug 2000 14:34:32 GMT
Hi all
I need to do some performance tuning on a large scale
IO intensive multi-threaded app. I usually use
sar and/or iostat to help determine what the OS
is doing under heavy load, but they don't appear to
be available on Linux.
What have you all used to do serious performance
monitoring/tuning?
Any help appreciated
Ed
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: [EMAIL PROTECTED]
Subject: serial port access with read() delays
Date: Tue, 22 Aug 2000 15:11:41 GMT
Hey there,
After lots of late hours, testing, trying and guessing I need
help of someone who�s familiar with Linux i/o-port-access.
I try to read just one byte with the read()-funktion from a hardware-box,
connected to /dev/cua1 (8databits,2 stopbits, 4800-19200baud).
That box` answertime is guaranteed within 100 ms.
But there�s a "big" delay of 8 ms between each calls of read().
It must work faster, because it�s called for every byte!
We made a write() directly after a read(), just for testing wheter read()
or write() is delaying, and stopped the time between that two calls with
an oscilloscope.
read() is returning immediately after calling, but most of the times
without a read char.
We guess Linux�s kernel is doing some other threads within that read()
and write() calls.
This is the stuff:
INT nRecv,
nRead = 0,
cCount = 0,
nMaxTry = 30000;
do {
nRecv = read (pDev, &nRead, 1); // pDev is the open File_Handle
//nRecv = write (pDev,&nRead, 1);
cCount++;
} while ((nRecv <= 0) && (cCount < nMaxTry));
Is Linux first reading to buffers and then, if there�s no other thread to
do
it�s telling read() that it has read a char or is Linux polling that
ports by
interrupts on several times?
cCount is counting to about 2500 until a char is recognized.
Is there anybody who could tell me how that delay is built, and wether I
could speed that up or not?
Could I force that read() with any termios-constants to return
immediately after recognizing a char?
Thank�s a lot
Markus
Sent via Deja.com http://www.deja.com/
Before you buy.
------------------------------
From: Laurent Pourchet <[EMAIL PROTECTED]>
Subject: Message queues (IPC) under Linux Redhat 6.0
Date: Tue, 22 Aug 2000 17:21:29 +0200
Hello!
I would like to use a message queues (Interprocess Communication) with
no stand by when receiving from the message queues: function msgrcv()
with IPC_NOWAIT flag set. But msgrcv() stay in stand by...
If you have an example who perform using a message queues and IPC_NOWAIT
flag set in msgrcv(), I take it!
Thanks.
------------------------------
From: Kasper Dupont <[EMAIL PROTECTED]>
Subject: Re: Adding a system call dynamically
Date: Tue, 22 Aug 2000 17:29:16 +0200
Marc SCHAEFER wrote:
>
> Kasper Dupont <[EMAIL PROTECTED]> wrote:
> : posibility to have. For some purposes a /proc interface might
> : be ineffective or even imposible.
>
> Another way is to register a new dynamic misc module (with a minor
> in the reserved range), and to add a few ioctl()s.
>
> This will be even portable between kernel revisions (I have such
> modules running in 2.0.x and 2.2.x, with just a minor differences
> in e.g. release()).
That will be a good solution for a lot of purposes,
but it will not work if you want to remain compatible
with existing software expecting a syscall interface.
For example I am currently considering to make a
module implementing the vm86 system call.
--
Kasper Dupont
------------------------------
From: Andi Kleen <[EMAIL PROTECTED]>
Subject: Re: Performance tuning tools
Date: 22 Aug 2000 17:16:52 +0200
[EMAIL PROTECTED] writes:
> Hi all
>
> I need to do some performance tuning on a large scale
> IO intensive multi-threaded app. I usually use
> sar and/or iostat to help determine what the OS
> is doing under heavy load, but they don't appear to
> be available on Linux.
>
> What have you all used to do serious performance
> monitoring/tuning?
>
> Any help appreciated
Standard Linux unfortunately does very poorly on IO statistics.
You can use Stephen Tweedie's sard patches to get some better stats
about the device driver queues, available from ftp.uk.linux.org
In case you're using 2.2.16 or 2.4 also consider tuning the elevator
using elvtune (again a bit poorly documented so far), the defaults there
are ATM a bit suboptimal.
-Andi
------------------------------
From: Andi Kleen <[EMAIL PROTECTED]>
Subject: Re: how to insert my protocol into the linux protocol stack?
Date: 22 Aug 2000 17:14:08 +0200
[EMAIL PROTECTED] (Grant Edwards) writes:
> In article <[EMAIL PROTECTED]>, Andi Kleen wrote:
> >Josef Moellers <[EMAIL PROTECTED]> writes:
> >> [EMAIL PROTECTED] wrote:
> >>
> >> > im trying to implement a data link protocol and i wonder how i
> >> > can enter this protocol in the kernel protocol stack?? any
> >> > books or internet sources will be welcome
> >>
> >> UTSL (Use the source, Luke)! You may want to look at the linux
> >> sources, e.g. the TCP code.
> >
> >It is probably easier to read packet(7) instead and implement
> >the protocol in user space.
>
> It's actually pretty simple to do it in a driver.
>
> You just call dev_add_pack() with a pointer to a struct
> containing the Etherenet protocol number and a pointer to a
> callback routine to be called when packets of that type are
> received.
You just have to learn lots of kernel code specific rules then, which
are not too well documented. It is also harder to debug. Unless you
need to get the most performance out of your new protocol an user space
implementation probably makes more sense, at least initially.
-Andi
------------------------------
From: [EMAIL PROTECTED] (Kaz Kylheku)
Subject: Re: kenel remains the same...
Reply-To: [EMAIL PROTECTED]
Date: Tue, 22 Aug 2000 16:14:34 GMT
On Tue, 22 Aug 2000 16:29:04 +0300, Christos Karayiannis
<[EMAIL PROTECTED]> wrote:
>I am trying to change Linux code at the following way (which is mine)
>From /usr/src/linux/net/ipv4 I add code inside udp.c and then I go to
>linux directory and make:
>make dep ; make clean ; make zImage;
That's a waste of time if you just made changes to a single source file.
Just do make zImage; it should only recompile net/ipv4/udp.c and then
link the whole thing.
>I also do cp /usr/src/linux/arch/i386/boot/zImage
>/boot/vmlinuz-2.0.36.-0.7
^
Is that extra period a spelling error that you only made in the
Usenet posting?
>and the kernel keeps working the old way. Why nothing has changed?
What does your /etc/lilo.conf say? Is there an entry actually pointing
at /boot/vmlinuz-2.0.36-0.7 You should add a new experimental kernel as a
secondary image, rather than overwrite your stable kernel.
Also, you must always re-run lilo when you replace one of the existing
images. The sector map may have changed. It's not enough to just copy
the zImage to the boot directory.
--
Any hyperlinks appearing in this article were inserted by the unscrupulous
operators of a Usenet-to-web gateway, without obtaining the proper permission
of the author, who does not endorse any of the linked-to products or services.
------------------------------
From: [EMAIL PROTECTED] (Michael Wojcik)
Crossposted-To: comp.os.linux.development.apps
Subject: Re: all threads in a process share the same pid?
Date: 22 Aug 2000 15:35:23 GMT
Reply-To: [EMAIL PROTECTED]
In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED] (Kaz
Kylheku) writes:
> Note that the SetCurrentDirectory() function in Win32 is also process-wide.
> I just checked MSDN online. (Could it be that this is where the POSIX guys
> have been getting their ideas lately? :)
>
> If Win32 does it, it's usually wrong. QED. :)
On a related note, I just ran into a problem in a multithreaded app
server that runs on Win32 and Linux (among other platforms), where a
piece of Win32 third-party code I call does indeed do a (wholly
unexpected) SetCurrentDirectory under the covers.
Which, in turn, was causing other threads to sporadically fail to
open files using relative pathnames.
Currently I restore the old cwd immediately after returning from the
call, but for production I'll clearly have to be making the call in
a separate process, unless I can get the library fixed.
So there's an example of a process-wide cwd causing significant
trouble. Sure, the immediate fault is in the third-party library,
which shouldn't be messing with the working directory, but it would
be nice if I had some way to compensate for that in my code.
--
Michael Wojcik [EMAIL PROTECTED]
AAI Development, MERANT (block capitals are a company mandate)
Department of English, Miami University
What is it with this warm, quiet, nauseating bond between them?
-- Rumiko Takahashi, _Maison Ikkoku_, trans. Mari Morimoto, adapt. Gerard
Jones
------------------------------
From: [EMAIL PROTECTED] (Grant Edwards)
Subject: Re: how to insert my protocol into the linux protocol stack?
Date: Tue, 22 Aug 2000 16:28:18 GMT
In article <[EMAIL PROTECTED]>, Andi Kleen wrote:
>[EMAIL PROTECTED] (Grant Edwards) writes:
>
>> In article <[EMAIL PROTECTED]>, Andi Kleen wrote:
>> >Josef Moellers <[EMAIL PROTECTED]> writes:
>> >> [EMAIL PROTECTED] wrote:
>> >>
>> >> > im trying to implement a data link protocol and i wonder how i
>> >> > can enter this protocol in the kernel protocol stack?? any
>> >> > books or internet sources will be welcome
>> >>
>> >> UTSL (Use the source, Luke)! You may want to look at the linux
>> >> sources, e.g. the TCP code.
>> >
>> >It is probably easier to read packet(7) instead and implement
>> >the protocol in user space.
I gues we're all assuming it's something that runs on Ethernet. If it's a
serial line discipline module (like PPP), then packet(7) isn't much help.
>> It's actually pretty simple to do it in a driver.
>>
>> You just call dev_add_pack() with a pointer to a struct
>> containing the Etherenet protocol number and a pointer to a
>> callback routine to be called when packets of that type are
>> received.
I meant that adding a protocol to the stack is pretty simple. Implimenting
and debugging the protocol in kernel space is another question.
>You just have to learn lots of kernel code specific rules then, which
>are not too well documented. It is also harder to debug. Unless you
>need to get the most performance out of your new protocol an user space
>implementation probably makes more sense, at least initially.
Definitely. If I were doing it, I'd try to write my protocol code in such a
way that it can be used in either kernel space or user space. Unfortunately,
it's hard to know how do that until you've already done a bit of kernel-mode
programming.
Debugging in user-space is _way_ easier.
Once it's debugged in user-space, you can think about moving it into kernel.
--
Grant Edwards grante Yow! Now I understand the
at meaning of "THE MOD SQUAD"!
visi.com
------------------------------
From: [EMAIL PROTECTED] (David Konerding)
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.networking,comp.os.linux.setup
Subject: Re: Camera Application software run on Linux PC
Date: 22 Aug 2000 16:38:40 GMT
Reply-To: [EMAIL PROTECTED]
On 21 Aug 2000 22:02:31 GMT, bill davidsen <[EMAIL PROTECTED]> wrote:
>
>In article <8msmmb$qrt$[EMAIL PROTECTED]>,
>bill davidsen <[EMAIL PROTECTED]> wrote:
>
>| Is the hardware still available? I have several webcams which I
>| ordered only to find that the model had changed, and by the time there
>| was Linux support the camera was out of production.
>
> And to answer my own question, no.
>
> Since being able to capture pictures is a solid requirement, and none
>of the dozen or so packages for Linux seem to work with a camera which
>is still made, I guess MS wins. I can't tell someone to base their
>business plan on buying used stuff from eBay.
>
> Thanks for the info, lots of these packages should work for personal
>use if you can find the camera at a garage sale near you.
How about the WinTV/Web cam combo from Hauppauge? Not only do you get a webcam,
you get a full TV tuner, "for free"! The bttv driver is the most mature and usable of
the
V4L drivers, anyway.
Also, there's the Winnov videum, which has a linux driver, although it's buggy and
doesn't
work well with many applications (since it only captures 1 field at a time, and many
apps assume they can capture a whole frame).
Dave
------------------------------
From: Brian Horton <[EMAIL PROTECTED]>
Subject: Re: how to see large prinouts from the kernel
Date: Tue, 22 Aug 2000 10:52:08 -0500
printk's also appear in /var/log/messages. 'man syslog.conf' has info on
this as well.
.bri.
Randy Lexvold wrote:
>
> Hello,
> I am developing a hardware driver with RedHat Linux. From the driver code,
> I call the function printk() to print debug info. I have 2 choices:
>
> 1. Without the startx Linux GUI: I have just one Linux screen with a
> prompt. When I start the debugged driver, it quickly overflows the screen
> with debug messages, and then I can't see them all. I also could not find a
> way to redirect the messages to a file so I can see them all.
>
> 2. With the GUI, I can have a terminal window with a prompt. However,
> there no messages printed with printk() from the kernel show up.
>
> So, how can I see large amounts of printk() output in Linux?
>
> Thank you very much,
>
> Mark Galecki
------------------------------
From: Peter <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: Re: Signal to a process
Date: Tue, 22 Aug 2000 19:20:18 +0200
Reply-To: [EMAIL PROTECTED]
Marco Fabiani wrote:
> Hello,
> I would send a sort of signal to a specific process in order to increase
> a variable inside of it:
I had a script that does that (see modified version below, tested only
on suse 6.3!!). It listens for signal 10, then exits (this is your
process B)...
I don't know much C, but I believe it works much the same :
-setup a handler-function for a signal, using signal();
-send this signal from the other process, using kill()
more info is in:
man 7 signal
man 2 signal
man 2 kill
man 8 pidof
apropos signal | more
Have fun,
Peter
Here's the script....
If you run it, and from another shell send a signal 10 (kill -10 <PID>)
to it, the signalhandler Inc() will run, and the script will exit.
<shellscript>
#!/bin/sh
# Signal-handler
Inc() {
A=`echo "$A + 1" | bc`
echo "Updated, variable A is now $A"
}
# Set the new traphandler
trap Inc 10
# Do something usefull
A=1
echo "variable A is now $A , PID is $$"
while [ $A -lt 2 ]
do
sleep 1
done
# Ignore signal 10 from here
trap '' 10
</shellscript>
------------------------------
Reply-To: "H.W. Stockman" <[EMAIL PROTECTED]>
From: "H.W. Stockman" <[EMAIL PROTECTED]>
Subject: Re: "Best" x86 Linux C/C++ compiler??
Date: Tue, 22 Aug 2000 10:50:15 -0700
Thanks for the info, comments below...
"Stefaan A Eeckels" <[EMAIL PROTECTED]> wrote in message
news:[EMAIL PROTECTED]...
> In article <Bbro5.3115$[EMAIL PROTECTED]>,
> "H.W. Stockman" <[EMAIL PROTECTED]> writes:
> > I'd like to purchase a shrink-wrapped x86 Linux package with a
> > "good" C/C++ compiler. I define "Good" to mean: the compiler
> > produces efficient executables for source code comparable to
> > the Spec2000 suite. The main use would be for compiling
> > memory-intensive floating point code, with some middling
> > graphics. An IDE is nice, but not the most important feature;
> > I would like a good, intuitive text editor.
> The standard compiler on Linux is gcc. It was designed
> to produce reasonable code on a wide variety of platforms,
> and it generates reasonable code on x86.
>
> If you want commercial compilers, you'd have to buy them
> separately (check CodeWarrior, and I think Lahey also
> offers a C/C++ compiler).
I've checked the Metrowerks pages pretty carefully, and it still seems
that CodeWarrior for Linux is just an IDE on top of gcc.
[...]
> > Currently I use Intel C/C++ 4.5 under Microsoft VC++ in
> > Windows. I did numerous tests with gcc and egcs back in
> > 1997 and early 1998, and generally found they produced
> > slower executables than did the Intel compiler (by 20 to 50%).
> I'd refer you to a recent comparison in Linux Journal
> <http://www2.linuxjournal.com/lj-issues/issue67/3425.html>,
> which concludes:
>
> | Is Linux a viable alternative to Windows in terms of computational
> | performance? Our results indicate that a substantial application
> | compiled with freeware compilers under Linux can probably be expected
> | to execute no more than 1% slower than when compiled and
> | executed under one of the most efficient Windows NT commercial
[...]
I'd truly like to see comparable performance. Nonetheless,
I read the linuxjournal link pretty carefully and noted:
(1) they tested a single home-grown app with a relatively small
data set -- 4.5 MB (I use up to 400 MB), and it appears
to be a standard image processing app: many memory
reads for every memory write.
(2) They used VC++ . I found Intel C/C++ 4.5 produced
substantially faster code than VC++, on _my_ applications.
> > I'd like to set up a system for dual-boot, and would like to
> > have the Linux system communicate with the 4 windows PCs
> > on my TCP/IP network
> Dual-boot is no problem, and if you want to use your Linux box
> to serve files to the PCs, Samba will do that nicely.
>
> > It would be nice to have conio.h features available in the Linux
> > compiler; it would be nicer still to have some equivalents of
> > the Windows console functions like SetConsoleScreenBuffer().
> I guess you mean "SetConsoleActiveScreenBuffer" (yuck!, the
> lenght of that name).
> By definition, UNIX does "console IO". You'll have to accept
> that the programming model for UNIX using X as a windowing
> system is totally different from Win32 using a simulated
> console. On the whole, there's no need to write a "console
> application", just use an xterm (or rxvt, or kterm...) and
> launch your application. If you need "character graphics",
> or screen positioning, you'll have to get used to using
> "curses". If you're talking about computationally intensive
> programs that do line-by-line output, then the UNIX model
> is a lot easier to program.
I remember writing curses and TeK4010 consoles in Unix. They seemed
awfully analogous to win32 console apps. I've seen conio.h show
up in linux Discussions, so _someone_ must do this sort
of thing. My guess is that x-windows programming must be easier than
win32 GUI programming; I remember my wife writing x-windows apps, and they
seemed far less bogged down in protocol. Perhaps that's because in 1992,
they were still written in C ;^) .
------------------------------
Subject: NT/HFS-Style Multiple "Resources" In A Single File
From: Questioner <[EMAIL PROTECTED]>
Date: 22 Aug 2000 11:02:49 -0700
If someone on kernel-list thinks this might be useful, please pass it on:
I noticed in Kernel Traffic today under "13. NT/HFS-Style Multiple
"Resources" In A Single File" that NT, HFS, and BeOS were mentioned.
I just wanted to note that the old Apollo (later HP) computers running
(IIRC, probably not) Apollo DOMAIN Unix (?over non-Unix Aegis OS?) at
least as (possibly patched and ) configured for use with a source code
control system (the name of which eludes me) supported such file forks.
Talking with some old Apollo people might add some ideas to the mix.
With that system we could refer to (at the shell level, at least) the
latest file version as xxx or get the version 3 as xxx/3, etc. IIRC,
the files were actually kept together in one big compressed thing from
which they were very quickly extracted. I don't know how deep into the
OS it got; probably not very.
------------------------------
** 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: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
