Visit Our Sponsors:		Table of Contents: The MailBag More 2-Cent Tips The Answer Gang News Bytes Linux User Caricatures , by Franck Alcidi The Art of Atari ST Emulation , by Matthias Arndt Audio Processing Pipelines , by Adrian J. Chung Microsoft's New Briar Patch , by Jim Dennis Winning the Battle for the Desktop , by Dennis Field Visual Debugging with ddd , by Wolfgang Mauerer Installing Linux on Root Devices Unsupported by Your Distribution , by Zwane Mwaikambo The Answer Gang's Posting Guidelines , by Ben Okopnik The Foolish Things We Do With Our Computers , by Mike "Iron" Orr Free-Software Appreciation , by Mike ("Iron") Orr Content Management with Procmail , by Pradeep Padala and Prakash Bulusu Writing Documentation - Part 1: POD , by Christoph Spiel Managing MP3 Playlists, The One UNIX Way , by zhaoway The Back Page

Linux Gazette Staff and The Answer Gang

Editor: Michael Orr
Technical Editor: Heather Stern
Senior Contributing Editor: Jim Dennis
Contributing Editors: Ben Okopnik, Dan Wilder, Don Marti

TWDT 1 (gzipped text file) TWDT 2 (HTML file) are files containing the entire issue: one in text format, one in HTML. They are provided strictly as a way to save the contents as one file for later printing in the format of your choice; there is no guarantee of working links in the HTML version.
Linux Gazette, http://www.linuxgazette.com/ This page maintained by the Editor of Linux Gazette, gazette@ssc.com Copyright © 1996-2001 Specialized Systems Consultants, Inc.

Send tech-support questions, Tips, answers and article ideas to The Answer Gang <linux-questions-only@ssc.com>. Other mail (including questions or comments about the Gazette itself) should go to <gazette@ssc.com>. All material sent to either of these addresses will be considered for publication in the next issue. Please send answers to the original querent too, so that s/he can get the answer without waiting for the next issue.

Unanswered questions might appear here. Questions with answers--or answers only--appear in The Answer Gang, 2-Cent Tips, or here, depending on their content. There is no guarantee that questions will ever be answered, especially if not related to Linux.

Before asking a question, please check the Linux Gazette FAQ to see if it has been answered there.

• radius
• HTML/CSS question - useful for dicussion
• neighbour table overflow
• Ethernet: Slow download, fast upload

radius

Hello,

I wonder if it's possible to make a radius server on linux to authenticate the users on a remote mssql server 7.0 database.we use ms radius server and want to try linux.and if it's possible which radius server would you recommend for this job?

Thank you

Hakan Bilginer

HTML/CSS question - useful for dicussion

I am currently trying to write html which will insert page breaks for printing, which is is CSS2, and just happens to be part of css2 not implemented in mozilla.

Is any anyone aware of any solutions to this using HTML/CSS1

neighbour table overflow

Hi all,

I just set up a nice little p120 with 2-NICs and RedHat7.0 for my mom and it is working great except for one thing which you have spoken of before, the "neighbour table overflow" message being printed out to the console.

In a response to a letter from James Zhuang, Heather spoke of how ifcfg-lo might be missing or that lo might not be up but on my system, ifcfg-lo is there and appears correct and lo is up and running. Also JimD mentioned that pump might be screwing up the loopback configuration and I am running that as my dhcp client; how might i remedy this problem? I'd be happy to provide more system information if it would help or if you have any other ideas on where i might look i would appreciate it.

Thanks,

-- Ian Berry

Old symptom, maybe a new problem. Anybody out there encountering this too? Even better if you've got a Tip sized answer. -- Heather

Ethernet: Slow download, fast upload

Hi all there,

I recently bought a OVIS Live FSH8R 10/100 MPS autosensing switch and a Davicom 9102 network card (dmfe.c) (They deliver Linux drivers on disk!!) So I cabled all together and connected to a friends notebook. From my Apache server he can download with speeds up to 9 MB/s, that's what I expected. Now when I download files (big files to measure the throughput) from his Apache (on Windows or via smbmount/Samba, I'll just get a rate of 2,5 MB/s. With iptraf (an ip traffic analyser, it's very good, IMO) I get a lot of big packets (1400-1500 bytes), which are the data packages. But my network interface also receives/sends alot small (<100 bytes) packages. These are confirmation packages, I assume.

Both network cards run at 100 Mps, changing switch ports did not help. The switch has only two cables connected.

So, nice story . The question: Why can't I get a fast downlink, but only a fast uplink.

Thanks, Matthias

• Diablo under Wine
• Tnx Ben & Breen
• re Nov issue - Dennis Field article
• Two readers address -- RE:Battle for the Desktop: Why Linux Isn't Winning
  Ian Carr-de Avelon
  Robin Rowe

Diablo under Wine

I read your article in Published in issue 71 of Linux Gazette October 2001 where Jefferson said Diablo ran fine under Wine at LWE. Which Diablo? I or II. I have II running fine, but I can't find anyone who has Diablo I running.

thanks,
chuck

I really don't recall whether it was I or II - but it was the honest to goodness CD from the Windows software package.

It was in the TRANSGAMING.com booth; Transgaming makes ActiveX extensions for WINE, and with those extensions, many games run fine. -- Heather

Tnx Ben & Breen

Ben & Breen
Thank you people who offered me help on loading modules, into the 6.2 system I havent got it right yet, but learnt something for sure. I dumped the system and will start from scratch again with a different card.
Thanks agn bill

re Nov issue - Dennis Field article

Would you kindly pass on to Mr Field that "Lunux Canada" seems to have exactly what he is looking for. It is not free, but moderately (compared with MS) priced. Certainly it is worth checking them out.

I found them at www.linuxcanada.com

They didn't have anything that interested me but they seem to have exactly what Mr Field needs.

Thanks for your wonderful magazine.

LG has received announcements from Linux Canada before about their POS products. I sent the URL to Dennis, but I also noted that I couldn't get into the site when I retested it. However, I can get into the site now. I think Dennis is looking for more of a software solution, whereas Linux Canada is more geared toward special hardware. But maybe Dennis will find what he needs. -- Mike

Thank you for your rapid reply.

So far as I know Linux Canada only sells software. There has (in the last 2 years) never been a suggestion of hardware sales or availability - only accounting and point of sale SOFTWARE.

The only time I have had trouble getting into their website was when the backbone was clogged, you could not even get anywhere in Canada then. You may have had a similar problem.

Harold A. J. Isaacs

RE:Battle for the Desktop: Why Linux Isn't Winning

There are a whole series of relatively common problems related in this article, like applications software which does not run smoothly on all distributions and hardware which is not suported by a distribution, or sometimes by any Linux driver at all.

The real point is what can be done about this, and who should do it. The author, Dennis Field, seems to put the blame with the Linux distribution he chose. He, or his employer, paid them good money and so Linux should do what he needs it to do. There is nothing wrong with the logic of this, but the prospect of all distributions, or the most comercial distributions, or even one Linux distribution running out of the box on all old and new PCs is pritty well nill. Also even if that did happen, it would not get Mr Field to his goal of using Linux for the whole bookshop in one giant leap.

The problems which would need to be fixed even in this one "case study" are spread accrose the developers of: Linux kernel, X windows, StarOffice, and the distribution's firewall. Each of these have nowhere near the resources or assistance from hardware designers which Microsoft has, and yet they produce software which (on other criteria than out of the box installation on all PCs) far outperforms Microsoft's products. They have a right to pride in their work and respect from others, and simply calling for any of these teams to work harder till the problems nolonger exist would really mean accepting that Linux has lost, because even if the developers gave up all their free time, the extra improvement could never have the kind of impact which the author desires.

I have writen elsewhere (http://www.suite101.com/article.cfm/15359/84340) how I would generally advise SMEs to move towards Linux, but given the evidence of interest in building fairly complex systems at low cost, maybe we need a business by business (Bookshops, Chemists, Garages) Linux forum to give system integration with Linux the kind of boost which Linux Documentation Project (LDP) and sourceforge have done in other ways.

Yours
Ian

[Dennis also has another article in this issue. -Iron.]

Re: Battle for the Desktop: Why Linux Isn't Winning

Dennis,

It is unfortunate that you had trouble installing Linux on your ThinkPad. You don't say what distros you tried, and each distro works a bit differently. Since installation troubles are really the domain of the distro and not the operating system itself it seems unfair that you name the problem as Linux but avoided naming the distros that actually caused you the trouble.

Have you tried Debian Linux? This is a very popular distro, supported entirely by volunteers. You could download the boot/root floppy images from debian.org, copy them on to two floppies, then boot and install a minimal Debian Linux OS. This is usually quite easy. After configuring your Linux network settings and adjusting your sources.list to point to the Debian download site you could then use dselect or apt-get to install the rest of whatever you want automatically over the Internet.

That you would have the troubles you did isn't too surprising. Most Windows users would have trouble in the similar circumstances trying to install Windows on their own for the first time without ever having used it before. A more realistic approach when installing Linux for the first time is to enlist the aid of other Linux users. Most Linux users groups host monthly installfests for this reason, so that new users get the install help they need from experienced hands. Had you done that I expect you would have had an operational Linux laptop within minutes. Another approach is to join the debian mailing list where anybody usually gets answers to install problems within a few minutes.

Your analogy comparing Linux to a Ferrari with no wheels is unjust. The wheels are right there, you just had a problem with your "Some Assembly Required" situation. You were not picking up your new Ferrari at the dealer, were you? If you were buying a new Thinkpad purchased from IBM with Linux already pre-installed (look under "personal systems" at www.ibm.com/linux) you should have no installation problems whatsoever.

Please let me know if I can be of any help to you installing Debian Linux on your laptop.

Cheers,
Robin

• Re: new LG article: The Art Of Atari ST Emulation
• Issue 13

The distinction between Tips and Articles

Articles go to gazette@ssc.com in HTML format. Tips, tech-support questions and tech-support answers go to linux-questions-only@ssc.com in text format.

Basically, a tip covers just one simple topic in a screenful or two (or less). An article covers several subtopics under separate headings and/or is more than a couple screenfulls.

... and Tips rarely involve more than one Answer Gang member's comments. I do like to use this rule of thumb: would the answer (possibly without explanation) fit in a .sig block? If not, it's probably not small enough to make a good Tip.

When the discussions get going even a short thread might go into TAG, and some of the better long answers in TAG still aren't as long as a standard article. I'd also like to remind people that for article submissions, we prefer simple HTML to the font-laden stuff automatically generated by most web browsers when "mail as HTML" is turned on. -- Heather

Issue 13

The following section isn't displayed correctly because you don't use &lt; and &gt; for the <stdio> part.

      #include
      main()
      {
              printf("test\n");
              fflush(stdout);
      }
      They will not echo what I print.

Thanks Russell. We don't generally go through the back issues to correct things like this; things have gotten a lot better since then, and I'm proud to say that things have to be sneakier that that to put bugs into our HTML code these days -- Heather

This page edited and maintained by the Editors of Linux Gazette Copyright © 2001
Published in issue 73 of Linux Gazette December 2001

HTML script maintained by Heather Stern of Starshine Technical Services, http://www.starshine.org/

• Command-line calculator
• Apache startup script improvement
• Re: De-enhancing text
• Fun with chroot jails
• Password list
• DNS
• using m-w online dictionary.
• PacHell DSL w/LINUX
• How we fixed "FW-I/LINUX kmalloc" problem
• DSL Drivers for USB
• gtkmm-config problem
• newbie question --or--
  Linux equivalent for Active Directory?
• Re: [LG 72] 2c Tips #4 translated oddly

Command-line calculator

One of the things I've always found amusing is watching people working at a PC suddenly stop and go digging through their desk for a calculator. I mean, good grief - all that processing power, and they have to go back to the Stone Age! Well, if you're one of those unfortunates, suffer no more. Just put the following lines in your "~/.bash_profile":

calc(){ perl -wlne'print eval'; }
export -f calc

The next time you log in (or if you source ".bash_profile"), the function will be available to you.

ben@Baldur:~$ calc
3.141592653*6**2   # What is the area of a circle 6 meters across?
113.097335508
( 3 - 117 ) % 7	   # If today is Tuesday, what day was it 117 days ago?
5
sqrt(115) * 1.34   # Hull speed of a ship with a load waterline of 115'
14.3698990949832
ben@Baldur:~$

Note that I actually typed those comments into "calc"; it chews and swallows them without a problem.

"calc" is actually a 'gateway' into Perl (via the "eval" mechanism); that makes it into quite a powerful gadget. It supports all the math/trig/etc. operations that are built into Perl - functions like "abs", "atan2", "cos", "exp", "hex", "int", "log", "oct", "sin", "sqrt", and even "rand" (rolling dice, anyone?)

ben@Baldur:~$ calc
print int rand(6) + 1 for 1..20		# Roll 20 6-sided dice
6
1
6
5
3
5
3
5
1
1
6
4
6
3
3
4
1
1
1
4

"calc" can be as simple as you like - or provide you with the kind of power that calculators just can't match. It's all in what you choose to do with it. By the way, be aware: there's nothing in "calc" that restricts you to "math-only" commands; if you type "unlink my_important_file", Perl will happily obey your orders (i.e., delete that file.) So, as with everything in Linux, be careful - and have fun.

[Python's interactive mode can also be used as a calculator. -Iron.]

Apache startup script improvement

Every thime I setup Apache I add two lines to the startup script to parse the config file for the variable containing the name of the file to store the PID at.

It seems logical to me to automate this, since the script has an entry for the pidfile, but really should also "knows" the location of the config file, why not parse any redundant information from it and remove the risk of conflicting parameters.

Here is what I add to the /etc/init.d/apache start|stop script:

CONFIG_FILE=/etc/apache/httpd.conf
PIDFILE=`sed -e '/^PidFile /!d; s/PidFile //' $CONFIG_FILE`

or for you bashers:

 PIDFILE=$(sed -e '/^PidFile /!d; s/PidFile //' $CONFIG_FILE)

Also, I usually pass the name of the config file to apache explicitly, so that it's obvious via "ps ef" what configuration is currently being used.

Seems to make sense to me. In fact, I'd hope this makes it into the scripts included in the distro.

[JimD] It's a good suggestion.

Personally I think the start-up (rc) scripts from most distributions are a bit lacking. For example I've always thought that it was remiss of the start up script that mounts the /proc filesystems fails to check that the mount point is a properly empty directory.

In the case of your suggestion, you are eliminating what I call a "moving part" (an opportunity for different configuration elements from different sources to get out of sync with one another).

Of course there are many other failure opportunities which could be mitigated with additional tests. For example: what if there are multiple PidFile directives? what if the case doesn't match your sed expression (doesn't Apache tread PidFile as equivalent to PIDFile, etc)?

[Ben]

The usual way that Debian does it is also fairly sensible. This is from "/etc/init.d/skeleton" (the template that you're supposed to use when writing an "init.d" script under Debian), by Miquel van Smoorenburg and Ian Murdock:

See attached apache.init-d-fragment.txt

Any daemon, when started via this mechanism, gets an individual pidfile.

[JimD]

I think you miss his point. Debian's rc scripts are no better than Red Hat's in this respect. If one changes the PidFile directive in the .conf file, then Apache's notion of its PID file location disagrees with Debian's startup/shutdown scripts.

That could be reported as a bug to the maintainer --- but it's unclear how far we should go in making the rc scripts more dynamic. It would be a bit absurd to do comprehensive failure-mode analysis and mitigation for all of the rc scripts. At some point we must just give up (maybe calling on logger -s to emit and error message).

The problem with making foolproof systems is that the universe keeps creating more ingenious fools.

Re: De-enhancing text

Richard Bly sent us:

Just in case you were not aware, the utility colcrt will take a man page output and format it without all the weird stuff. The underlining is put on the next line so both the text and the underline are visable.

[Thomas Adam] Why not just use the following......:

man manname | col -b > ./mymanpage.man

where "manpage" is the man page (obviously). The "col" command in this case (with the -b flag) will filter reverse line feeds.

There is also the option of using "man2html" for the adventurous......

Guys,

I think that following method is much simpler:

man thttpd | col -bx

Regards
Peter Dzimko

Fun with chroot jails

Ben asked:

There's one you could write up (assuming you ever got the time to do it, that is) - creating those "chroot" jails. That's something I'd love to have the specifics of; I understand the concept well enough, but having never implemented one, I'm short on the actual mechanics.

There's a fairly current Freshmeat entry called "cage". Initial release. Not my stuff, but it's exactly the right idea - some support for a bash-shell centered chroot jail, so you can jail more complex apps a little more safely, e.g. make chroot a one way trip, nicking off a few linux-privs along the way.

Sounds like cool fun; I'll definitely check it out.

In the "barely enough to run an app" category, there's a helpful document for BIND, and a different one for Postfix, iirc, but I don't have their URLs memorized and I'm trying to avoid getting -too- distracted. (too late!)

<grin> I'll search for those some time this coming week...

There are a few patches and at least one kernel module (capsel) around now, that offer to stop the chroot() call from happening more than once, preventing the usual script-kiddy method of getting out of one, among their other helpful efforts.

Uh... what's the usual script-kiddy method? I mean, I know I can type 'exit' if I've started a regular 'chroot' without specifying a prog... but... maybe I'm not visualizing it right. I'm seeing a chroot jail as a "system within a system" - if you exit, you end up at a login prompt. That's it. Real "root" is only available via a different IP; in effect, you're logging into a different system. Correct?

Minimum Mechanics:

1. blank hard disk
2. install parent level with syslog, cron, ssh, sudo.
3. create subdirs for jail areas (e.g. /home/HTTPD-jail, /home/MAIL-jail, etc.)
4. run installer again, using "already mounted directory". Once per jail of course.
   Mhm. I wonder how hard it would be to create a stripped-down installer just for the purpose. Might make a nice project, don't you think?
5. tweak each jail like it was a seperate machine you could boot into normally that was dedicated to the purpose. Each jail's ssh must be on a unique IP address/port number combo.
6. grafting - setup top level so it runs services out of their jails, already chrooted there.
7. time to make an IPL backup
8. stripping - take more stuff out of the jails, that they will NEVER need because they are really not the top level after all. e.g. fsck, copy of the kernel and modules. This may require some brutal adjustments to the packaging systems so they won't get put back if you choose to upgrade the jails later. Possibly make it so there should never be a need to be root inside the jail anyway. etc.
9. time to make IPL backup #2, on a different media from #1. Allows for return to this point, or to decide you went overboard and try shaving that differently by starting again from #1.

Eh... you lost me there on #6; that's the part I'm not seeing. What's the interaction mechanism between the two levels? How does the "top" see the "bottom" without the "bottom" seeing the "top"?

I usually run a lot of things from /etc/inittab so they can be respawned if they die.

For #8 I agree, that's the way I would do it - since root can twiddle anything on the mounted filesystems, there shouldn't even be root access in there. Although I would set up some sort of an "admin" account, with carefully decided powers.

Might be helpful to have more hard disks, or seperate partitions for each jail. I gotta stop procrastinating like this ;>

I'm glad you did. Thanks - I'll dig into it some more!

Password list

OK, so this is straight out of any security FAQ: whatever you do, _don't_ keep a list of your passwords on your machine. Right? Right.

Now, since you're going to do it anyway... Here's a somewhat safer way to do it - note that I did not say "safe", just "safeR". The way I see it, those of you who don't keep one won't be affected, and those of you that do will notch up the security just a tad.

To make this work, you'll need something to keep your secrets for you:

See attached pass.bash.txt

Here's what you do: put this script in a directory that's in your path, say "/usr/local/bin", then set the ownership and permissions as follows:

chown root:root /usr/local/bin/pass	# You must be root to do this
chmod 755 /usr/local/bin/pass		# And this, too

You now encrypt the file that contains your list of hosts, usernames, and passwords, one per line:

www.cia.gov		JohnDoe			cRYpTo
www.kgb.ru		IvanIvanov		bOLsh0isEkRET
www.mossad.il		PloniAlmoni		sHiN8eT
kempeitai.jp		NanashiNoGombe		haITTeM0ikEmAsEN
www.mybroker.com	FulanoMengano		QuIenSaBE
www.mybank.bm		MattiMeikalainen	sAipPUakAuPPIAs
www.centralbank.an	JanModaal		fInanCIeeL

...with a command like:

crypt My1SecretPasswD < mysecrets > ~/pass

Move the original ("mysecrets") to a floppy and put it somewhere safe (yes, that usually means where nobody - not even you will ever find it again. . Remember to update it once in a while. As to the encrypted file, all anyone is going to see when they look at it (you did set its permissions to 0600, right?) is a bunch of binary-looking gobbledygook.

Now, let's say you want to see what the combo is for "mossad". Easy enough:

spy@Hideout.com:~$ pass mossad
Enter password (screen echo disabled):
www.mossad.il		PloniAlmoni		sHiN8eT
spy@Hideout.com:~$

If you want to edit the file, just type "pass -e"; this will invoke your editor ("$EDITOR" - "vi" by default) on the decrypted version of the file.

"grep"-related tip: if you want to just see the entire file, call it as

pass $

DNS

David Menegat asked us the following:

I am trying to set up a name server on my mandrake 8 system and I believe I installed the dns package I just don't know how to configure it do you know where there is a faq or have any advice for me. I just bought a domain name and this is the last piece in the puzzle before the final configuration and I transfer the name to my machine.

Thank you David Menegat

[Faber] Well, there's always the HOWTOs: http://www.linuxdocs.org/HOWTOs/DNS-HOWTO.html

To which David replies:

Thank you very much I'm sure I'll have no problem now
thank you
David Menegat

# def <word> goes to Mirriam Webster page of it definition
lynx "http://www.m-w.com/cgi-bin/dictionary?book=Dictionary&va=$*"

def word

bash-2.05$ g++ test.cc -o test `gtkmm-config --cflags --libs`
In file included from /opt/gnome/include/gtk--/base.h:34,

from /opt/gnome/include/gtk--/object.h:30,
from /opt/gnome/include/gtk--/widget.h:32,
from /opt/gnome/include/gtk--/container.h:27,
from /opt/gnome/include/gtk--/bin.h:27,
from /opt/gnome/include/gtk--/button.h:27,
from test.cc:2:

/opt/gnome/include/gtk--/proxy.h:6: sigc++/signal_system.h: No such file
or directory
/opt/gnome/include/gtk--/proxy.h:7: sigc++/bind.h: No such file or
directory
/opt/gnome/include/gtk--/proxy.h:8: sigc++/convert.h: No such file or
directory
test.cc:4: `#include' expects "FILENAME" or <FILENAME>

Answered By Dan Wilder, Udo Puetz, Mike Orr

to sir,

thanks for reply i wanted to know information regrading linux device drivers books or manuals pages from net and tcp/ip for professional people.

thanks ranjeet

[Dan] You've reached a mailing list administrative address. I'm forwarding your query to the <linux-questions-only@ssc.com> mailing list.

You might try a search for

linux device drivers

on http://www.google.com

I just tried it and got 304,000 matches, of which most of the matches in the first two pages (as far as I got) looked worthwhile to visit.

[Udo] Hi!

You could download "writing linux device drivers" on the oreilly web-page (www.oreilly.com) some time ago (I think 2-3 months ago). This was not the last release, but hey, it's for free and online

[Mike] There was an article about writing device drivers recently in LG:

http://www.linuxgazette.com/issue69/mathew.html

This article is about a PC speaker driver, but it serves as a general example. Republished from Linux.com with theirs and the author's kind permission. I requested this article for LG because we had a need for articles on device-driver programming.

fine-grained delay in shell scripts

From Ben Okopnik

Answered By Thomas Adam, Mike Orr, John Karns

So, you're writing a shell script, and you want to add a little pizzazz: you know, a little blinking indicator, or a slow display that prints out "Please wait" one letter at a time or something. You look around, and the only choices you've got are a) "sleep" (a minimum 1-second delay), or various strange things involving loops and "cat"ting large files (which makes you CPU usage shoot up into the 90% range.) Blechhh. What's a poor but honest scripter to do?

Farm the job out, of course.

See attached nap.pl.txt

It doesn't get much simpler. "nap" will give you a delay in milliseconds, plus a tiny machine-dependent fudge factor for starting Perl. Here, as an example, is that famous "rotating dash" indicator, using "nap":

while :; do for x in - \\ \| /; do printf "%c\b" $x; nap 100; done; done

[Thomas Adam] Tut tut Ben. For this kind of use, I always tweak the millisecond usage of the command:

"usleep"

Then I can use a for i in....loop and a usual "echo" in Bash.

Works everytime.

But, I prefer your script!!

OK, I'll admit my ignorance - what's a "usleep"? There's nothing like that on my system, or indeed in the Debian "Contents-i386.gz" file list. Please enlighten me. (I do seem to _vaguely_ remember something like that in C, but that's less than helpful.)

[Thomas] But, I prefer your script!!

Well, you got _something_ useful out of it. That's a plus.

[Mike] ..From "man 3 usleep": "The usleep() function suspends execution of the calling process for usec microseconds."

It looks like it's available only as a C function. Somebody should wrap it up in a command.

<smirk> I did.

[Thomas] ....and they did just that . I believe that on RedHat systems, it was supplied as part of the "initscripts" rpm, thus:

http://www.rpmfind.net//linux/RPM/redhat/7.2/i386/initscripts-6.40-1.i386.html

/sbin/usleep

is where my copy resides (despite the fact im running SuSE 7.1 professional).

Hope that helps

[John K] No such animal on my SuSE 7.2 install ...

jkarns@jkInsp8000:~ > locate usleep /home/jkarns/Dwnlds/Linux/XScreenSavers/xscreensaver-3.32/utils/usleep.c /home/jkarns/Dwnlds/Linux/XScreenSavers/xscreensaver-3.32/utils/usleep.h /home/jkarns/Dwnlds/Linux/XScreenSavers/xscreensaver-3.32/utils/usleep.o /usr/share/doc/packages/mod_php/doc/function.usleep.html /usr/share/doc/packages/phpdoc/manual/function.usleep.html /usr/share/man/allman/man3/usleep.3.gz /usr/share/man/man3/usleep.3.gz

As I'd mentioned, it's not part of Debian - whereas Perl is in every distro. I'm sticking with portability. Besides, when would you ever need microsecond precision in a shell script? Even milliseconds is splitting it frog-hair fine.

[Mike] You don't, but sometimes you want to delay for a quarter second or half a second.

BTW, "usleep" isn't described in "libc.info.gz", either - although there's an interesting-looking "nanosleep".

Serial Programming on an i486 in Linux

From V Sreejith

Answered By Chris Gianakopoulos, Ben Okopnik, Heather Stern

hi all,

This is my first posting. Here I have a problem related with linux serial programming in C.Hope u can help me in this.

I have a C program that communicates with a remote terminal unit via serial port.The program uses termios structure to initialise the port. The program works as expected in kernel version 2.4.2-2 on an i686 macine.

This same program when tested on a 586 (Kernel 2.2.14-12 on an i586) machine fails to read the port properly. Writing to the port was working properly. The setserial and stty commands produced the same result on both machines.

[Chris G.] What does "failing to read the port properly" mean? Do the data look rather odd? Does it look garbled?

Later i found that minicom(communicating with hyperterminal in windows) also showed the same problem while reading the port on the 486 macine. Writing to the port was working properly. Communication was proper when hyperterminal was used on both sides.

Here is the o/p of the setserial command.

/dev/ttyS0, Line 0, UART: 16550A, Port: 0x03f8, IRQ: 4

Baud_base: 115200, close_delay: 50, divisor: 0
closing_wait: 3000
Flags: spd_normal skip_test

and here is the settings on the port using stty command while my program is running.

[Chris G.] I am guessing that your baud rate is 115200 baud.

speed 9600 baud; rows 0; columns 0; line = 0;
intr = ^C; quit = ^\; erase = ^?; kill = ^U; eof = ^D;
eol = <undef>;
eol2 = <undef>; start = ^Q; stop = ^S; susp = ^Z;
rprnt = ^R; werase = ^W;
lnext = ^V; flush = ^O; min = 1; time = 0;
-parenb -parodd cs8 hupcl -cstopb cread clocal
-crtscts
-ignbrk -brkint -ignpar -parmrk -inpck -istrip -inlcr
-igncr icrnl ixon -ixoff
-iuclc -ixany -imaxbel
opost -olcuc -ocrnl onlcr -onocr -onlret -ofill -ofdel
nl0 cr0 tab0 bs0 vt0
ff0
isig icanon iexten echo echoe echok -echonl -noflsh
-xcase -tostop -echoprt
echoctl echoke

[Chris G.] I am guessing that your baud rate is 9600 baud.

If anybody knows what is creating the problem... please do help..

regards, sree

[Chris G.] Hey Sree, Are you running both sides of the link at the same data rate? The parameters that you displayed give me the indication that the data rates are not the same.

When you type on one terminal, do you see anything displayed on the other terminal? If they are different, one cause could be mismatched baud rate. Your stty parameters seem okay -- if they are not, someone else will probably jump in and correct me.

Get back to me (at you convience of course), and let me know what you see on the displays.

I hope that this helps, Chris Gianakopoulos

hi chris,

I think this isn't the problem of the code as this same code is working as expected in an i686 machine.

[Ben] I agree with you here, Sree; if "minicom", etc. are having problems, then your programming isn't what's at fault - it's "lower down". That could still mean that there are problems with the baud rate, however - the "lower down" part means anything from the serial port settings, through hardware initialization, to actual hardware problems (the last of which you've eliminated by trying it in Wind*ws.)

So it couldn't be the problem of baudrate,parity etc. Ofcourse my baudrate is 9600 and the stty command showed that i have initialised the port as expected.stty command showed the same o/p for i686 and i586(not i486 ..sorry for the mistake).

[Chris G.] That's ok. If you are at the same baud rate, you might expect things to work.

[Ben] Why the "of course"? I almost always set my port speed to 115200. One of the few exceptions is when I'm reading from my Minolta Dimage V camera: it will not work with anything past 38400.

Have you tried your program on another 586, preferably a different make and model? It could be that the serial hardware on that specific machine is wonky, or that it's sufficiently strange that the serial driver is having problems. I would also urge you to study the "setserial" man page; if indeed the driver is having problems, "setserial" gives you _tons_ of configuration options, and you might be able to "configure the problem away."

My problem(in an i586) is that i can't read from the port properly.I could get the no of bytes available on the port with an ioctl call.The program fails to read that bytes..

Sometimes it reads correctly a few bytes.. But most of the time the read function returns with Zero bytes read even if bytes are available on the port. Writing was working properly(verifying..by the return value of write function)

Earlier i thought it could be some problem with my code.Later minicom also showed the same problems.That is, whatever was typed on my keyboard was displayed on hyperteminal on the windows machine on the other side correctly.But when i typed something on the other side(in hyperterminal) it was not reaching here properly.That is, once it displayed some ..2 or 3 characters correctly and failed to show the rest.

[Ben] Hmmm... that _does_ sound like a speed problem. If the port on the other machine was set to, say, 57600, and this machine was set to 9600, it would "miss" most of the characters sent to it (although the classic speed problem shows up as random garbage mixed with a few of the characters being sent.)

When hyperteminal was used on these side also ..everything was ok. So i guess..there is no problem with the serial port.

Could this be some problem with the serial driver in linux.

i think the problem is more clear now...

sree

[Ben] Wind*ws, depending on the version, sets the serial port speed to 57600 or 115200, if I remember correctly.

[Chris G.] Hi sree,

You have got an interesting problem here. First, let me mention that I read Ben Okopnik's response, therefore, I will attempt to provide information that does not overlap his response. Take his response into consideration.

Let's pretend that your baud rates are the same on both sides. When you type on your Linux machine, the characters appear on the Hyperterminal display. So, you Linux machine can send, and your Hyperterminal machine can receive.

Now, when you type on your Hyperterminal machine, I get the impression, that the first few characters that you type appear on the Linux machine (my assumption may be wrong -- correct me if I am wrong). After that, when you type any more characters on your Hyperterminal machine, the Linux machine displays nothing. My first guess would be that flow is enabled on your Hyperterminal machine, and you are using a 3-wire connection.

Flow control being enabled would seem like a logical cause for your problem except for one thing. I get the impression that if you type ten characters on your Hyperterminal machine (excluding the enter key), that your ioctl() returns a value of ten. The results seem contradictory. I believe that you are telling me the truth -- therefore I am now confused.

One trick that I use to test a serial port on a computer is to test each serial port one at a time. Go ahead and remove the cable that connects your two computers. On your Hyperterminal computer, connect a serial cable to it, but, do not connect the other end of the cable to your Linux machine. On the cable, connect pins 2 and 3 together. This will connect your transmit data line to your receive data line. Bring up Hyperterminal and start to type characters. If things are working, you should see whatever you type appear on the display.

Do the same thing on your Linux machine. That is, connect a serial cable to the serial port of your Linux machine, and connect pins 2 and 3, of the connector, together. Bring up minicom, and go ahead and type characters on your Linux machine. If things are working, you should see whatever you type appear on the display of your Linux machine.

In both cases, we are performing a loopback test. If you do not see a proper display, you have isolated the machine for which a problem exists on a serial port.

If you do see proper display, I would expect the following causes: 1. Mismatched baud rates between the two computers (maybe not) 2. Problems with your RS-232 drivers (voltage level problems).

If you don't see proper display on each machine, I would expect the following causes: 1. Problems with your RS-232 drivers. 2. Flow control enabled (you could disable flow control)

Try Ben's suggestions too. My analysis may be incomplete.

Let me know what happens, and good luck, Chris Gianakopoulos

Hi chris amd ben,

Sorry for the delay in replying as there was some problem with our mailserver.

[Chris G.] No problem.

Now, when you type on your Hyperterminal machine, I get the impression, that the first few characters that you type appear on the Linux machine

Not always...it appeared rarely...

[Chris G.] Hmmmmmm.....

(my assumption may be wrong -- correct me if I am wrong). After that, when you type any more characters on your Hyperterminal machine, the Linux machine displays nothing. My first guess would be that flow is enabled on your Hyperterminal machine, and you are using a 3-wire connection.

Hardware Flow Control is enabled in hyperterminal(That's the default..) That is set in minicom also.I changed only the baudrate. (I am in doubt about this cable whether it is using 3 pins only or more)

But the cable I use in my program uses the rts,cts pins and uses them in data transmission since the rtu expects that.Everything is working fine in 686 machine.

Flow control being enabled would seem like a logical cause for your problem except for one thing. I get the impression that if you type ten characters on your Hyperterminal machine (excluding the enter key), that your ioctl() returns a value of ten. The results seem contradictory. I believe that you are telling me the truth -- therefore I am now confused.

I haven't checked the ioctl call with hyperterminal.I am using minicom with hyperterminal.

[Chris G.] Understood.

The ioctl call is used in my program that communicates with an rtu through serial port. I am sending some data to the port and expecting some data in return. The ioctl call returns the no of bytes available on the port. But the read function fails to read them and returns zero bytes.

By using hyperterminal on both computers communication is perfect. So that proves there is no problem with port or cable.

[Chris G.] I agree.

If there is some problem with the settings mismatch in hyperterminal and minicom ...how is the communication working correctly in one direction?

[Chris G.] The direction that is not working would be the one that is not getting the proper handshake -- if my flow control idea is valid.

I will check with flow disabled in hyperterminal.

I want to know whether this is a problem of linux? I am asking this because ..when hyperterminal was used on both sides communication was working. Have anybody encountered this type of problem in 586.

[Chris G.] That is what is odd. A serial port is not really associated with a processor -- Ok, ok, some embedded microprocessors do have lots of serial devices incorporated into their package. I would think that you are using something like a 16450, 16550, 8251, or someother UART or USART. It is just a device that is addressed by the Intel processor in I/O space. I understand what you mean, though. When you say 586, you are referring to your particular 586 motherboard with its oddly behaving serial interface (as far as its behavior with Linux is concerned).

I am waiting to check this problem with another 486 machine.

I wan't to know more about the flow control aspect. What is its significance?

[Heather] That's "ready to send" and "clear to send". One system says it wants to .. has data ready to go... and the other system has to give it the thumbs up... clear, you can send now.

[Chris G.] When you want to send, you assert rts. Your system will monitor cts, and it will not send the data unless cts is asserted (when I say asserted, I mean set to an active state). If the computer, at the other end of the link, does not activate the cts line, your system will never send the data byte. It will be flow controlled. As a quick test, you could make a cable that jumpers rts to cts. That would be pins 7 and 8 on your db9 connector.

I am using the rts/cts pins in my program. I do it like this.. I will make the rts pin high before writing and wait for cts to become low. Only when cts becomes low will i write to the port. However i am not using any setting/resetting while reading the port. Everything is working fine in 686.

bye sree

[Chris G.] Okay. You seem to be troubleshooting things in a good manner. Don't forget to look at Ben O's suggestions, too.

Shut down when turn computer off

From pclouds

Answered By Richard Adams, Frank Rodolf, Mike Orr

What I think he is talking about is that with some computers (dell is the only one I know) that are running windowsNT if you hit the power button on the front of the pc they begin to do a software shutdown.

Thats what he is asking how can he set it up so when he hits the power button that the machine notices this and goes into a software shutdown..

[Richard] Well look at it this way, by depressing the power button on the computer you start a shutdown, ok yes i now know what wa meant, what Linux has that no windows software has is "ctrl-alt-del" invokes the defined call in /etc/inittab by default its in short, "shutdown -r now" that one can change to use 'poweroff' 'shutdown' or 'reboot'.

So instead of pressing the power button on the machine itself, hit ctrl-alt-delete, advantage of that is no need to streach to reach the machine as the keyboard is right in front of you.

[Mike] There is also a POWERFAIL signal Linux uses to signal the detection of an imminent power failure. It's meant for emergencies, but maybe you can somehow latch onto that infrastructure. But an option like 'ctrlaltdel' in '/etc/inittab' would be more ideal.

ATX-style computers have a momentary on/off switch that can (somehow) be made to trigger the "shutdown" command -- maybe. I'm not sure how it's done. Look through "man inittab", "man init" and the kernel documentation, and maybe you'll get an idea.

This won't work with older AT-stype cases and motherboards because they have a true 2-position switch. When you turn the switch off, it cuts the power mechanically, and Linux doesn't know about it. Even if Linux did know about it, there's not enough processor cycles left for Linux to do a clean shutdown before it dies.

[Frank] I did some quick lookup... It seems acpid can do this - should be available with your distribution, or otherwise you can easily find it online.

If not, there is also something called "Linux PowerSwitch Driver", which is meant exactly for what you want... You can find it at:

http://deadlock.et.tudelft.nl/~joris/powerswitch

Hope this helps you!

Thanks for all advices. I have got powerswitch. It work very well!

slib installation

From Dann S. Washko

Answered By Ben Okopnik

While compiling gnucash I needed to install the slib libraries. I grabbed the latest files from SLIB site and looked over the install information. I initially put the slib files in /usr/lib/slib, but this was not working, g-wraper kept puking saying it could not find require.scm. Looking at the path's listed, I moved slib to /usr/share/guile/1.4. I reconfigured g-wraper and ran make with no errors.

Is this where slib belongs?

[Ben] Just a datapoint: on my (Debian) laptop, "slib.scm" is in /usr/share/guile/1.3.4/ice-9/. It sounds like the right neighborhood.

Yeah, that is where it is on my slack system. The g-wrapper program was looking for it in /usr/share/guile, so that is where I put it. I managed to compile gnucash successfully. It puked on the first run saying I should run it as root. Again, this was an issue with the slib libraries.

strace gnucash showed some issues with opening slib.cat, but after running it as root once, I was able to run it as a user.

[Ben] Hm. That makes me wonder if it creates some sort of a config file that it relies on in a place wher you normally don't have write permissions. <Shrug> Just a wild guess.

This makes me wonder if my placement of slib is not 100% correct.

[Ben] I would think that if it wasn't, it would continue to fail. Either way, it's working now, right?

Thanks for your reply, Ben.

[Ben] You're welcome, Daniel; little enough that it was, I'm glad that it was of use to you.

SuSE 7.1 installation CD not recognized

From Tom Zabosky

Answered By Dan Wilder, Karl-Heinz Herrmann, Jim Dennis

Good day

I came across the undermentioned question to yourselves and it is the identical problem I have been having. Could you please send me the answer given to this question. Thanking you in advance for your kind consideration in this matter.

yours

Tom Zabosky

[Dan] I'd suggest you ask SuSE this question.

[K.-H.] I agree with Dan on this -- SuSE would be the place to ask.

If you didn't fetch a boot disk as well and burned the CD explicitly as bootable CD with that disk-image of course it can't be bootable. See cdrecord man pages and README's for details or move to a search engine of your choice and type "burn bootable CDR" or something similar.

The partitioning improvements of yast2/7.1 are that you still cant choose any more complicated layout than a continous part of a HD. (seems to have improved with 7.3).

SuSE/yast requires a list of packages and lots of information for administration of the system. probably you forgot that.

You've no problem with filesystem type/extensions on the CD? like all caps filenames instead of lower case ones?

I guess SuSE put something there so it's not that easy to just bake your Installation CD yourself where lots of needed stuff might be needed. Aren't there CD images which would even be bootable?

[JimD] Actually it's probably a replay of the old question:

I downloaded a .ISO image and it doesn't work after I burn it to a CD

(i.e. the user/querent is using CD burning software that is not recognizing the target file as a pre-mastered ISO image).

The best answer to this question seems to at the LinuxISO web page:

http://www.linuxiso.org/cdburninginfo.html

... where one can also find ISO images of over twenty Linux distributions, as well as FreeBSD, NetBSD, and OpenBSD (which I usually call *BSD or {Free,Net,Open}BSD for brevity's sake).

What I particularly like about this page is that they used to have a Javascript animation that showed a sequence of dialogs and options in "Easy CD Creator" that selected the appopriate "Options, Track Type" settings for burning pre-mastered ISO images. (Even though that seems to have disappeared, the information is still there; MS Windows users just have to READ it and find the

Installing tulip.o in 6.2

From William Laing

Answered By Ben Okopnik, Breen Mullins

hello

Can someone instruct me how to install the module/ driver in Linux 6.2 .for the Linksys networking card.

The following module came with the card on a floppy and I was able to load it in as follows as per the instructions. I have tulip.o loaded in at this location.

# locate tulip.o
   /lib/modules/2.2.17-14/net/old_tulip.o
   /lib/modules/2.2.17-14/net/tulip.o

Does the old file require to be deleted or may it stay ? I am be having some fun, but not making any headway on insmod.At this point I cant go ant further.

Thanking you. Bill

[Ben] Did you try loading the module that was already on your system? I would do that before using the new one, by preference - consider that to be your fallback position if the new one fails, or gives you any problems. Presumably, you're using the 2.2.17 kernel; all you have to do is type

insmod tulip

If the module loads without errors, check to see if the system actually "sees" the card as it should:

ben@Baldur:~$ ifconfig eth0
eth0      Link encap:Ethernet  HWaddr 00:4C:69:6E:75:79
          inet addr:192.168.0.3  Bcast:192.168.0.255  Mask:255.255.255.0
          BROADCAST PROMISC MULTICAST  MTU:1500  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100
          RX bytes:0 (0.0 b)  TX bytes:0 (0.0 b)
          Interrupt:11 Base address:0x200

Note that your numbers may not be the same as mine; in fact, the line starting with "inet addr" will most likely be absent. This is normal.

Truth to tell, I'm a little surprised at the fact that the manufacter included a module: they're normally compiled for a given kernel, and will error out (although they can be force-loaded) when pushed onto a different one.

The old module (I assume, from the above, that you renamed the original to "old_tulip.o") does not need to be deleted. If you want to test the system with it instead of the new one, you should just be able to type

insmod old_tulip

Don't try to load more than one of them at a time; unload the one that you don't want with

rmmod <module_name>

You can always see what's loaded by typing

lsmod

Ben

Thank you for Clear instructions, much apprecated. The following happens while using the commands you suggested.

Presumably, you're using the 2.2.17 kernel; all you have to do is type...

Yes a 2.2.17-14 kernal

insmod tulip

Yeilds as follows:

Using /lib/modules/2.2.17-14/net/tulip.o: init_module:
Device or resource busy Hint: insmod errors can be caused by incorrect
module parameters including invalid I0 or IRQ parameters.

[Ben] That sounds like it's looking for parameters. This can also be caused by I/O or IRQ conflicts, but ...

[Breen] The tulip cards are all PCI and you shouldn't attempt to specify either IRQs or IO addresses for them. The system assigns these for a PCI device, not the driver.

[Ben] Right. Check "/proc/pci".

insmod old_tulip

Yields the same as the new one. I believe I am guilty of copying both modules myself to the linux box.

[Ben] It sounds like either one will work, once you have the correct parameters.

lsmod

Yeilds:

# lsmod
Module                        Size             Used by
#

With nothing listed.

[Ben] This is fine.

Both machines can ping each other while Window applications are installed suggesting the network path is funtional.

[Ben] Good - knowing that will narrow down any troubleshooting that you may do.

Ben, Breen

Agn thank you people for you kind assistance. The results of your suggestions follow. (It is a bare bones machine text only.)


The kernel is a 2.2.17-14
The card is a Linksys LNE Ver.4.1 TAIMAG HE-012D

I do have other D-Link Cards I have tried, using RTL8139, but the results were the same.

# /var/log/dmesg
    Permission denied
#

[Ben] Erm, you're supposed to read it - not execute it. "/bin/dmesg" will print out the contents; for a bit more scrolling control, try

more /var/log/dmesg

Reading "daemon.log" and "messages" in "/var/log" would be of even more use - they would tell you what happened when you tried to load the module. A quick look at the available parameters for "tulip.o" t