Linux-Development-Sys Digest #566, Volume #6 Fri, 2 Apr 99 10:15:12 EST
Contents:
Re: Anyone around with knowledge about memory barriers in SMP environments? (Aurel
Balmosan)
Re: Kernel build problem (Michael Travis)
Re: Programming tools for Linux/Unix: Editor, IDE, Frontend to GCC. (Peter Samuelson)
Re: [ANN] CodeWarrior for Red Hat Linux (Peter Samuelson)
Re: Patch-2.2.5 fails on linux-2.2.4 (Peter Samuelson)
Re: 2.2.[2345] cannot mount MO devices (Igor Zlatkovic)
BUG? fcntl problem in kernel 2.2.4/NFS/mutt (Wout van Albada)
Re: polling an interface at 12 KHz (Peter Samuelson)
Re: Put another way: Linux from the Ground Up (John Forkosh)
Re: Kernel 2.3 when? (Peter Samuelson)
Limit ??? (Vincent)
Re: Linux Kernel documentationI (John Burton)
Re: 4 Gb memory? (Harry Heinisch)
Linux Kernel documentationI (root)
Re: Idea: Make a seperate "i686" tree for Redhat Linux 6.0 (Johan Kullstam)
Re: Compiling??? (Roman)
----------------------------------------------------------------------------
From: Aurel Balmosan <[EMAIL PROTECTED]>
Subject: Re: Anyone around with knowledge about memory barriers in SMP environments?
Date: Thu, 1 Apr 1999 07:56:26 GMT
David Wragg <[EMAIL PROTECTED]> wrote:
> impossible to track down. 99% of the time you should just use mutexes
> or semaphores, because they hide all of these issues, saving you from
> risk and effort.
Thanks David.
My intenion was to find a way around mutexes/spinlocks/semaphores for
reader processes. From your answer it looks like it is possible without
major problems. Currently a specific database implementation needs about
30% system time on DEC-Unix 4.0d and 15% system time on Linux-2.0.36 due
to simple sempahore operations. I expect to gain those percents with the
new implementation.
Bye,
Aurel.
--
================================================================
Aurel Balmosan | [EMAIL PROTECTED], [EMAIL PROTECTED]
http://gaia.owl.de/~aurel/ |
================================================================
------------------------------
From: [EMAIL PROTECTED] (Michael Travis)
Subject: Re: Kernel build problem
Date: Thu, 01 Apr 1999 09:56:03 GMT
Reply-To: NewsClient
It is simple to solve.
You do make bzImage (instead of make zImage).
You may need a more uptodate version of LILO.
Regards,
Michael.
On Sun, 28 Mar 1999 02:05:50 +0000, John Bullock
<[EMAIL PROTECTED]> wrote:
>I am trying to rebuild my 2.0.36 kernel, I go through the config and
>make. Towards the end of the build (when it is making zImage) I get the
>following:
>.............
>tools/build bootsect setup compressed/vmlinux.out CURRENT > zImage
>Root device is (3, 1)
>Boot sector 512 bytes.
>Setup is 4396 bytes.
>System is 521 kB
>System is too big
>make[1]: *** [zImage] Error 1
>make[1]: Leaving directory `/usr/src/linux-2.0.36/arch/i386/boot'
>make: *** [zImage] Error 2
>
>Has anyone seen this, and if so what resolution.
>
>Thanks
>John Bullock
>[EMAIL PROTECTED]
>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.help,comp.unix.programmer
Subject: Re: Programming tools for Linux/Unix: Editor, IDE, Frontend to GCC.
Date: 1 Apr 1999 03:19:16 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Nix <$}xin{[email protected]>]
> Some of us find `evil GUIs' useful. (For viewing DVI files without
> bothering to print them, and for having lots of rows and columns on
> the screen in a decently-sized[1] font.)
> [1] read `200 cols x 160 rows' ;)
I have yet to see any use for more than 116x60 (well, perhaps an
occasional insane piece of ASCII art, or when trying to edit by hand a
wide xpm file, which is really the same thing). I get 116x60 in text
mode and with a very readable font. (Of course, as for readability,
the 21" monitor doesn't hurt.) I find X considerably harder on my
eyes. Also I don't see any advantages of having multiple windows open
at once that I can't get with Linux vt switching and gpm. GUIs aren't
evil, of course -- only superfluous for a lot of what I do.
No flamewars, please: I already know most of you out there disagree. I
also know there are things you need GUIs for -- I work in a CAD shop.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.unix.programmer
Subject: Re: [ANN] CodeWarrior for Red Hat Linux
Date: 1 Apr 1999 05:35:52 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Kendall Bennett <[EMAIL PROTECTED]>]
> The catch is that there *are* so many Linux distributions and they
> don't all follow a standard distribution structure. That to me is
> something that is very much missing from the Linux community, and I
> really hope that the Linux standard distribution effort takes
> hold.
Sounds wonderful in theory. However...
> Specifically ensuring we all standardise on the C runtime library
> (ie: all new distributions use glibc!!),
You like glibc and so do I, but what about distributions that target
small space or memory requirements? Why does everyone need nsswitch?
> the use of RPM as the standard installler
And lose all the convenience of dselect and apt? Is there an RPM-based
equivalent to those? How about dpkg's smooth handling of the config
files you may or may not want to overwrite? My point is not to
resurrect the old flame war, just to note that there are good reasons
not everyone uses RPM -- and good reasons not everyone uses dpkg.
> and standard locations for configuration files and standard locations
> for the Xserver.
Problem: The XFree86 Project prefers /usr/X11R6/lib, The Debian Project
prefers /etc/X11. It's nice to agree on one or the other, but what if
you *don't* agree with whatever the consensus is?
I like having choice.
> This will mean compromises for some of the distributions
There's a reason so many distributions exist. They each have distinct
features and advantages. I have always been confused by Linux
standards efforts, since as I see it, many of them seem to want to make
everyone look like Red Hat (or whomever), and thus eliminate the whole
point of having more than one in existence.
Filesystem hierarchy is a good thing to standardize, and certain
minimal expectations about shell tools available, libraries available,
etc. would make software distributors' lives easier. Which is where
LSB comes in. But there are many things that really do not need to be
standardized.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Patch-2.2.5 fails on linux-2.2.4
Date: 1 Apr 1999 06:01:17 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Mike Dowling <[EMAIL PROTECTED]>]
> Remember to
>
> # cd /usr/src
> # zcat patch-2.2.5.gz | patch -p0
Inflexible and dangerous. What if /usr/src/linux is some other kernel
tree (or symlink), not the 2.2.4 one you want to patch? Also what if
whoever made your patch didn't see fit to call the destination dir
"linux"? Linus always does, but that doesn't mean he always will, and
many other kernel developers who release patches don't.
A much safer habit to get into:
$ cd /usr/src/linux-2.2.4 # or wherever your 2.2.4 tree is
$ gzip -cd ../patch-2.2.5.gz | patch -p1
(Note that I also don't do stuff like this as root.)
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: Igor Zlatkovic <[EMAIL PROTECTED]>
Subject: Re: 2.2.[2345] cannot mount MO devices
Date: Thu, 01 Apr 1999 12:19:25 +0000
Andries Brouwer wrote:
> Very recent versions of fdisk can be used with the `-b 2048'
> option to create partition tables fro these MO devices
> under 2.2.2 or later.
> If there is interest I might try to add a conversion option.
Yep, if you can do that, go ahead :-)
I just tried a bit and saw that what you say is true. MO diskettes formatted
under 2.2.5 work under 2.2.5 normally. MO diskettes formatted under 2.2.1
work under 2.2.1 normally. It is however not possible to exchange information
using MO between, lets say, two computers one running 2.2.5, the other
running 2.2.1 without extra work.
I�m personally using MO for backup and data transfer between machines at home
and in the office. Both machines run the same kernel, so this is not the
problem. Since data on the backup diskettes can also be found on the hard
disk, there is no problem in reformatting few MOs and recopy the data.
However, there are people out there that would certainly benefit from this,
perhaps because it is a lot easier to convert, let�s say 35 MO diskettes than
reformat them and recopy the data, or because MOs are the only place where
some important data reside and reformatting under new kernel would be
disasterous.
If you can, and can get the time, do it. Many will be thankful.
Cheers
Igor
------------------------------
From: Wout van Albada <[EMAIL PROTECTED]>
Subject: BUG? fcntl problem in kernel 2.2.4/NFS/mutt
Date: Fri, 02 Apr 1999 10:52:26 +0200
Hi,
I encountered a problem with fcntl. When using mutt, it complained that
it could not obtain a write lock on a mailbox file. The mailbox was on
an NFS mount.
Everything had been working fine until I upgraded to kernel 2.2.4
(had kernel 2.0.36 before). Yes, I also upgraded the 'kernel related'
packages.
The error message I get came from line 170 in mx.c in the mutt source
code.
This was triggered because the call to fcntl earlier resulted in ENOLCK.
..........
while (fcntl (fd, F_SETLK, &lck) == -1) /* returns -1, errno =
ENOLCK */
{
struct stat sb;
dprint(1,(debugfile, "mx_lock_file(): fcntl errno %d.\n", errno));
if (errno != EAGAIN && errno != EACCES)
{
mutt_perror ("fcntl"); /* <= line 170 */
return (-1);
}
..........
I have the following setup:
Intel style platform.
Redhat Linux 5.2 running kernel 2.2.4.
Mutt version 0.95.4i.
NFS server running Solaris (SunOS 5.6) for Intel.
Oh yes, I recompiled mutt with the `--disable-fcntl' flag which works
fine.
But that doesn't really fix the problem.
Anyone have any ideas?
If you need any more details, feel free to send an email.
Wout
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: polling an interface at 12 KHz
Date: 2 Apr 1999 07:09:43 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Mark Hahn <[EMAIL PROTECTED]>]
> of course. it's basically trivial to do, if you busy-wait in a
> realtime thread. of course, the machine will not respond like a
> multitasking box during that time.
That's why I said Linux might not be very well suited for hardware it
has to poll at 12 kHz. Presumably the project calls for driving the
hardware for an arbitrary length of time. The issue looks a lot
different if you only need the 12 kHz during a relatively short
initialization process.
> it's somewhat harder, but still hardly a big deal, to turn HZ up to
> 12 KHz, and be a little kinder. the ideal solution is to do this in
> kernel space, since then you can make the performance impact
> near-zero.
Actually (according to my favorite source for this sort of information,
Kernel Traffic) there was a thread in linux-kernel recently that
touched on this. Apparently not all device drivers are really ready
for arbitrary HZ tuning yet, though that is definitely a general goal.
Some places in the kernel -- parts of the SCSI subsystem came up --
reportedly disable interrupts for long enough that you can start losing
timer ticks if your HZ is too fast.
In some cases, continues the thread, this unfortunate effect results
from having to drive broken hardware. The example that came up was a
joystick driver (MS Sidewinder, IIRC), which apparently needs a few
hundred ms of hard realtime just to initialize -- and to reinitialize,
whenever that is necessary.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (John Forkosh)
Subject: Re: Put another way: Linux from the Ground Up
Date: 2 Apr 1999 04:44:48 -0500
[EMAIL PROTECTED] wrote:
: I am simply curious to build a Linux system WITHOUT using a distribution.
You might take a look at Tom's root/boot
http://www.toms.net/rb/
which is a customizable one-floppy "distribution".
It's pretty easy to see how it works, and you might
be able to start from there.
John ([EMAIL PROTECTED])
P.S. There are several similar projects, but I can't
recall them offhand. I think Tom's page links to some
of them, which you might also want to check out.
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Kernel 2.3 when?
Date: 2 Apr 1999 07:30:49 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Christopher B. Browne <[EMAIL PROTECTED]>]
> It seems likely to me that there is more likely to be an "ext3fs"
> than there is an "augmented ext2fs."
Yep. Stephen Tweedie is working on things like journalling extensions
to ext2 but has already renamed the work "ext3". Linus will *not*
allow ext2 to destabilize in place.
> The other "killer feature" that I hope for is for EGCS to stabilize
> (e.g. - to become the new GCC), and for C9X functionality to
> thereupon provide us "portable" 64 bit values.
Stable egcs would be wonderful (it's already plenty stable enough for
my needs but those aren't very rigorous), and whatever contributions
from the PCG they manage to merge in and QA would only be to the good.
Portable 64-bit values would be aesthetically nice, but...
> 64 bit values buys us:
> - Bigger-than-2GB files ("ext64fs," maybehaps?)
> - Big dates (--> no more 2038 problem)
> - Univerally-available 64 bit ints
For kernel work, "portable types" have never been particularly
important. If you really need a definite word size you use __s32 or
__u64 or whatever. The fact that those are not standard does cause a
few #ifdefs in header files but hasn't really deterred anyone yet. Any
ugliness or unportability in kernel/user interfaces will mostly hide
behind libc.
Having portable 64-bit ints *would* be nice for user-space development,
though typedeffing `long long' in an #ifdef works for now.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: Vincent <[EMAIL PROTECTED]>
Crossposted-To:
alt.os.linux,alt.linux,comp.os.linux.x,comp.os.linux.hardware,comp.os.linux.development,comp.os.linux.development.apps
Subject: Limit ???
Date: Thu, 01 Apr 1999 16:07:45 +0200
How can I configure the limit of coredump for standard user ?
When I'm root it's not a problem. I tried to set the value in the
/etc/profile file but nothing result.
And for the /etc/login.* it's the same.
In what file are store this information ???
[EMAIL PROTECTED]
------------------------------
From: John Burton <[EMAIL PROTECTED]>
Subject: Re: Linux Kernel documentationI
Date: Fri, 02 Apr 1999 14:38:25 +0100
I've got the book "Linux device drivers". You can find out abaout it at
:
http://www.amazon.co.uk/exec/obidos/ASIN/1565922921/n/qid=923060259/sr=2-3/026-9763839-6482227
which only covers what you need to know for writting device drivers but
it is a very good book indeed.
root wrote:
>
> I'm a novice about Linux but i'd found it wonderful.
> I'm an Andersen Consulting's consultant and when i can work on linux I'm
> very happy.
> I wrote to this group to ask you about a book (downloadable or not)
> which can teach me more about
> the inner workings of the kernel away from the university books (a lot
> of theory and a few of practice)
> Thanks in advance
>
> write to [EMAIL PROTECTED]
------------------------------
From: Harry Heinisch <[EMAIL PROTECTED]>
Subject: Re: 4 Gb memory?
Date: Thu, 01 Apr 1999 22:11:32 -0500
All,
Although the limit for all platforms (AFAIK) is 2GB memory, several
individuals are proceeding to move the limit to 8 or 16 (under Alpha
mainly, since newer systems from Compaq break the barrier). Linus will be
getting a system with more than 2GB, so I presume he will join in the work
to get larger memory systems working as well. Just wait until the newest
1GB DIMMs come out on the market - VERY large memory!
Harry
Compaq
------------------------------
From: root <[EMAIL PROTECTED]>
Subject: Linux Kernel documentationI
Date: Wed, 31 Mar 1999 22:17:08 +0200
I'm a novice about Linux but i'd found it wonderful.
I'm an Andersen Consulting's consultant and when i can work on linux I'm
very happy.
I wrote to this group to ask you about a book (downloadable or not)
which can teach me more about
the inner workings of the kernel away from the university books (a lot
of theory and a few of practice)
Thanks in advance
write to [EMAIL PROTECTED]
------------------------------
From: Johan Kullstam <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.misc,linux.redhat.misc,alt.linux,alt.os.linux,comp.os.linux.hardware
Subject: Re: Idea: Make a seperate "i686" tree for Redhat Linux 6.0
Date: 02 Apr 1999 08:14:51 -0500
Enkidu <[EMAIL PROTECTED]> writes:
> It's a mixed blessing. Count the number of times there are questions
> on this group from someone who has bought or downloaded Redhat, and
> doesn't know how to partition a disk.
is this a redhat problem or a generic linux problem? if debian,
slackware &c were as popular i am sure we'd hear the same questions
about them.
> Or even that they can't run Linux under Windows! A recent question
> posted was "Where's the setup.exe for Linux".
^^^^^^
notice this says linux and not redhat.
so they are clueless. we all were at one point.
> However if these people *can* learn, then they become an asset to
> the Linux community, and to the non-Microsoft world.
yes.
> Redhat does put *barriers* to understanding, by making things look
> more GUI, and hiding the nuts and bolts. Again this is both good
> and bad.
whatever.
--
johan kullstam
------------------------------
From: Roman <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,fido.belg.linux,it.comp.linux.development
Subject: Re: Compiling???
Date: Thu, 01 Apr 1999 19:35:52 +0200
Ivan Van den Bossche wrote:
>
> Hello everybody !!!
>
> I have written a simple program which displays 'hello, world' on the screen:
>
> The source file calles 'test.cpp'
>
> #include <iostream.h>
>
> int main()
>
> {
>
> cout << "Hello, world !";
>
> return 0;
>
> }
>
> I have some questions: How can I compile the test.cpp file, so I get an
> executable file ???
>
> I'm using Redhat Linux 5.0.
>
> I've already used gcc and g++ but I don't know what library I have to link
> with it.
>
> I'm also interested in programming modules and X Window - applications.
>
> How can I compile a source file as a X Window-applications or as a module
> written with C/C++ ???
>
> Where can I find source-code on the web for these areas???
>
> my email:
>
> [EMAIL PROTECTED]
>
> Thanks !!!
Look for XWPE
it's just like borland C for dos
------------------------------
** 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
******************************
