Linux-Development-Sys Digest #369, Volume #6 Wed, 3 Feb 99 05:14:06 EST
Contents:
problems with netscape on a glibc system... (Ronald Cole)
2.2.1 and modules -- modules.dep is empty (Joe Pfeiffer)
Re: A question about Intel... (Peter Samuelson)
Re: Linux Phase 2: A Consumer Operating System (Christopher Browne)
Re: Easy(?) kernel question. (Alexander Viro)
Re: egcs 2.91.60 and the 2.2.1 kernel. (Peter Samuelson)
Re: Diskless systems? (Martin Maney)
Re: Newbe tar question (Timothy J. Lee)
Re: Linux apps in win2000 port news! (Colin Smith)
Re: glibc-crypt-2.0.112 somewhere ? (Andreas Jaeger)
Re: I'm confused. (Juergen Heinzl)
Re: Easy(?) kernel question. (Michel TALON)
Re: Linux Phase 2: A Consumer Operating System (Jon Taylor)
----------------------------------------------------------------------------
From: Ronald Cole <[EMAIL PROTECTED]>
Subject: problems with netscape on a glibc system...
Date: 02 Feb 1999 18:16:16 -0800
The problem is with libXpm.so.4.11. Netscape requires one built
against libc-5, so I put one in /usr/local/lib/netscape and performed
the following gyrations:
$ ls -l netscape*
-rwxr-xr-x 1 root root 85 Feb 2 17:45 netscape*
lrwxrwxrwx 1 root root 32 Sep 4 13:46 netscape-4.5 ->
/usr/local/lib/netscape/netscape*
$ cat netscape
LD_LIBRARY_PATH=/usr/local/lib/netscape
export LD_LIBRARY_PATH
exec netscape-4.5 $*
For the most part, this works. However, I have a glibc compiled
libXpm.so.4.11 in /usr/X11R6/lib and I recently compiled freeamp-1.1.0
against it. Calling freeamp from within Netscape nets me the
following error:
/usr/local/lib/netscape/libXpm.so.4: undefined symbol: _fxstat
Clearly, I don't want to set LD_LIBRARY_PATH when calling Netscape
because all helper apps will attempt to link in the libc-5 libXpm.
So, how do I get Netscape to link against a libc-5 libXpm while all
helper apps called by Netscape link against a glibc libXpm?
--
Forte International, P.O. Box 1412, Ridgecrest, CA 93556-1412
Ronald Cole <[EMAIL PROTECTED]> Phone: (760) 499-9142
President, CEO Fax: (760) 499-9152
My PGP fingerprint: 15 6E C7 91 5F AF 17 C4 24 93 CB 6B EB 38 B5 E5
------------------------------
From: Joe Pfeiffer <[EMAIL PROTECTED]>
Subject: 2.2.1 and modules -- modules.dep is empty
Date: 02 Feb 1999 20:38:12 -0700
I haven't been able to get modules to work with 2.2.1. I have
modutils 2.1.121; I compiled and brought up a version of 2.2.1
without modules, and built modutils with that running.
But, at boot time, depmod -a is leaving /lib/modules/2.2.1/modules.dep
empty. As you can imagine, everything is down hill after that! I
could use any suggestions anybody might have as to where to look for
my error...
--
Joseph J. Pfeiffer, Jr., Ph.D. Phone -- (505) 646-1605
Department of Computer Science FAX -- (505) 646-1002
New Mexico State University http://www.cs.nmsu.edu/~pfeiffer
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: A question about Intel...
Date: 2 Feb 1999 18:35:05 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Indigo <[EMAIL PROTECTED]>]
> I've heard that Intel high-end processors, like PentiumII and
> PentiumPro, work as a sort of RISC processor that emulate an 80x86
> processor.
Yes.
> Can I directly work with this set of instructions? (= Can I work with
> a PentiumII or an AMD K6 treating them as RISC processors?)
No.
And even if you could, it wouldn't be the same between Intel and
AMD. And maybe not between the PPro and P2.
> Could be this change in the instruction set a way to increase
> performances?
Not sure I understand your question. Intel definitely did it this way
because of performance, yes. If you *could* access the microcode
directly it might increase performance (though, who can say since the
translation is done in parallel anyway) but you can't. Note that if
Intel had built in some sort of logic to switch modes to the other
instruction set it would make the chip as a whole *more* complex,
*less* RISC-y....
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Christopher Browne)
Crossposted-To: alt.os.linux,comp.os.linux.development.apps,comp.os.linux.setup
Subject: Re: Linux Phase 2: A Consumer Operating System
Reply-To: [EMAIL PROTECTED]
Date: Wed, 03 Feb 1999 04:09:12 GMT
On Tue, 02 Feb 1999 21:15:09 GMT, steve mcadams <[EMAIL PROTECTED]> wrote:
>[Snipped for brevity, quoted material marked with ">"]
>On Tue, 02 Feb 1999 01:23:49 GMT, [EMAIL PROTECTED] (Christopher
>Browne) wrote:
>
>>If it takes more than "bolting it on," then the OS starts to suffer.
>
>There's no reason why it should. I'm currently (off and on) working
>on a GUI component that will run on both Win32 and Linux with
>identical application source code. If it's possible for GUI
>components to bridge that gap, I'm sure it would be easy enough to
>make the jump to the belowmentioned forks.
>
>>- Some of the features of Hurd involve breaking off things that Linux
>>does. It adds the ability to mount $PATH as /bin, which makes the
>>various binary paths go away. Similarly, /usr goes away.
>>
>>- Xos involves throwing away virtually the whole kernel, and
>>reimplementing kernel functionality as sharable libraries. Applications
>>may still be usable, supposing a reasonably compatible API, but the base
>>kernel goes away *entirely.*
>>
>>- FluxOS lies somewhere in between. It pushes a lot of kernel
>>functionality out to libraries, and exposes additional pieces of system
>>management. Again, while portably written applications may not notice a
>>change, the kernel gets changed quite completely.
>
>I guess these guys had reasons for their forks. I don't see why a
>single GUI layer couldn't ride on top of all of them who wanted to use
>it.
They're not "forks." FluxOS has a scheme where they use something
remarkably like COM to connect FreeBSD and/or Linux device drivers, but
is not otherwise based on Linux.
They're *COMPLETELY NEW CODE.*
The point of their efforts was generally speaking to support different
sorts of research efforts to try to come up with fundamentally new
approaches to building OSes.
They're not evolutionary; they're revolutionary.
And they can't coexist with the Linux 2.3 effort (that likely starts
Real Soon Now) because they're not building a Linux-like kernel.
--
"What's wrong with 3rd party tools? Especially if they are free? What
the **** do you think UNIX is anyway? It's a big honkin' party of 3rd
party free tools." -- Bob Cassidy ([EMAIL PROTECTED])
[EMAIL PROTECTED] <http://www.hex.net/~cbbrowne/lsf.html>
------------------------------
From: [EMAIL PROTECTED] (Alexander Viro)
Crossposted-To: comp.os.linux.misc
Subject: Re: Easy(?) kernel question.
Date: 2 Feb 1999 23:43:23 -0500
In article <[EMAIL PROTECTED]>,
cano_jonathan <[EMAIL PROTECTED]> wrote:
>I can see how it might be considered a joke but no, I wasn't joking.
>
>I imagine that the question is quite easy for the right sort of
>person: someone who knows the x86 instruction set and architecture
>well (e.g. someone who has had to write assembler instructions for
>kernel scheduler)
It's not a scheduler question. On ix86 you don't need to do anything
special to flush TLB on a context switch. Any assignment to cr3 (address of
root pagetable) will do the trick. Look in include/asm-*/pgtable.h for
flush_tlb() and notice that on ix86 it will happen automatically as a side
effect of MMU context switch.
--
"You're one of those condescending Unix computer users!"
"Here's a nickel, kid. Get yourself a better computer" - Dilbert.
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: comp.os.linux.advocacy
Subject: Re: egcs 2.91.60 and the 2.2.1 kernel.
Date: 2 Feb 1999 19:31:03 -0600
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[mlw <[EMAIL PROTECTED]>]
> I know what I am doing is "bad" but, I compiled the 2.2.1 kernel with
> egcs. It seems to work, even with the XFree86 Mach64 server.
You misunderstand. It is "bad" to compile a 2.0.x kernel with egcs.
The 2.1.x/2.2.x series do not (for the most part) have that problem. I
have been using 2.1.x and egcs on my machines for several months now.
> Anyone have any experience with this? Anyone know any hard core facts
> why this is a bad idea?
Here's the story: embedding assembly language in a C program is pretty
easy if you use GCC, and you can share variables back and forth using a
"constraints" syntax to tell GCC what registers should be loaded and
stored to/from variables before and after the asm, which registers are
clobbered by your asm, stuff like that. Kernel hackers have introduced
bugs in the past relating to asm constraints, and didn't notice because
GCC 2.7.2 did not optimize code heavily enough to break the assumptions
they forgot to put in.
When GCC 2.8 was released it contained some additional optimization
strategies, some of which can break kernel code by (for example)
mistakenly optimizing out some essential asm. EGCS is derived from GCC
2.8 so it does this too. Kernel hackers started to notice the kernel
bugs and fix them, and thus 2.1.x has been made progressively safer and
safer to compile with the more agressive compilers. Some constraint
funnies are still allegedly known to exist in current kernels, but none
serious enough to break under current compilers.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: Martin Maney <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development,comp.os.linux.redhat
Subject: Re: Diskless systems?
Date: 3 Feb 1999 01:16:37 GMT
Matt Kressel <[EMAIL PROTECTED]> wrote:
> The Diskless system HOWTO:
> http://metalab.unc.edu/LDP/HOWTO/mini/Diskless.html
> No data bout RAM reqs. though, sorry.
I don't have the URL handy (hmmmm, wait a bit, maybe I can find it), but the
smallest web server yet displayed in public is a Linux machine runnign on a
very compact 486 board with, IIRC, 16M of RAM and 16M of flash. The only
cheesy part was using PLIP for the uplink. :-)
Now let me see... nope, sorry, that must have been in the bookmarks on the
lab machine that I keep meaning to copy and merge with the local set. :-(
The teeny server machine was mentioned in slashdot sometime last week, I
believe. I expect it could also be found in Linux Weekly News, and probably
on other like-minded news sites as well.
[followups reduced]
------------------------------
From: [EMAIL PROTECTED] (Timothy J. Lee)
Subject: Re: Newbe tar question
Reply-To: see-signature-for-email-address---junk-not-welcome
Date: Tue, 2 Feb 1999 19:52:41 GMT
[EMAIL PROTECTED] (Bucky4me) writes:
|I am new to UNIX and Linux. I uncompressed a .gz file (JDK) and now have a
|.tar file. How do I use it. It is my understanding that this is a archive
|file. Is this correct? I am trying to install Java on my Linux system.
"tar tvf file.tar" tells you what is in the tar file
"tar xvf file.tar" extracts the files in the tar file
See "man tar".
--
========================================================================
Timothy J. Lee timlee@
Unsolicited bulk or commercial email is not welcome. netcom.com
No warranty of any kind is provided with this message.
------------------------------
From: Colin Smith <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.os.linux.hardware,comp.os.linux.misc,comp.os.linux.advocacy
Subject: Re: Linux apps in win2000 port news!
Date: Tue, 02 Feb 1999 20:24:48 +0000
M Sweger wrote:
> Hi all,
>
> Just saw on the biz wire that a company http://www.interix.com has
> developed a 64 bit application that runs a a unix/linux emulator within
> the MSoft Win2000 (NT5.0) platform environment. They want alot of
> Linux apps ported to their API interface so that Unix runs within Windows.
> Seems kind of ironic since Linux is trying to run Windows apps in Unix;
> but we don't have the source code to their stuff, but they have ours.
> Hmmm! seems like there isn't any competition again. Msoft is supporting them
> though.
>
Hmmm....... Yes lets spend lots of money to run our Linux applications inside an
emulator on Windows NT rather than on Linux...........
Doh.....
Why bother?
>
> They have a limited list of things they want ported such as gcc (all compilers and
>assemblers) and emacs and Tex to name a few. You have 2.5 months to do it
> by -- they end of January 1999. Anything else is nice but they won't pay for.
> Those they'd like you to port will only pay a maximum of $1000 to a cheap
> payment of $500. They didn't mention [conveniently] that they would like the
> linux kernel or Sunos x86 kernel etc to run within the interix Win2000.
> The reason it was supposed to be done by the end of January 1999 (now past)
> is due to todays press release about this product.
>
> Hmmm! if you can't get the Unix software vendors to port their stuf from Unix
> to Window Nt, then make the Unix apps run within NT! In this way you can
> say that your OS is a engineering workstation.
>
> --
> Mike,
> [EMAIL PROTECTED]
--
|Colin Smith: [EMAIL PROTECTED] | Windows 2000: |
|Linux: Delivers on the promises Microsoft make. | The Zeppelin of |
| http://www.linux.org/ | operating systems.|
------------------------------
From: Andreas Jaeger <[EMAIL PROTECTED]>
Subject: Re: glibc-crypt-2.0.112 somewhere ?
Date: 02 Feb 1999 18:30:37 +0100
>>>>> Mike Dowling writes:
Mike> I would like to hear about these test releases. Specifically, I
Mike> need to update from glibc-2.0.6 as, for some reason, the
Mike> compiler pgcc-1.1.1 wants to link some function from libm that
Mike> is not present in libm-2.0.6. My temporary fix is to copy
Mike> libm-2.0.7 from a red hat distribution.
Which function? What's the error message? Did you grab a binary of
pgcc?
Mike> I've down loaded
Mike> glibc-crypt-2.0.111.tar.gz
Mike> glibc-2.0.111.tar.gz
Mike> glibc-linuxthreads-2.0.111.tar.gz
Mike> glibc-localedata-2.0.7pre3.tar.gz
Mike> I could not find a 2.0.111 version for localedata; hopefully it
Mike> will live happily together with the other libs.
The localedata add-on is now integrated into glibc, don't use the
2.0.7 version!
Mike> Is there anything I have to watch out for with this release?
Read the FAQ and the INSTALL file first.
Mike> I'm a little uneasy about this as my Linux distribution is a
Mike> DIY, and if I screw up with libc, it will cost me time that I
Mike> don't have at the moment.
glibc 2.0.111 shouldn't have problems - but libc is one of the most
vital parts of your system.
Andreas
--
Andreas Jaeger [EMAIL PROTECTED] [EMAIL PROTECTED]
for pgp-key finger [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Subject: Re: I'm confused.
Date: Tue, 02 Feb 1999 20:49:43 GMT
In article <796kba$[EMAIL PROTECTED]>, Dr. Unk wrote:
>Ok, I may be ignorant, so don't bash my head in for that. I have a couple
>of topics I need clarified.
>
>1. Is egcs-1.1.1 and all it's other addons stable enough for production?
>If not which gcc should I get?
Some question ... depends on the platform too. On an AIX I could produce
pretty interesting assembler code with the combination egcs-1.1.1 and
GNU binutils, although the CPU seemed somewhat surprised about what it
saw. Using the IBM linker still causes trouble that has to be sorted out
... waiting for the original compiler package (the reason I tried egcs,
time scales).
On Intel ... I've compiled both, 2.0.7pre6 as well as 2.0.112 with egcs;
no noticeable bugs yet, which does not mean there are none; they just
have not showed up yet (as with any software more complex than hello world).
For gcc ... 2.7.2.3 is stable and at the same time too buggy to compile
the glibc (funny, isn't it), 2.8.0 and 2.8.1 is "use on your own risk",
though I did not use it on a non Intel platform and on others it might
be better. Just not the time to play around with that one too; we need
a production C++ compiler.
>2. I'm not too clear on the 3 glibc versions I've seen, which one is
>stable (2.0.6, 2.0.7pre6, 2.0.112)? What are the differences?
I jumped from 2.0.7pre6 to 2.0.112; the bad news first ... the libstdc++
from egcs compiled with egcs against 2.0.7pre6 and kernel 2.0.36 is a no
go anymore and both, the library as well as, for instance, ddd had to
be re-compiled due to the unchanged version number of the shared lib. Up
to now though all C binaries compiled against 2.0.x are fine (still some
caveats, see the FAQ).
2.0.112 is a test release though, so all right, I'd to expect things like
that and I want to test it (and installed it after a complete system
backup first too) while 2.0.7pre6 is a release version, considered to
be reliable. Does not mean 2.0.112 breaks your machine and bootstrapping
egcs while in X, downloading news and having Netscape running did neither
cause a hick nor an up.
In any case, the 2.0.112 is much more standards aware, something very
important for me and in combination with 2.2.1 ... it screams. I expected
a slowdown to be honest ... really a *great* job girls and guys 8-)
Lots of ISO C 9x, POSIX, Unix98 stuff and IPv6 support that makes life
easier.
Up to now I dare say, if you are not in urgent need of having a glibc
system do not care about it and wait for 2.1. Versioning support is in
2.0.112 already at least, so even if the internal interfaces change
programmes will still run. If you use a distribution though they are left
holding the baby anyway. You might look at it this way ...
2.0.36 (or 37) + libc5 (play safe), 2.2.x + libc5 (pretty safe),
2.2.x + 2.0.7x (pretty safe too), 2.2.x + 2.0.112 (not if you need the
machine 24h a day, 7 days a week, are the only administrator and off on
a North Pole expedition).
Minor note, the libstdc++ that Netscape requires runs out of the box with
2.0.112 too, so *big phew* ... you still can use it to visit /.
>3. I need a minimal-list of the packages to make a distribution free
>system. I currently have the list for the GNU version 0.2 system. Is there
>a place that I can get samples of all the configuration files, do they come
>with the packages they use?
None I know of, altough with some thinking it over it would be possible to
add up the stuff required. It's not that much and if in doubt ... remove
what you think is not necessary and if it still comes up ...
Cheers,
Juergen
--
\ Real name : J�rgen Heinzl \ no flames /
\ EMail Private : [EMAIL PROTECTED] \ send money instead /
\ Phone Private : +44 181-332 0750 \ /
------------------------------
From: Michel TALON <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc
Subject: Re: Easy(?) kernel question.
Date: 03 Feb 1999 10:19:41 +0100
You can get full documentation on Pentium processors by going to:
http://www.x86.org/intel.doc/686Manuals.html
and downloading
Pentium(R) Pro Family Developer's Manual, Volume 3
------------------------------
From: [EMAIL PROTECTED] (Jon Taylor)
Crossposted-To: alt.os.linux,comp.os.linux.development.apps,comp.os.linux.setup
Subject: Re: Linux Phase 2: A Consumer Operating System
Date: 3 Feb 1999 09:20:09 GMT
Reply-To: [EMAIL PROTECTED]
On Wed, 03 Feb 1999 04:09:12 GMT, Christopher Browne <[EMAIL PROTECTED]> wrote:
>On Tue, 02 Feb 1999 21:15:09 GMT, steve mcadams <[EMAIL PROTECTED]> wrote:
>>[Snipped for brevity, quoted material marked with ">"]
>>On Tue, 02 Feb 1999 01:23:49 GMT, [EMAIL PROTECTED] (Christopher
>>Browne) wrote:
>>
>>>- FluxOS lies somewhere in between. It pushes a lot of kernel
>>>functionality out to libraries, and exposes additional pieces of system
>>>management. Again, while portably written applications may not notice a
>>>change, the kernel gets changed quite completely.
>>
>>I guess these guys had reasons for their forks. I don't see why a
>>single GUI layer couldn't ride on top of all of them who wanted to use
>>it.
>
>They're not "forks." FluxOS has a scheme where they use something
>remarkably like COM to connect FreeBSD and/or Linux device drivers, but
>is not otherwise based on Linux.
>
>They're *COMPLETELY NEW CODE.*
>
>The point of their efforts was generally speaking to support different
>sorts of research efforts to try to come up with fundamentally new
>approaches to building OSes.
>
>They're not evolutionary; they're revolutionary.
>
>And they can't coexist with the Linux 2.3 effort (that likely starts
>Real Soon Now) because they're not building a Linux-like kernel.
Flux is really interesting. It is based on the concept of
polymorphic, recursize virtual machines. A microkernel exports process
spaces which spawn new subprocess spaces. Page faults are used for
kernel call gates. Another team has taken Flux and merged it with Kaffe
and added the concept of a process to Java, and all of the 'sandbox'
concerns disappear from Java. It is another cool idea. I think the
OSKit is going to generate a wave of OS innovation.
Jon
---
'Cloning and the reprogramming of DNA is the first serious step in
becoming one with God.'
- Scientist G. Richard Seed
------------------------------
** 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
******************************
