Linux-Development-Sys Digest #314, Volume #6 Thu, 21 Jan 99 08:14:06 EST
Contents:
How can I build a Linux system from scratch - NO distribution? ("Thomas T.
Veldhouse")
Re: disheartened gnome developer ([EMAIL PROTECTED])
How to install glib and gtk+ (�̺�ȣ)
Re: Linux Phase 2: A Consumer Operating System (Alexander Viro)
Hacking glibc's dynamic linker (Dave Bailey)
Re: Will 2.2.x support removable medias better? (Dietmar Kling)
Re: Linux Phase 2: A Consumer Operating System (Rick Onanian)
Re: PROPOSAL: comp.unix.year-2038 (WAS: 2038 and Linux) (Andre Fachat)
Re: K6-400 "kernel paging request" errors (Larry)
kernel: VFS: No free inodes - contact Linus (Matt Corddry)
Re: Linux Sound Engine (Dickon Reed)
Re: 2.2.0-pre7 and APM - system clock loses time! (Chris Rankin)
----------------------------------------------------------------------------
From: "Thomas T. Veldhouse" <[EMAIL PROTECTED]>
Subject: How can I build a Linux system from scratch - NO distribution?
Date: Wed, 20 Jan 1999 10:43:14 -0600
I would like to build a Linux system from scratch, without using a premade
distribution as a starting point. I don't know exactly where to begin.
I would like the glibc2 libraries as a base. I would need the the gnu
development tools and the sysvinit. However, how do I create the partitions
and get things in place without installing a distribution of linux in the
first place (and thus all the extra bagage I don't want). I would like to
create a system similar to the way FreeBSD does things. Install a very lean
base system with just the development tool required to build the system.
Everything else is a port which can be installed from source. I love the
FreeBSD system of ports, but I would like to base my system on Linux.
Is there a way to put a minimal system together on my hard drive (multiple
ext2 partitions) and get the tools on to it so that I can build what I need
from there?
Also, as a side note, do you thing that using GNU inetutils would be a good
choice for [some of] my networking daemons?
Tom Veldhouse
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Crossposted-To: comp.os.linux.advocacy,comp.os.linux.development.apps,comp.os.linux.x
Subject: Re: disheartened gnome developer
Date: Thu, 21 Jan 1999 07:18:51 GMT
In article <784e1r$[EMAIL PROTECTED]>,
Navindra Umanee <[EMAIL PROTECTED]> wrote:
> My mistake, which I tried to correct by making the code more
> equivalent! In any case, feel free to compare single lines of Qt code
> and GTK C code. In any case, feel free to post equivalent snippets of
> code.
Both GTK and Qt are quite complex and quite different. I don't think posting
random snipplets of code is going to prove anything.
> > You said the code was ugly.
>
> No, I said the *C* code was ugly and I didn't even intend to single
> out GTK from the various toolkits. Why do you feel the need of
> wrongly rephrasing everything I say? What's the point of that if not to
> FUD?
What was the point of slopping out the code example if not to FUD?? Why not
let new developers go thru each tutorial and judge for themselves. Just refer
them to the tutorials.
> > Then you ommitted the comments from both sources, replacing them
> > with your own opinions.
>
> The reason I removed the comments is because it seemed to make the
> code even longer and uglier. I gave you the URLs in *both* cases,
> Perry.
>
> Does any of this change the look'n'feel of the code? Does C code not
> suffer from trying to enforce some kind of OO on top of it?
It goes both ways. In Qt you have to create a new class just to create a
callback function. The only time I need to create a new class in GTK is to
create a custom signal.
> > David is comming from the perspective of is own experiences writing code.
You
> > seem to be making sweeping generalizations. What code have you written with
> > Qt??
>
> Nothing I'd like to mention to you.
Nothing.
> >> Way to go, Perry Pip. If there's anyone FUDing here, it's you.
> >> Followups set to .advocacy.
> >
> > No, your not going to get off that easy. It's pretty obvious from the other
> > two posts you sent this morning you have a pretty strong grudge againsts
FSF,
> > GNU, GPL, Redhat, Gnome etc. etc. And your usenet posting history at
dejanews
> > verifies it. You're the one who's putting out the FUD.
>
> This is completely false. If you have a strong desire to put words in
> my mouth, at least have the decency to waste other people's time only
> in .advocacy.
False?? I suggest you read the posts you made Tuesday morning 1/19/99.
> FSF --> No grudge there. Just pointing out your naive assumptions
> about one organization over another. This is where you paint
> the FSF to be a saintly organization and Troll some form of
> potential evil entity.
I never said FSF was "saintly" nor have I ever said Troll Tech is evil. Who's
putting words in who's mouth??
I will tell you some facts:
FACT #1: FSF is classified as a non-profit, tax exempt charity under section
501c3 of the U.S. Tax code. That means if the people runing FSF tried to
profit from the contributions of other they would face both jail time and
serious tax penalties.
FACT #2: Troll Tech is a resistered corporation in Norway. They have to make a
profit to survive.
FACT #3: FSF licences all it's code under GPL/LGPL, which guarantees that
developers will have never the right to redistribute modifications of the code
the ever taken away from them.
FACT #4: Troll tech has yet to release Qt under any license that allows
developers redistribute modifications of the code.
That's not FUD, that's FACT. Intelligent people know the difference.
> Gnome --> Sure, I didn't/don't like the political ickiness, initial
> anti-KDE propaganda and all that stuff. But I hold no
> grudge and I, in fact, like many of the things I've seen in
> GNOME from a non-political point of view.
>
Like it or not, it's because of Bruce Perens and others that Troll Tech
announced the new license for your toolkit. Yet you flamed Bruce six months
ago and you flamed him again this week. Looks to me like you have a grudge.
============= Posted via Deja News, The Discussion Network ============
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
------------------------------
From: [EMAIL PROTECTED] (�̺�ȣ)
Subject: How to install glib and gtk+
Date: Wed, 13 Jan 1999 17:54:57 -0800
I have glib(1.1.12) and gtk+(1.1.12) in `tar.gz'
format.
how to install them?
just `./configure' and `make'?
*** Posted from RemarQ - http://www.remarq.com - Discussions Start Here (tm) ***
------------------------------
From: [EMAIL PROTECTED] (Alexander Viro)
Crossposted-To: alt.os.linux,comp.os.linux.development.apps,comp.os.linux.setup
Subject: Re: Linux Phase 2: A Consumer Operating System
Date: 21 Jan 1999 03:18:10 -0500
In article <hMyp2.16725$[EMAIL PROTECTED]>,
Christopher Browne <[EMAIL PROTECTED]> wrote:
[snip]
>This is an unkind response.
And it bloody supposed to be one. Excuse me, but WTF do we need marketdroids
on our necks? Sorry, I'm sick and tired of people who can't program their way
out of the manhole (perfectly OK with me) *AND* have a temerity to decide what
should we do. To all such persons: Linux is not controlled by your ilk. Yes,
there are things done without Management. If it shatters your world it's *your*
problem. Suffer or change your parasitic beliefs or scream and die if you can't
survive *that*.
>Funny thing of the day... Corman LISP comes with the (HURL-INSULT)
>package:
[snip]
C|N>K
Try telnet insulthost.colorado.edu 1695
>At any rate, while dramatic insults may drive people away, it does
I don't see anything dramatic here, but then English is not my first
language. Sorry if it was not obvious from my posting, but I really couldn't
care less whether this in-duh-vidual was insulted or flattered.
>little to advance any understanding of how/why the idea of "making Linux
>into a consumer product" is a bad one.
>
>>[snip]
>>>To accomplish this, developers must find out what consumers want.
>>
>>Then go, find it out and do it. And don't DARE to tell "must" to people
>>who don't depend on you and owe you nothing.
>
>This thread came up in discussions yesterday; this is well worth
>elaborating on, perhaps with a little less vitrol.
>
>... Moving along the vitrolic path ...
>
>My local LUG went through a similar-but-opposite problem over the last
>24 hours. Some kid, of what persuasion, we're unable to identify, who
>had joined [EMAIL PROTECTED] a while back, apparently felt that some
>question that he asked was not answered with what he felt was "due
>haste."
Oh, you got a share of those idiots too?
[snip all-too-familiar story]
>One very unfortunate thing was that the fellow seemed to have the
>expectation that he could *demand* that we help him. He was *demanding*
>that our LUG write an "A-Z guide to Linux."
Where does it come from? I don't believe that this kind of expectations
applies only to Linux. Yes, articles in PC Sick re: great community support
*might* be misinterpreted such way, but I don't think that it's the only
reason. Where do they get infected by that?
--
"You're one of those condescending Unix computer users!"
"Here's a nickel, kid. Get yourself a better computer" - Dilbert.
------------------------------
From: Dave Bailey <[EMAIL PROTECTED]>
Subject: Hacking glibc's dynamic linker
Date: Thu, 21 Jan 1999 10:34:27 +0000
Well, sort of. I'm trying to write a program that
lists all dynamically linked ELF files in $PATH which
have a dependency on libxxx.so, libyyy.so, libzzz.so,
etc, where xxx, yyy, zzz, ... are supplied as the
arguments to the program. The program would work
sort of like ldd, only backwards - instead of listing
the dependencies of a program, it would list the
programs with a given dependency. I'm doing this
because whenever I upgrade gtk, everything breaks
and I need to know what needs to be rebuilt. I
realize that I could ldd everything in $PATH, send
the output to a file, and process the file in search
of binaries with a particular dependency. But I sort
of wanted to spend some time learning stuff about the
run time dynamic linker, don't ask me why.
So far I've learned that ldd works by setting an
environment variable, verifying that the argument is
an ELF binary, then execl'ing the program in question.
The run-time linker tests for the environment variable
when it's loading the program. If the variable is set,
it just prints out the names of the libraries it would
need to load, then it bails out.
Is it possible for me to extract the part(s) of the
RTLD in glibc2 which determine the dynamic dependencies
of an ELF? It should be just a matter of reading the
header in the right way, right? I'm just starting on
this, and wanted to make sure I'm not about to waste
tons of time.
--
Dave Bailey
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Dietmar Kling)
Subject: Re: Will 2.2.x support removable medias better?
Date: Thu, 21 Jan 1999 11:16:30 GMT
On Tue, 19 Jan 1999 07:22:54 +0100, [EMAIL PROTECTED] wrote:
There _are_ ways as far as I know to detect if a user wants to change
a cdrom / tape .
In the IDE specs is a bit (in scsi too) where you can check if the
user has pressed the eject-button -> you could umount the mounted
media. There is a bit for media inserted too.
Nobody seems to like though (but me).
(Search for "IDE Automount" topic in this newsgroup).
Regards
Dietmar
>With 2.0.x and removable medias you have to care about mount/umount or
>use mtools if the media has a FAT fs.
>
>supermount addresses this issue in older kernels. Now I wonder how 2.2.x will
>care for removable media?
>
>Ideally, I would want to be able to flip in a media (Floppy, CD, ZIP,
>whatever) and use it right away without mount or umount. It should be allowed
>to remove the media whenever the activity light is off.
>
>supermount is rather old, anyone knows why it didn't find a way into the
>official kernel?
>
>--
>Olav "Mac" W�lfelschneider [EMAIL PROTECTED]
>PGP fingerprint = 06 5F 66 B3 2A AD 7D 2D B7 19 67 3C 95 A7 9D AF
>Things which try to look like things often look more like things than things.
------------------------------
From: Rick Onanian <[EMAIL PROTECTED]>
Crossposted-To: alt.os.linux,comp.os.linux.development.apps,comp.os.linux.setup
Subject: Re: Linux Phase 2: A Consumer Operating System
Date: Wed, 20 Jan 1999 14:45:28 -0500
Christopher Browne wrote:
> Note that if things change such that these changes really do take place,
> this will make Linux a much less interesting platform for the people
> that are interested in "playing with an OS" for its own sake.
>
> With the result that many of the sorts of folk that made Linux grow into
> what it has become are liable to migrate off to other systems that
> permit them to "play" again. Whether that's Hurd, Xos, FluxOS, *BSD, or
> whatever.
>
> At that point, Linux starts to die. It may continue on for years, but
> if the sources of innovation leave, that causes great injury.
This result is not required. Noone is required to kill Slackware, or Debian, or your
favorite distribution. I would envision a new distribution set up commercially similar
to...well...Microsoft. :) Granted, if one does this, than most other distributions will
want to follow, but I think that basically, it will still be as play-with-able as it
was
before.
Just my two cents. (Imagine how cheap my thoughts would have been before 40 years of
inflation? <G>)
-Rick
------------------------------
From: Andre Fachat <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.misc,comp.os.linux.advocacy,comp.software.year-2000,comp.unix.advocacy
Subject: Re: PROPOSAL: comp.unix.year-2038 (WAS: 2038 and Linux)
Date: 21 Jan 1999 11:04:35 GMT
In comp.os.linux.development.system Bloody Viking <[EMAIL PROTECTED]> wrote:
> Yeah, before the time_t expiry, you tar up the ext2fs partitions, move the
> tars, then format the ext2fs partitions into ext3fs, and untar right back.
I'm no expert, but how does tar save the dates of a file?
Andre
--
Email address may be invalid. Use "fachat AT physik DOT tu-chemnitz DOT de"
======Fight SPAM = join CAUCE http://www.cauce.org======Thanks, spammers...
Andre Fachat, Institute of physics, Technische Universit�t Chemnitz, FRG
http://www.tu-chemnitz.de/~fachat
------------------------------
From: [EMAIL PROTECTED] (Larry)
Crossposted-To: linux.dev.kernel,comp.os.linux.hardware,comp.os.linux.misc
Subject: Re: K6-400 "kernel paging request" errors
Reply-To: [EMAIL PROTECTED]
Date: 20 Jan 1999 21:16:35 GMT
On Wed, 20 Jan 1999 08:13:13 -0500, David R. Bergstein <[EMAIL PROTECTED]>
wrote:
>If it helps at all, I am also seeing similar paging request errors under
>linux 2.0.36 with an AMD K6-200 and 128MB RAM. I will need to change
>my syslog.conf to obtain a dump next time it happens (will post).
>
>
>
>[EMAIL PROTECTED] wrote:
>>
>> Suffering from an unstable system.
>> K6-400 (stepping 12), Motherboard FIC PA-2013 (VIA MP3),
>> 256 MB Ram (PC-100), (the board allows to downclock the RAM to 66 Mhz, what I
>> did),
>> AGP Matrox G200, 2 SCSI-Controller, EATA-DPT (only Disks)
>> and ncr53c825 (DDS-3, CDROM ).
>> RedHat 5.2 Kernel 2.0.36 and I tried as well all 2.2.0-preXX. the last
>> 2.2.0-pre7ac2.
>> The system keeps chrashing.
>>
>> I am trying to fix the system now since christmas. Getting frustrated....
>>
>> Any ideas, suggestions??
>>
>> Thanks in advance
>>
>> Mario Dix
>>
>> This is from 2.0.36:
>>
>> Unable to handle kernel paging request at virtual address e8f7ce98
>> current->tss.cr3 = 083ac000, %cr3 = 083ac000
>> *pde = 00000000
>> Oops: 0000
>> CPU: 0
>> EIP: 0010:[<00125af0>]
>> EFLAGS: 00010202
>> eax: 28f7ce98 ebx: 08c30814 ecx: 00000400 edx: 00000025
>> esi: 00000814 edi: 00067831 ebp: 00000400 esp: 0313de68
>> ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
>> Process tar (pid: 3707, process nr: 53, stackpage=0313d000)
>> Stack: 08c3f398 00067831 00000814 00000002 00000000 00000025 0012702d 00000814
>> 00067831 00000400 0313df18 00002484 00000000 0dfddf00 00067831 00000814
>> 00000100 08c3f498 036c0814 00000000 08c3f498 08c3f418 036cf918 0313df14
>> Call Trace: [<0012702d>] [<001274af>] [<0011d25c>] [<0011d341>] [<0011d6b4>]
>> [<00123c70>] [<0010a941
I was getting this type of error with a cyrix p166 and finally took the
processor fan off and cleaned it and the fins and put it back together and
VOILA, it's been working famously ever since.
------------------------------
Date: Thu, 21 Jan 1999 02:36:45 -0800
From: Matt Corddry <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: kernel: VFS: No free inodes - contact Linus
Well... the subject really explains my problem. I get hundreds of
thousands of these at console and in /var/log/messages.
I'm running a chat server on a 2.0.36 redhat box. It's a p2-450 with
512m of ram, an adaptec 7890 + quantum 4.5 lvd, and a tulip 21140 nic.
This error occurs when I start stressing my server, which involves a
large number of open tcp connections -- about 2300 or so. I run 40
'chatservers' on ports 5000-5039 to handle the load.
One of our developers modified the kernel to get more user processes and
open files (he followed instructions from the net... I can find out the
changes exactly if need be). A 'ulimit -a' spits out:
core file size (blocks) 1000000
data seg size (kbytes) unlimited
file size (blocks) unlimited
max memory size (kbytes) unlimited
stack size (kbytes) 8192
cpu time (seconds) unlimited
max user processes 1792
pipe size (512 bytes) 8
open files 2048
virtual memory (kbytes) 2105343
The status lines in 'top' looked like this as it was failing:
84 processes: 41 sleeping, 43 running, 0 zombie, 0 stopped
CPU states: 0.9% user, 21.3% system, 0.0% nice, 80.9% idle
Mem: 515440K av, 507048K used, 8392K free, 83040K shrd, 70420K buff
Swap: 128516K av, 88K used, 128428K free 7348K
cached
If anyone knows what's going on, I'd love some help here!
thanks in advance
matt corddry
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Dickon Reed)
Subject: Re: Linux Sound Engine
Date: 21 Jan 1999 11:14:42 GMT
I have done some work on a audio device driver abstractions that might
be useful. In my design and implementation, the DMA buffer is made
available to all applications using sound playback, and the
applications render (ie mix) their audio straight in to the buffer. My
main motivation for doing this was to reduce latency, and I got
results which were basically as good as the sound card I was using can
do. See http://www.cl.cam.ac.uk/~dr10009/audio/ for an academic short
paper on the work I did.
My work was done on Nemesis
(http://www.cl.cam.ac.uk/Research/SRG/pegasus/nemesis.html), and it is
because of the high quality of service, millisecond timescale
scheduling that we do on Nemesis that I could get the low latency
stuff to work. Basically, for streaming audio, the latency achieved
reliably will be roughly the same sort of timescale as the application
gets to run.
Basic plus and minus points to my scheme on Linux, as I see them:
+ low latency, depending on how often Linux can let applications
run. Different applications get to choose their latencies. If they
pick too low, they just break up but the system is still stable.
+ mixing with all the work being done by clients; if anything, the
kernel does less than it would under something like OSS. It isn't
necessary to make any system calls to do the sample by sample work of
playing, either.
+ no packetisation delay issues
+ clients get to know the exact playout rate of the sound card (ie
44100.05 Hz), which is useful for streaming audio from another audio device.
- It *is* different; IANA kernel hacker, so I don't really know what's
involved in doing this.
- Some thought has to be given to security; basically, if you let
applications play this way, they can make whatever noise they want
come out of the sound card and there isn't really any way to stop
them. So clearly you need to make sure that only authorised in some
sense clients can get DMA buffer access.
- Maybe not all sound cards use DMA buffers which are suitable for
this scheme.
- Different sound cards have different limits on the size of their DMA
buffers. 64K is quite a common buffer size to use; that gives about a
1/3rd second at CD quality. Under heavy load, some processes might not
run that often so they would break up alot.
Quick summary of how it works; check the paper
(http://www.cl.cam.ac.uk/~dr10009/audio/) for slightly more detail:
The device driver exposes a buffer and a control block. The control
block gives enough timing information for the clients to work out
what sample in the buffer the sound card will be playing at any given
moment, using the processor cycle counter for timing. The clients then
choose some safety delay and start reading what is in the buffer
already, adding in their samples and then writing back in to the
buffer the samples with their sound mixed in to it. A volume control
mixer advises clients what volume to mix at, using shared memory. The
device driver periodically zeroes a small part of the buffer after the
sound card has played it.
If we were doing this on Linux, there would be several compatibility
issues.
- For existing programs that use the OSS stuff as a character device
(ie not mmaping it) we'd need to implement probably a kernel service
that mixes their audio in to the shared buffer on their behalf. This
might be a bit of a resource hog; actually, on modern processors I
don't think it would be too bad.
- For existing programs that mmap the DMA buffer of the OSS stuff, we
still need either another user space application or the kernel to act
like a sound card playing that buffer, but actually mix the stuff in
to the shared buffer.
- For device drivers that can't expose DMA buffers, something would
have to copy stuff from the shared "DMA" buffer to the sound card.
Comments?
Dickon
------------------------------
From: Chris Rankin <[EMAIL PROTECTED]>
Subject: Re: 2.2.0-pre7 and APM - system clock loses time!
Date: Wed, 20 Jan 1999 14:17:10 -0500
Chris Rankin wrote:
> I just noticed this: normally, I can suspend the machine (using apm
> v2.4, and YES I am running the apmd daemon - for what it's worth) and
> xclock shows the correct time when I start the machine up again in the
> morining. Well, this is my first morning with 2.2.0 and the system clock
> was not correct. In fact it was about 5 hours slow ...
Clunk! (* rattle! *) The penny drops. I thought my CMOS clock was GMT,
but it isn't. And by a strange coincidence I am 5 hours behind GMT ... I
have adjusted my kernel config and all is well again.
Chris.
------------------------------
** 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
******************************
