Linux-Development-Sys Digest #712, Volume #6 Fri, 14 May 99 18:14:32 EDT
Contents:
Hilfe wie kriege ich meine Matrox-G100 AGP einegebunden ? ("J�rg Gro�er")
Rh6.0/LinuxThreads debugging? ("Phil Estes")
Re: Get client machine's IP-address (Iond Research Srl)
Accurate profiling with Pentium TSC (Andre McCurdy)
accessing real-mode ram on pc, how ? (Lars Frenzel)
Re: more than 65000 ids impossible ? (Phil Howard)
Re: get client machine's IP-address (Barry Margolin)
PhD studentship available in component based OSes & database systems architecture
(Greg Law)
Re: libc5/libc6 ([EMAIL PROTECTED])
Re: Translation of linux to minor languages ("Stefan Monnier "
<[EMAIL PROTECTED]>)
Re: A most peculiar bug! ("Stefan Monnier "
<[EMAIL PROTECTED]>)
Re: Hostile Takeover of Linux ("Tom Emerson")
Re: Accurate profiling with Pentium TSC ("Stefan Monnier "
<[EMAIL PROTECTED]>)
Re: more than 65000 ids impossible ? (Joel Klecker)
Re: TCP/IP guru help needed (Modemch)
Re: get client machine's IP-address (Lew Pitcher)
Re: Translation of linux to minor languages (Johan Kullstam)
Re: 2.2.8 - Evil behavior (Johan Kullstam)
portmap problems (Ulrich Eckhardt)
Repost: Help needed with Mobitex Driver for linux (Fong Fatt Chee)
Re: Hostile Takeover of Linux (Marco Anglesio)
----------------------------------------------------------------------------
From: "J�rg Gro�er" <[EMAIL PROTECTED]>
Subject: Hilfe wie kriege ich meine Matrox-G100 AGP einegebunden ?
Date: Fri, 14 May 1999 18:20:48 +0200
Ich habe ein Problem, ich besitze Red Hat 5.2 Linux und m�chte meine
Grafikkarte MGA-100G einbinden leider habe ich kein Linuxtreiber
gefunden ich hoffe auf HILFE.
besten dank schon mal
------------------------------
From: "Phil Estes" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.development.apps
Subject: Rh6.0/LinuxThreads debugging?
Date: Fri, 14 May 1999 10:53:53 -0500
I had my RedHat 5.2 with (1) kernel 2.2.3 and (2) glibc
2.1.1pre1 (3) with LinuxThreads (033199 drop) and
(4) H. J. Lu's gdb 4.17.0.11 patches. This setup allowed
me to debug multithreaded applications very well with
no problems.
Now I have RedHat 6.0 which has in it (1) kernel 2.2.5
(2) glibc 2.1.1 ? (later drop than I had it seems?), (3)
LinuxThreads 04??99 drop (later than I had) and (4)
gdb 4.17.0.11. Debugging a multithreaded app hangs
many times. All threads are either in sigsuspend (as they
should be) or pthread_handle_sigrestart code, but nothing
ever advances at this point. These programs run fine (just
as they did in my 5.2 configuration) outside the debugger,
but under the debugger they no longer are debuggable
because I hit this shortly after all threads are created (early
on in the program). I have rebuilt gdb 4.17 with H.J. Lu's
patches just to make sure gdb had everything it needs, but
still no luck.
Any ideas? Should I try and be more like my 5.2 system
(build glibc myself with LinuxThreads add-on?) and see if
that fixes it or is there some other problem with this
setup.
Thanks,
Phil
--
======================================
Phil Estes / [EMAIL PROTECTED]
IBM Austin - replace _NO_SP_ with austin to reply...
------------------------------
From: Iond Research Srl <[EMAIL PROTECTED]>
Subject: Re: Get client machine's IP-address
Date: Fri, 14 May 1999 18:09:21 +0200
Mark Hahn wrote:
> getpeername. this would be obvious if you had read even basic
> network programming books.
Thanks Mark for your answer, but I'm not sure getpeername can solve my
problem.
I'll try to explain myself better with an example:
You are connected from a linux machine (client) to a solaris machine
(server) with
a telnet session during which you want to execute this c-program
(complex.c)
main()
{
printf("Hi, Mark\n");
/* Now I wish to know the IP-address of the machine on which
the string "Hi, Mark" is displayed"
*/
}
This program runs in the server but thru the pseudo-tty its output is
displayed
at the client machine.
No sockets are used so we haven't the first arg to pass to the
getpeername.
You can try to pass 0 as socket-fd to getpeername but it returns the
local-host
IP-address (127.0.0.1).
So what's the matter ?
Maybe work around the slave part of the pseudo-tty ?
Sorry for this silly question but I'm not a network programmer, really
this is
my first networking problem, and also reading Stevens' I haven't yet
found a
solution.
Anyway, thanks a lot
Lucio
[EMAIL PROTECTED]
------------------------------
From: Andre McCurdy <[EMAIL PROTECTED]>
Subject: Accurate profiling with Pentium TSC
Date: Fri, 14 May 1999 15:51:37 GMT
I'm trying to optimize some code by timing its execution using the Time
Stamp Counter on a Pentium.
What seems to be happening though is that the average executon time
changes by a few percent between runs - even without any changes to the
source. I remember reading somewhere (possibly this group ??) about
paging/caching effects which can produce this type of behaviour, but
don't remember the workaround.
Can anyone suggest a solution ?
Andre
--
------------------------------
From: Lars Frenzel <[EMAIL PROTECTED]>
Subject: accessing real-mode ram on pc, how ?
Date: Fri, 14 May 1999 16:04:09 +0200
Hi,
how can I access the 'real-mode' ram (specified by seg:offs) on a pc ?
(using c/c++)
i tried (but it didnt work ;-( the following:
char *ptr = (char *)(seg<<4)|(offs&0xF);
ptr[0] = ...
--or--
fd = open("/dev/mem");
ptr = mmap(...);
ptr[0] = ...
in fact i simply have to access the video-memory directly due to
speed improvement
thx,
Lars.
------------------------------
From: [EMAIL PROTECTED] (Phil Howard)
Subject: Re: more than 65000 ids impossible ?
Date: Fri, 14 May 1999 17:15:12 GMT
On 14 May 1999 10:03:09 GMT Tibor Weis ([EMAIL PROTECTED]) wrote:
| I would like to ask you if linux can support more than 65000 user IDs.
| We would like to set up main MAIL HUB for more about 100 000
| users. I found out that SOLARIS 7 support it but
| LINUX does not. Do you have any idea how to solve it ?
If the mail software disassociates mail users from system users, then
you are basically free of limitations. Incoming mail addressed to,
and POP/IMAP mail pickups authenticated as for, any user@host name
(where the @ might be translated to some other character for the user
login purposes), gives you freedom to not be bound to the system user
names. Choose your mail software wisely.
--
Phil Howard KA9WGN
[EMAIL PROTECTED] [EMAIL PROTECTED]
------------------------------
From: Barry Margolin <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.apps,comp.programming,comp.unix.programmer,comp.os.linux.networking
Subject: Re: get client machine's IP-address
Date: Fri, 14 May 1999 18:24:20 GMT
In article <[EMAIL PROTECTED]>,
Lew Pitcher <[EMAIL PROTECTED]> wrote:
>I'm in a telnet session to my unix host (Linux, if it makes a difference),
>and I enter the "who am i" command. My unix host replies with
> srdscs05!pitchl ttyp1 May 14 13:03 (49.21.20.105)
>and I note that the IP address in parenthesis is the IP address of my
>Windows NT client machine (from which I telnet to unix).
>
>OK, how does 'who' determine that IP address? Could that technique
>(what ever it is) be used to answer this thread's original question?
"who" gets it from /etc/utmp, which is where login stores it, based on the
-h parameter passed by telnetd.
However, if reverse DNS is configured for the client address, this will
contain the hostname, not the IP address. And on most systems the field is
only 16 characters, so if the hostname is longer than that it will be
truncated.
--
Barry Margolin, [EMAIL PROTECTED]
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
------------------------------
From: Greg Law <[EMAIL PROTECTED]>
Subject: PhD studentship available in component based OSes & database systems
architecture
Date: Wed, 12 May 1999 13:01:15 +0100
Apologies if this is repost - something strange happened on previous
attempt!
Applicants are sought for a Ph.D. studentship researching either
component based operating systems or database systems architecture at
City University, London.
Successful candidates will become a member of the Hipex (High
Performance Extensible Systems) group:
http://web.soi.city.ac.uk/research/hipex/
See http://web.cs.city.ac.uk/informatics/stuship.html for details of the
studentship.
if you're interested, email [EMAIL PROTECTED]
Greg.
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: libc5/libc6
Reply-To: [EMAIL PROTECTED]
Date: Wed, 12 May 1999 14:23:32 GMT
[EMAIL PROTECTED] wrote:
> Could anybody explain why libc6 has a much larger size
> than libc5? Thanks.
Because it provides a richer set of functionality ?
------------------------------
From: "Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]>
Subject: Re: Translation of linux to minor languages
Date: 14 May 1999 15:07:36 -0400
>>>>> "Johan" == Johan Kullstam <[EMAIL PROTECTED]> writes:
> both ways of skinning the cat are valid, it's just that one way has
> proven itself to be better in practice.
>From the programming side, that might be true (it seems to depend on how
complex your regexps can be), but there is also the performance issue (I
usually measure performance in MB rather than in seconds and converting
UTF-8 files to UCS-2 or UCS-4 before running the regexp engine seems costly
(prevents the use of memory mapping, for instance)).
Stefan
------------------------------
From: "Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]>
Subject: Re: A most peculiar bug!
Date: 14 May 1999 14:42:41 -0400
>>>>> "Frampton" == Frampton Steve R <[EMAIL PROTECTED]> writes:
> tar zcvpf /archive/backup-13-may-1999.tar.gz --exclude=archive
> --exclude=mnt --exclude=proc --exclude=var/spool/squid .
You don't want to exclude manually mnt, proc, ...
Instead you want to use the -l (aka --one-file-system) option.
Much much very cleaner and safer.
Stefan
------------------------------
From: "Tom Emerson" <[EMAIL PROTECTED]>
Subject: Re: Hostile Takeover of Linux
Date: Fri, 14 May 1999 12:19:31 -0700
Joe Pfeiffer wrote in message <[EMAIL PROTECTED]>...
>dan <[EMAIL PROTECTED]> writes:
>
>> Too bad guy's it was all for nothing.
>> Sylvan has tagged us all as idiots if we
>> don't get their Linux certification.
>
>I know I'm going to be sorry I asked this, but... what in the *hell*
>are you talking about?
I must be in a weird mood, but I *think* I understand where he's coming
from:
-- Basically, Sylvan is a piss-poor "certification" testing center, but
they have a virtual monopoly on the market.
-- "pointy-haired" types like to see "certificates", so they'll hire
someone who has passed a sylvan test over someone with "years of field
experience".
-- groups of "test cheaters" go into these certification center and do one
thing: memorize 5 of the 20 questions given and the answers provided; these
questions/answers are researched, and the group goes back secure in the
knowledge of the exact answer sylvan is looking for.
-- the questions themselves are rarely relevant -- one in particular I
remember "missing" on a netware questionaire had to do with how to load a
certain driver in a SPECIFIC type of memory [i.e., the "extended" vs.
"expanded" memory]. The reason this is an irrelevant question is because if
you type "/?" as the parameter, you're TOLD by the program which switch
setting is for which type of memory. However, 90% of the "techs" out there
probably don't know which "kind" of above-640k-memory scheme the computer is
using, nor do they care...
Taken together, you get a group of people who know the exact answers to test
questions that never come up "in the field", so when something "happens",
they'll resort to "well, let's reboot the machine and see if it happens
again". When a large enough group of people are resorting to
reboot-instead-of-research responses, the "word on the streets" will be that
linux is no better than windows "because you have to keep rebooting it to
make it work"...
------------------------------
From: "Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]>
Subject: Re: Accurate profiling with Pentium TSC
Date: 14 May 1999 13:10:32 -0400
>>>>> "Andre" == Andre McCurdy <[EMAIL PROTECTED]> writes:
> What seems to be happening though is that the average executon time
> changes by a few percent between runs - even without any changes to the
[...]
> Can anyone suggest a solution ?
Deal with it ?
Face it: execution time is *not* deterministic.
Typical workaround include things like run several times and take the minimum
of all run. Note that the program and the libraries it uses will probably not
be swapped back and forth between those runs, so they keep the same
virtual-to-physical mapping. This seems irrelevant but it can have a
significant impact on caching. To workaround this problem, you want to do
something like:
mkdir $TMP/bench-$$ && cd $TMP/bench-$$
cp $theprogram $thelibsIneedincludingldso .
LD_LIBRARY_PATH=. ./ld-linux.so.2 ./theprogram
This defeats the caching and forces a reload of all the libs and the
program so that the location in physical memory will be different for
each run, giving you a better chance to hit the minimum.
Another reasonable solution is to not care about the few percents
since after all they don't matter much in most cases.
Stefan
------------------------------
From: Joel Klecker <[EMAIL PROTECTED]>
Subject: Re: more than 65000 ids impossible ?
Date: Fri, 14 May 1999 10:16:49 -0700
In article <7hgsct$[EMAIL PROTECTED]>, Tibor Weis <[EMAIL PROTECTED]> wrote:
> I would like to ask you if linux can support more than 65000 user IDs.
> We would like to set up main MAIL HUB for more about 100 000
> users. I found out that SOLARIS 7 support it but
> LINUX does not. Do you have any idea how to solve it ?
There is this: <http://www-personal.engin.umich.edu/~wingc/uid32/>.
--
Joel Klecker (aka Espy) Debian GNU/Linux Developer
<URL:mailto:[EMAIL PROTECTED]> <URL:mailto:[EMAIL PROTECTED]>
<URL:http://web.espy.org/> <URL:http://www.debian.org/>
------------------------------
From: Modemch <[EMAIL PROTECTED]>
Crossposted-To: comp.protocols.tcp-ip
Subject: Re: TCP/IP guru help needed
Date: 14 May 1999 16:12:52 -0400
Dmytro Myasnykov <[EMAIL PROTECTED]> writes:
> Hi
> I see your problem. When window size is changing (grows up from 0 back to 288 ,
> for example), I send a window update to peer with new size. Normally it happends
> when program reads data from TCP/IP buffers. Also, standard implementation apply
> next condition: if place in receive window more or equal than 1 MSS AND not a lot
> of byte left unreaded by program -> force window update send. This window contains
> no data (you can send data if you have), only ACK, new sequence number and new
> window size.
> After this peer knows that you are able to receive more data.
> And I didn't understand what do you mean about x.1025 acks....If you still have
> problems, try to describe again and make a new tcpdump.
Thanks for responding, but I've fixed it already. I have simply fogrot to
send a window update after the recv() is complete.. Works like a clock
now.
--
Regards,
Modemch
------------------------------
From: [EMAIL PROTECTED] (Lew Pitcher)
Crossposted-To:
comp.os.linux.development.apps,comp.programming,comp.unix.programmer,comp.os.linux.networking
Subject: Re: get client machine's IP-address
Reply-To: [EMAIL PROTECTED]
Date: Fri, 14 May 1999 18:07:48 GMT
I'm in a telnet session to my unix host (Linux, if it makes a difference),
and I enter the "who am i" command. My unix host replies with
srdscs05!pitchl ttyp1 May 14 13:03 (49.21.20.105)
and I note that the IP address in parenthesis is the IP address of my
Windows NT client machine (from which I telnet to unix).
OK, how does 'who' determine that IP address? Could that technique
(what ever it is) be used to answer this thread's original question?
On Fri, 14 May 1999 18:13:49 +0200, Iond Research Srl <[EMAIL PROTECTED]> wrote:
>Mark Hahn wrote:
>
>> getpeername. this would be obvious if you had read even basic
>> network programming books.
>
>Thanks Mark for your answer, but I'm not sure getpeername can solve my
>problem.
>
>I'll try to explain myself better with an example:
>
>You are connected from a linux machine (client) to a solaris machine
>(server) with
>a telnet session during which you want to execute this c-program
>(complex.c)
>
>main()
>{
> printf("Hi, Mark\n");
>/* Now I wish to know the IP-address of the machine on which
> the string "Hi, Mark" is displayed"
> */
>}
>
>This program runs in the server but thru the pseudo-tty its output is
>displayed
>at the client machine.
>
>No sockets are used so we haven't the first arg to pass to the
>getpeername.
>
>You can try to pass 0 as socket-fd to getpeername but it returns the
>local-host
>IP-address (127.0.0.1).
>
>So what's the matter ?
>
>Maybe work around the slave part of the pseudo-tty ?
>
>Sorry for this silly question but I'm not a network programmer, really
>this is
>my first networking problem, and also reading Stevens' I haven't yet
>found a
>solution.
>
>Anyway, thanks a lot
>
>Lucio
>[EMAIL PROTECTED]
>
Lew Pitcher
System Consultant, Integration Solutions Architecture
Toronto Dominion Bank
([EMAIL PROTECTED])
(Opinions expressed are my own, not my employer's.)
------------------------------
From: Johan Kullstam <[EMAIL PROTECTED]>
Subject: Re: Translation of linux to minor languages
Date: 14 May 1999 13:57:18 -0400
"Stefan Monnier <[EMAIL PROTECTED]>"
<[EMAIL PROTECTED]> writes:
> >> Regexp matching can be done on UTF-8 just fine (no conversion
> >> needed).
> > you *do* need conversion. what does "." mean for a multi-byte char?
>
> The regexp compiler needs a slight adjustment, but the string doesn't
> need to be converted.
the gnu emacs team tried this approach in MULE and, as it turned out,
adjusting the regexp parser rather more difficult and error prone than
just doing conversion to a fixed width form.
conversion isn't hard or all that time consuming. confirming that
conversion works isn't difficult. rewriting regexp and working out
all the bugs is.
both ways of skinning the cat are valid, it's just that one way has
proven itself to be better in practice.
just my observation.
--
johan kullstam
------------------------------
From: Johan Kullstam <[EMAIL PROTECTED]>
Subject: Re: 2.2.8 - Evil behavior
Date: 14 May 1999 13:53:47 -0400
Kevin Turnquist <[EMAIL PROTECTED]> writes:
> I upgraded from 2.2.7 to 2.2.8, and removed the reference to update in
> the rc.S file as directed from a couple sources.
>
> Within 12 hours, all the servers that received this OS were dead from
> "Divide Error (addresses) - Kernel Panic"
>
> One machine I can see, but this hit 6 computers. Fortunately, I just
> booted back into 2.2.7.
>
> Has anyone else seen this behavior? I've never had this problem until
> 2.2.8, and I compiled it exactly as I did 2.2.7 and previous.
there seems to be a 2.2.9 out now (since 13 may). i saw on the kernel
list a patch to fix some of the update stuff by the guy who made the
update elimination fix in the first place. i have no idea what the
2.2.9 patch entails since it seems like no one does release blurbs
with new kernels anymore.
in short, linux-2.2.8 seems to be a loser. i'd either go back to
2.2.7 as you did and wait it out or go for 2.2.9 and hope for the
best.
--
johan kullstam
------------------------------
Date: Fri, 14 May 1999 21:32:38 +0200
From: Ulrich Eckhardt <[EMAIL PROTECTED]>
Subject: portmap problems
Hi,
i want to run the informix database with kernel 2.2.9 . For getting
the database to work i must increase some kernel settings e.g.
__FD_SETSIZE in posix.h . With this changed kernel the database runs
fine, but the portmap failed now with an error in svc_init.
I have first updated and recompiled glibc2.1.1pre2 with the new kernel,
but this doesn't help. Now i want to recompile the portmap, but
couldn't find a recent version. I have only found a version
portmap_5beta which throws a lot of warnings and errors during
compilation.
Where can i find a recent version of portmap, which compiles
under glibc2.1.1 ?
Thanks
Uli
--
Ulrich Eckhardt
http://people.frankfurt.netsurf.de/uli/
Truly great madness can not be achieved without significant
intelligence. (Henrik Tikkanen)
------------------------------
From: [EMAIL PROTECTED] (Fong Fatt Chee)
Subject: Repost: Help needed with Mobitex Driver for linux
Date: Fri, 14 May 1999 19:41:45 GMT
Hi!
I need help with the mobitex driver for linux posted on freshmeat.
Pls share your configuration info; I have not been able to get it to
work properly.
Your input will be much appreciated
fong
------------------------------
From: [EMAIL PROTECTED] (Marco Anglesio)
Subject: Re: Hostile Takeover of Linux
Reply-To: [EMAIL PROTECTED]
Date: Fri, 14 May 1999 19:54:08 GMT
On Fri, 14 May 1999 12:19:31 -0700, Tom Emerson <[EMAIL PROTECTED]> wrote:
> -- "pointy-haired" types like to see "certificates", so they'll hire
>someone who has passed a sylvan test over someone with "years of field
>experience".
You're implying that Sylvan creates the tests; it doesn't. What it does is
distribute the test. They do quite a bit of that; it's their business.
What you're describing are weaknesses of multiple-choice tests that have a
fixed question pool - the program takes N questions from every area, so
that each test is reasonably different from the others but still of
roughly equal difficulty.
The hard but largely irrelevant questions are actually a good thing in
test design; they differentiate test-takers at the far end end of the
testing scale. This forestalls compression effects, which is important
especially for exams which require a high score to pass.
By compression effects, I mean inaccuracy caused by attempting to rank
multiple scores at the tip of a bell curve - on a standard test, the
difference in score between someone at the 95th and the 99th percentile is
almost trivial and very unreliable (the testees will have much more
inter-test variation between their own and their respective scores),
whereas the difference in score between 50th and 55th is much less so and
much more reliable.
Many modern testing schemes modify or abandon the multiple-choice model.
Microsoft uses case-study methods in some exams; newer exams use an
adaptive model. The computer-based GRE (produced by the College Board) is
adaptive. Don't judge the test as a failure until you've actually seen and
evaluated it..
marco
--
,--------------------------------------------------------------------------.
> Marco Anglesio | Alcohol, hashish, prussic acid, <
> [EMAIL PROTECTED] | strychnine are weak dilutions. <
> http://www.the-wire.com/~mpa | The surest poison is time. <
> | --Ralph Waldo Emerson <
`--------------------------------------------------------------------------'
------------------------------
** 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
******************************
