Mailbag

Mailbag

[[[Thomas]]] - It's a topic that has come up before, and I believe the kernelnewbies.org site is quite good also.

[[[[Deepak]]]] - Thank You very much for your advice and suggestions.

[John] - I'm not 100% sure of it, but I think the following method will not overwrite your existing modules:

- In the kernel configuration menu, choose "General setup"

- Then "Local version - append to kernel release"

- Input a string (64 char length max, but you'll probably want to make it quite a bit shorter than that).

Then after booting from that kernel, "uname" command will return the string you entered in the step above appended to the usual kernel version number. The modules corresponding to that kernel should be installed under /lib/modules/KVER-localVersion where localVersion is the string you entered during kernel configuration.

I say I'm not 100% sure here because I haven't used that method since trying it once back in the days of kernel 2.0 or 2.2 at least 8 yrs ago, and my memory of it is a bit tenuous.

If I've guessed wrong, then I invite you to read "Asking Questions of The Answer Gang" at http://linuxgazette.net/tag/ask-the-gang.html and write us back - hopefully with your question rephrased as per the guidelines (i.e., including sufficient information), and we'll see if we can help.

[[Brian]] - I get a blue line for quoted text in Mozilla Thunderbird under Linux when T'bird is being "polite" and rendering HTML email instead of doing everything plaintext as I instruct it. To be more explicit, when I'm looking at a reply to an HTML email in Mozilla Thunderbird in the message pane, the bits that are quoted from prior espondent(s) are bordered on the left with a vertical graphical line. The color of that line changes with depth of quoting...

For example, in a later message in this thread, where Dale quotes Thomas, there is a blue line down the left side. Where Dale quotes Thomas quoting Dale, there is a DarkRed line inside the blue line, indicating two levels of quoting. In a plaintext message, I would instead see two greater-than symbols (that is, '>').

It is most irksome to me that these lines don't revert to plaintext quoting methods, even when I explicitly instruct T'bird to display the messages in plaintext. The TLA "WTF" sproings to mind...

[Ben] - Very nicely done, Mahesh! I wrote a similar function in my Bash tutorial; I've tweaked it quite a lot since then. Yours is prettier, though, and includes the 'bold' versions.

  colsel () # Color selector - iterates through a $TERM's color choices 
  { 
  trap 'reset' 0 # Reset on exit 
  n=47 # Max foreground color value 
  while [ "$n" -gt 39 ] 
  do 
  m=37 # Max background color value 
  while [ "$m" -gt 29 ] 
  do 
  [ `expr $m % 20` -eq 0 ] && echo -e '\E[$40;1m|' 
  echo -en "\E[$m;${n}m| $n $m \E[$40;1m" 
  m=`expr $m - 1` 
  done 
  n=`expr $n - 1` 
  echo -e '\E[$40;1m|' 
  done 
  } 

Oh, and just because I can't resist:

  perl -we'for$a(40..47){print"\e[$a;${_}m$a $_|"for 30..37;print"\n"}' 

[grin]

Conversely, you could just use Moshe Jacobson's "color" program; it'll even show all the colors in much the same sort of layout when invoked with the '--list' option.

  > Well, the standard disclaimer applies. Any suggestions and corrections  
  > are welcome (all flames to /dev/null).

It may already be filled up; that's where we're sending ours...

  > PS: If you want any help, do an "nl -ba colors.sh". ;-)

I've always found 'vi' to be just as useful. :)

[[Thomas]] - I've always avoid the use of escape codes in favour of tput(1) -- at least then it works remotely, too.

[[[Ben]]] - How would you use 'tput' to set colors? I've just run 'infocmp', and it doesn't show anything related to doing so.

[[[[Thomas]]]] - tput(1) can query the terminal database, hence:

  tput setaf 1; echo this is red 
  tput setaf 2; echo this is green 
  tput setaf 0; echo now we are back in black 

etc...

[[[[[Ben]]]]] - Ok, I got you. Even though "tput" options aren't listed anywhere near the "tput" man page (for anyone interested, they're in "man 5 terminfo"), and there is a huge number of them with mostly-cryptic explanations of what they do, this is still an excellent tip.

So -

for n in `seq 0 7` 
do 
        for m in `seq 0 7` 
        do 
                tput setab $n    # Set background 
                tput setaf $m    # Set foreground 
                echo -n "$m $n" 
                tput op          # Reset to original color pair 
                echo -n "|" 
        done 
        echo 
done 

The data displayed can be used with 'tput' invocations to set the colors shown, much like Mahesh's script, but would be term independent. Pretty sweet...

[[Martin]] - Just to let you guys know that there is a BASH Prompt HOWTO at tldp.org which gives you more info on colouring your prompts and putting other things in there too... ;)

[[[Ben]]] - I recall reading it a couple of years ago. Unless it's been updated, it uses escape codes - which, as Thomas pointed out, aren't necessarily term-independent.

[[[Mahesh]]] - [Quoting Thomas]:

  > I've always avoid the use of escape codes in favour of tput(1) -- at
  > least then it works remotely, too.

Yeah, but this was inspired by reading the console_codes(4) man page and seq(1) info page. Combining the two resulted in a useful(?) script.

[[[[Thomas]]]] - There's nothing wrong with it, but hard-coding escape codes is going to turn around and bite you on the arse before long. Some terminals don't/can't render escape codes properly.

[[[[[Mahesh]]]]] - Just out of curiosity, how would a modified script using tput(1) would be able to print the colour coordinates to the o/p?

[[[[[[Ben]]]]]] - Just like using escape codes does - except you don't use the escape codes. :) The point to 'tput' is that it actually queries the terminfo database and outputs the correct escape codes for a given terminal. Using explicit escape codes will work much of the time - after all, many Un*x terminals use the same ones - but will fail miserably on terminals that use something different.

If you're asking about an example, here's one:

tput smso		# Turn on bold mode
echo This should print in bold
tput rmso		# Turn it off

On the other hand, if you use these a lot, you're going to get tired of typing - so it makes sense to request them once and save them:

a=`tput smso`
b=`tput rmso`
echo "${a}This should print in bold$b"

[[[[[[Thomas]]]]]] - The o/p being what or whom? (Abbreviations, unless they're generally well-recognised, are pretty much useless for communicative writing).

[[[[[[[Mahesh]]]]]]] - Wow... looks like you're in a hot mood! (uh, read: angry)

[[[[[[[[Thomas]]]]]]]] - Far from it. Welcome to the ambiguity of email.

[[[[[[[[Ben]]]]]]]] - Nah. The last time Thomas got angry... well, you heard of what happened at Troy, right? When he gets angry, it's obvious (although the people to whom it's obvious may not have much time to regret it.)

In other words, don't worry about it. Unless you've already been slaughtered and buried in an unmarked grave - in which case, disregard everything I've said. :)

[[[[[[[Mahesh]]]]]]] - o/p == out put

[[[[[[[[Thomas]]]]]]]] - Ah, I see. In the English language, "output" is all one word.

[[[[[[[[[Mahesh]]]]]]]]] - Here's some more that I'll be (or someone else'll be) using

	w/o == without
	h/w == hardware
	s/w == software
	s/o == Significant Other (*not* son of)
...

[[[[[[[[[[Kapil]]]]]]]]]] - Why not "son of"?

Anyway, you seem to be missing the point. It is generally up to the communicator (in this case you) to ensure that abbreviations that you use are known to the communicatees (in this case not just Thomas or TAG but possibly other readers of LG).

Otherwise it is rather like sending an ogg file to someone who uses a Macintosh or a Word file to me. The recipient might be able to figure out what you are trying to say---but might press the delete button instead.

[[[[[[[Mahesh]]]]]]] - How do you get the colour coordinates using tput?

[[[[[[[[Thomas]]]]]]]] - Coordinates for what, or where? All tput does is query the terminfo database.

[[[[[[[[[Mahesh]]]]]]]]] - Uh, ... fotget it!

[[[[[Mahesh]]]]] - And how would you create (a hacked up version) of files like, say, /etc/issue using tput w/o using hard coded escape codes?

[[[[[[Ben]]]]]] - Those are not terminal codes, they're 'getty'-specific escape codes; no relation between the two. Yes, it's possible to use ASCII escapes to create lots of special effects; heck, I used to do that myself, back in the early '80s for BBS "post-login" screens. I just seriously doubt that this hack from 25 year ago is still in any way "cool", or even interesting - and this is filtered through the fact that it's guaranteed to break on some percentage of terminals.

As a side note, I suspect that 'getty' replicates a fair amount of 'tput' functionality, since it has to deal with terminals at that level.

[[[[[[Thomas]]]]]] - You probably wouldn't, which is why the initscript in Gentoo and the like waste massive amounts of code in adding nice pretty colours which last for lesss than ten seconds on the screen, provide next to no manner in terms of usefulness and will fall over and die spectacularly on different terminals. Is this gain worth it? Of course not.

[[[[[[[Mahesh]]]]]]] - I was thinking of "private" use... in /etc/issue not /etc/issue.net?

[[[[[[[[Thomas]]]]]]]] - That depends on your distribution.

[[[[[[[[[Mahesh]]]]]]]]] - Arrgh! I don't seem to get my idea thru to you. See, when you feel like creating an /etc/issue that'll emit your '\l' getty escape codes using so-called hard-coded escape colour codes like "^[[33;1m\l^[[0m", how can tput come to the rescue? Like you code up a shell script to o/p (err.. output, sorry) a /tmp/issue that'll have correct escape codes querried form the terminfo database.

[[[[[[[[[[Kapil]]]]]]]]]] - As Ben pointed out you can indeed do that. For example, you can have /etc/default/issue which is used/parsed by /etc/init.d/generate_issue to generate the "real" /etc/issue by using the terminal settings appropriate to /dev/console (which may or may not be a serial console). Beware that this /etc/issue will break on some terminals that are "getty"-ed via serial line.

[[[[[[[Mahesh]]]]]]] - like it'd be useful to have ^[[31;1m\n^[[0m to see your hostname in bright red or something...

[[[[[[[[Thomas]]]]]]]] - Again, can be achieved via tput (you've been shown examples of this.)

[[[[[[[[[Mahesh]]]]]]]]] - Can tput escape codes be interpretted by getty?

[[[[[[[[[[Kapil]]]]]]]]]] - That "getty" only understands escape codes in /etc/issue instead of a generic description like that given by terminfo or slang is surely a fault of "getty"! Hack "getty" to fix this problem.

[[[[[[[[Ben]]]]]]]] - If I was going to do that - which I would not - I would generate '/etc/issue' via a script, with escape sequences created by 'tput'. However, the overriding problem that can't be handled at this level of interaction is that the console from which the user connects to your system is completely arbitrary - so there's no way for you to know what their correct escape sequences are. In other words, if (e.g.) a Solaris or a DOS user connects to you using a console that uses something different, all they're going to see is a bloody mess of square brackets and other line noise on their screen - and you can't prevent that from happening.

*This* is why it's a really bad idea to do those color tricks in '/etc/issue'. It's not because anybody's trying to prevent you from being cool; it's because you will inevitably make someone else's life harder by doing so.

[[[[[Mahesh]]]]] - I don't know, can u help?

[[[[[[Thomas]]]]]] - Nope, since I don't understand your question.

[[[[[[[Mahesh]]]]]]] - Like finding the codes o/p'd (err, outputted) by tput (maybe you wouldn't want it, but it could be found, isn't it?

[[[[[[[[Ben]]]]]]]] - Err, have you considered the word 'printed'? It's a perfectly good one, and causes no harm to anyone as far as I know.

[ In case you're still missing it: clarity, in a technical discussion, is far more important than saving a few keystrokes. WRITE OUT what you're trying to say to avoid confusion. ]

[[[[[[[[[[Ben]]]]]]]]]] - [ TAG forcibly re-CC'd. Mahesh, please leave the list in the CCs when replying; not doing so is considered rude. ]

[[[[[[[[[Mahesh]]]]]]]]] - Sorry, ed (warning: I may not be employed to write any public-readable stuff)

I wrote this, because I recently got a job offer from a leading magazine in India, to work as an in-house Linux Technical writer.

[[[[[[[[[[Ben]]]]]]]]]] - Much more of a reason to communicate clearly, then! Now, it's not just an adventure - it's your JOB...

(Not to carry the anvil too far, but some non-USians might miss it: the US Army's long-time advertising slogan was "It's not just a job - it's an adventure." These days, they've changed it to "An army of one" - presumably since anyone signing up these days will get ALL the "adventures" they never wanted.)

[[[Mahesh]]] - I even thought about fitting getopts in to select bold, underline, blink, etc.

[[[[Thomas]]]] - Depends what you want out of your script.

[[[[[Mahesh]]]]] - To help the newbies ease the complications...

[[[[[[Thomas]]]]]] - Pffft. Help them by teaching them, not dazzling their retenas with colours.

[[[[[[[Mahesh]]]]]]] - I was meaning those brainless twits who can't remeber all those escape codes. ("dazzling their retenas" -- is that a problem with my script?)

[[[[[[[[Thomas]]]]]]]] - You seem to be under the impression that your script is somehow lacking and isn't working. That's not true. Your script does work -- for some terminals. When you sent this script in to TAG, did think it was somehow not open for discussion? Constructive criticism such as the advice myself and Ben have been giving you is there to aid you; and not to pick any holes in your script.

[[[[[[[[[Mahesh]]]]]]]]] - Thomas (if I can call you by ur first name), my primary intention of writing the script was to find out the correct colour sequences to colour up the various getty control codes (\l, \m, \n) etc., in /etc/issue and for ${PS1}.

So I can take in any suggestions to make it more useful (beware, I'm working on v2.1 of the same) ...

useful, in the sense, anyone who feels like colouring their /etc/issue or /etc/motd can look up this script and decide on what to do (and what not to do.)

[[[[[[[[[[Kapil]]]]]]]]]] - As far as "motd" is concerned there is no short cut. You need to hack at "login" itself to make sure that it "knows" terminfo or slang in order that it can put out a colourful message of the day regardless of which terminal a user logs in from.

[[[Mahesh]]] - True, tput would have made it portable (a portable shell script???!!!).

[[[[Thomas]]]] - Careful -- the shell script itself already isn't portable (you use 'seq' which isn't on every system, think *BSD for instance), but it's portable in the sense that tput(1) is always guaranteed to work, where hard-coding escape sequences might not.

There's nothing wrong with your script, I was merely pointing out that tput(1) is by far the "better" choice when you wish to use colour for things such as prompts and the like.

[[[[[Mahesh]]]]] - Hey, I think the escape codes are handled by Linux console drivers, so how can they fail to work. At least thats what I understood after reading the man pages.

[[[[[[Thomas]]]]]] - No, they're handled by terminals which read the terminfo and termcap databases.

[[[[[[[Mahesh]]]]]]] - I think I misinterpreted what was in console_codes(4), can u correct me? I thought it said something like the Linux display driver handled the escape codes. Thanks.

[[[[[[[[Thomas]]]]]]]] - No, it's the terminal that interprets them. Note from console_codes(4):

It  is  generally not good practice to hard-wire terminal controls
into programs.  Linux supports a terminfo(5) database of terminal
capabili- ties.   Rather than emitting console escape sequences by hand,
you will almost always want to use a terminfo-aware screen  library  or
utility such as ncurses(3), tput(1), or reset(1).

[[[[[[[[[Mahesh]]]]]]]]] - Thanks, I might have missed it.

And one more thing, you don't mind me using this in my private /etc/issue, right? At least, decapitation is not an option for someone using hard coded escape sequences!

Thanks for ur time...

[[[[[Mahesh]]]]] - does the Linux cosole drivers in anyway deal with the terminfo/termcap facilities?

[[Mahesh]] - Ben, I'm happy I got a mail from LG. Thanks for mailing in..

[Quoting Ben]:

  > colsel () # Color selector - iterates through a $TERM's color choices 
  > { 
  > trap 'reset' 0 # Reset on exit

I didn't think of a trap ... sue me ;-)

[[[Ben]]] - I found that unless I used it, 'Ctrl-C'ing out of the script would often leave the xterm in some weird color mode. NOT a desirable outcome. :)

[[Mahesh]] - [Quoting Ben]:

  > > Well, the standard disclaimer applies. Any suggestions and corrections
  > > are welcome (all flames to /dev/null).
  > 
  > It may already be filled up; that's where we're sending ours...

I have written a script that will empty /dev/null the same way with /tmp at each boot time!!! :-D

[[[Ben]]] - Let me guess:

  # Save the old one - just in case, y'know... 
  ext=0
  while [ -f foo.$ext ]; do ((ext+=1)); done;
  mv /dev/null /dev/null.$ext 

  # Who needs that weird 'mknod' thing, anyway?
  # And by the way, why is my root partition all filled up? 
  touch /dev/null 

:)

[[Mahesh]] - [Quoting Ben]:

    > > > PS: If you want any help, do an "nl -ba colors.sh". ;-) 
    > > 
	> > I've always found 'vi' to be just as useful. :)

But do vi show line numbers (at default)

[[[Ben]]] - It does if you've got 'set nu' in your .exrc file.

[Kapil] - The discussion on pretty-printing of files naturally led to the discussion in the Unix-Hater's book (UHB) on how curses and terminfo are "twisted" technologies. According to the UHB the "right" way is that files should not be "streams of bytes" but should be "objects" which should (in particular) encode all pretty-printing information.

Interesting ... wonder if there is an implementation.

[[[[[[[[[Mahesh]]]]]]]]] - PS: Ben, I shoulda said this earlier: I learnt (well, most of) shell scripting thru a guide someone printed occassionaly in LG (I think his name was Ben Okopnik or something). It was a great guide. So you know whom to blame for my shell mistakes (OK, not escape-sequence mistakes ;) )

[[[[[[[[[[Ben]]]]]]]]]] - [laugh] My sins come back to haunt me. I'm glad that you found it useful, Mahesh (although I specifically remember using 'tput' in that series for "jazzing up" output in programs.)

bogo_router:
 domains = +local_domains
 no_verify
 condition = ${if !eq {$received_protocol}{bogodone} {1}{0}}
 driver = accept
 transport = bogo_transport

bogo_transport:
 driver = pipe
 command = /usr/sbin/exim4 -oMr bogodone -bS
 use_bsmtp = true
 headers_add = X-Bogofilterd: true
 transport_filter = /usr/bin/bogofilter -d /etc/bogofilter -l -p -e -u
 return_fail_output = true
 group = mail
 user = mail
 home_directory = "/tmp"
 current_directory = "/tmp"
 log_output = true
 return_path_add = false

sudo update-exim4.conf && sudo /etc/init.d/exim4 reload

2006-06-02 04:54:13 1Fm0jl-0006a4-8V ** thomas at edulinux.homeunix.org
R=bogo_router T=bogo_transport: Child process of bogo_transport transport
returned 2 from command: /usr/sbin/exim4

An error was detected while processing a file of BSMTP input.
The error message was:

 500 Command unrecognized

The SMTP transaction started in line 0.
The error was detected in line 1.
The SMTP command at fault was:

'exim -d9 -M <message-id>' - turning on max_debug and vgrepping through

OK -- it seems that the "-d9" idea isn't applicable on Exim 4. Looking in its manpage, there seems to be "-d all" which sounded promising, but which just held up Exim and wouldn't do anything after that. So I used the following command:

 sudo exim4 -d -M 1Fn2PM-0007BB-EV 2<&1 | tee /tmp/exim.out

... that message ID is of an email I had sent on the localhost -- the `same error happens from anywhere. I shan't paste the output from that command in here directly. Instead you can see the output here.

[[[Ben]]] - I note a couple of questionable log messages there - e.g.,

internal_search_find: file="/etc/aliases"
  type=lsearch key="n6tadam"
file lookup required for n6tadam
  in /etc/aliases
lookup failed
expanded: 
file is not a filter file
parse_forward_list: 
system_aliases router declined for n6tadam at edulinux.homeunix.org

which may or may not be of concern for you; however, the important bits are obviously toward the end:

calling bogo_router router
bogo_router router called for n6tadam at edulinux.homeunix.org
  domain = edulinux.homeunix.org
set transport bogo_transport
queued for bogo_transport transport: local_part = n6tadam
domain = edulinux.homeunix.org
  errors_to=NULL

That last line should probably be redefined for better troubleshooting.

  domain_data=NULL localpart_data=NULL
routed by bogo_router router
  envelope to: n6tadam at edulinux.homeunix.org
  transport: bogo_transport
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
After routing:
  Local deliveries:
    n6tadam at edulinux.homeunix.org

So far, so good.

  Remote deliveries:
  Failed addresses:
  Deferred addresses:
search_tidyup called
>>>>>>>>>>>>>>>> Local
deliveries >>>>>>>>>>>>>>>>
--------> n6tadam at edulinux.homeunix.org <--------
locking /var/spool/exim4/db/retry.lockfile
locked /var/spool/exim4/db/retry.lockfile
opened hints database /var/spool/exim4/db/retry: flags=0
dbfn_read: key=T:n6tadam at edulinux.homeunix.org
no retry record exists
search_tidyup called
changed uid/gid: local delivery to n6tadam <n6tadam at
edulinux.homeunix.org> transport=bogo_transport
  uid=8 gid=8 pid=28344
  auxiliary group list: <none>
  home=/tmp current=/tmp
set_process_info: 28344 delivering 1Fn2PM-0007BB-EV to n6tadam using
bogo_transport
direct command:
  argv[0] = /usr/bin/bogofilter
  argv[1] = -d
  argv[2] = /etc/bogofilter
  argv[3] = -l
  argv[4] = -p
  argv[5] = -e
  argv[6] = -u
direct command after expansion:
  argv[0] = /usr/bin/bogofilter
  argv[1] = -d
  argv[2] = /etc/bogofilter
  argv[3] = -l
  argv[4] = -p
  argv[5] = -e
  argv[6] = -u

Have you double-checked to make sure that

'/usr/bin/bogofilter -d /etc/bogofilter -l -p -e -u'

actually works? I'd do it with the options spelled out, exactly as Exim sees it.

bogo_transport transport entered
direct command:
  argv[0] = /usr/sbin/exim4
  argv[1] = -oMr
  argv[2] = bogodone
  argv[3] = -bS
direct command after expansion:
  argv[0] = /usr/sbin/exim4
  argv[1] = -oMr
  argv[2] = bogodone
  argv[3] = -bS

Ditto here. Obviously, these two are hooked up via a pipe - so that's the manual process to try.

Writing message to pipe
writing data block fd=10 size=13 timeout=3600
writing data block fd=10 size=40 timeout=3600
writing data block fd=10 size=5 timeout=3600
process 28347 running as transport filter: write=11 read=12
added header line(s):
X-Bogofilterd: true---
process 28348 writing to transport filter
copying from the filter
set_process_info: 28346 reading output from |/usr/sbin/exim4 -oMr bogodone
-bS

Just for grins... what does Exim return when run that way? '$?' should tell you.

writing data block fd=11 size=2252 timeout=3600
waiting for filter process
filter process returned 3
waiting for writing process
end of filtering transport writing: yield=0
errno=-24 more_errno=3

Whoops. I wonder what that means. You might want to Google for it... incidentally, I just noticed that my '/usr/share/doc/exim' directory contains a discussion of filters and how to use them. Again, this is Exim3 - but you might have a parallel document that describes the specifics, and might be worth checking out.

search_tidyup called
bogo_transport transport returned PANIC for n6tadam at
edulinux.homeunix.org
post-process n6tadam at edulinux.homeunix.org (9)
LOG: MAIN
  <n6tadam at edulinux.homeunix.org>: bogo_transport transport
output: An
error was detected while processing a file of BSMTP input.
LOG: MAIN PANIC
  == n6tadam at edulinux.homeunix.org <thomas at
edulinux.homeunix.org>
R=bogo_router T=bogo_transport defer (-24): Filter process failure
2006-06-05 01:17:30 1Fn2PM-0007BB-EV == n6tadam at edulinux.homeunix.org
<thomas at edulinux.homeunix.org> R=bogo_router T=bogo_transport
defer
(-24): Filter process failure
2006-06-05 01:17:30 1Fn2PM-0007BB-EV == n6tadam at edulinux.homeunix.org
<thomas at edulinux.homeunix.org> R=bogo_router T=bogo_transport
defer
(-24): Filter process failure
exim: could not open panic log - aborting: see message(s) above

Hmm, looks like another place where you could get troubleshooting info. Do you have a '/var/log/exim/paniclog' file on your system?

So, I'd run through all of the above stuff and see what shakes out. If you don't get definitive results from the above (and I'm sorta betting on testing the piped progs), then feel free to describe what you've got and we'll bang on it some more. :)

[[Thomas]] - I'm not turning up many hits on Google which sheds much light on this -- despite triple checking the existing filter for bogo_filter matches the general concensus which purportedly works on other's machines; and it does match.

[Neil] - [Quoting Thomas]:

  > 2006-06-02 04:54:13 1Fm0jl-0006a4-8V ** thomas at edulinux.homeunix.org 
  > R=bogo_router T=bogo_transport: Child process of bogo_transport transport 
  > returned 2 from command: /usr/sbin/exim4

Someone else had the same problem. I'm not sure that the answer he got was very useful :-(

http://www.exim.org/mail-archives/exim-users/Week-of-Mon-20030811/msg00247.html

FWIW according to http://bogofilter.sourceforge.net/man_page.shtml return values are "0 for spam; 1 for non-spam; 2 for unsure ; 3 for I/O or other errors.

If both -p and -e are used, the return values are: 0 for spam or non-spam; 3 for I/O or other errors."

It's interesting that you seem to be using -p and -e, but still get a return value of 2.

According to http://exim.org/exim-html-4.62/doc/html/spec_html/ch29.html#id2641142: "If the command exits with a non-zero return code, the delivery is deemed to have failed, unless either the ignore_status option is set (in which case the return code is treated as zero), or the return code is one of those listed in the temp_errors option, which are interpreted as meaning “try again later”."

So the problem would appear to be the return code of 2 from bogofilter, which AFAICS you should not be getting.

[Rick] - [Quoting Thomas]:

  > I'm running Exim4 from Debian stable (which by definition means it's the 
  > Debianised configuration of Exim -- something I am still undecided as to 
  > whether this is a good idea or not).

My middle initial "A" honestly doesn't stand for "Ambivalent", but I have ongoing doubts about that too -- and I rely on it. ;->

  > From my own knowledge about Exim and 
  > how emails are relayed at SMTP time, I think I understand how this all 
  > fits in, but I'll just make sure, as I don't understand why I get the 
  > error I do, and whether there's anything else I ought to be doing. 
  > (Rick, this is surely one for you... :P) 

OK, but I'm a bit of a seat-of-my-pants flier, when it comes to MTA configuration. I've never had a machine set up for experimentation; anything I attempt is on a production mail system, where people call and scream at me if I break anything. Accordingly, opportunities for tinkering have been few, and cautious.

  > I'm not using the whole "split-file" configuration thing which Debian 
  > seems to like. Instead, if I want to make any changes, I just edit 
  > /etc/exim4/exim4.conf.template  .

That's one of the things on which I have greatest ambivalence, actually. The split-file option is best for long-term maintenance, in that the system can incrementally upgrade the parts you haven't touched, while leaving alone others. On the other hand, the single-file option's significantly easier to grok.

  > Now, into that file, I have added two 
  > things. The first thing is the router. I suppose this ought to come 
  > after the DNS stuff. Hence I added: 
    
  >   bogo_router: 
  >   domains = +local_domains 
  >   no_verify 
    
  >   condition = ${if !eq {$received_protocol}{bogodone} {1}{0}} 
  >   driver = accept 
  >   transport = bogo_transport 

Hmm, there used to be a third-party modification for Exim4 called Exiscan (http://duncanthrax.net/exiscan-acl/). I see that it's now merged into Exim4 mainline code -- but you might double-check that your chosen Debian package includes that functionality. E.g., it might be that package exim4-daemon-light doesn't, but exim4-daemon-heavy does.

The Exiscan patch added support for a broad range of content-scanning (virus, spam, other), while Marc Merlin's SA-Exim patch is specialised for SpamAssassin integration and (I think) SA-specific. Exiscan's most recent versions (dubbed Exiscan-ACL) work by adding additional functions to Exim4's DATA ACL, while also providing an entirely new ACL (acl_smtp_mime) called for each MIME part in a message. You might want to look into that approach, rather than attempting to hack transports and routers.

  > One of the reasons I'm not using the Exim-SA package or spamassassin in 
  > general is because it's a pig of a program and reduces poor fluffy (a P166 
  > that acts as the kitchen sink) to a mere crawl.

However, note that one of the key aims of an SA-Exim/SpamAssassin setup is to use SA as little as possible. If you start out (as I did) with J.P. Boggis's EximConfig set of default ACLs / configuration files, most of the heavy lifting is performed not by SA but rather by the Exim4 ACLs that handle the mail first, and reject most of the junkmail.

You might want to start by having a look at the linux bootproces and understanding the role that the kernel fills in this proces.

Recently someone published a nice overview of the boot process on the ibm developerworks site

Additionally I always found the linux from scratch project an excellent resource on the ecosystem that the linux kernel + the (GNU) userland tools & programs really is.

You can find it here: http://www.linuxfromscratch.org/

  > I have 3 questions.
  > 1. Is this approach to understand the linux kernel code all right?

No, see above. Basically the kernel handles access to the hardware resources in an uniform way and offers hooks for user-space programs to get access to these resources. You will find that the login process is handled by a program that calls the kernel to get access to the appropriate resource (filesystem, screen, keyboard, serial line, whatever)

  > 2. If any one can give me some hint where I can find the login related
  > code or where I should start .. I will be very glad.
  > 3. Is there any documentation available which I can use as reference?

See above.

If you have a spare machine lying around somewhere that you can use I can really recommend linux from scratch. It will give you an excellent insight in how what most people refer to as linux is actually build from a number of different components, among them a kernel. It doesn't have to be a high-specced machine, I originally build a linux from scratch on an old intel 386 machine.

Hope you enjoy the ride ;-)

[Lew] - You probably should read the "From PowerUp to Bash Prompt HOWTO" which covers that topic quite well

You might also want to read up on Operating System design and implementation, as the background will help you understand what's going on. A /really/ good book is "Operating Systems: Design and Implementation" by Andrew Tanenbaum and Albert Woodhull (ISBN: 0-13-142938-8, Prentice Hall). You might also want to study how Minix was built (Minix is a unix-like OS built as a teaching tool, and is documented in OSD&I). That might give you a good background on how things work in the Unix and Linux world, and where you might reasonably find the facilities and tools implemented.

  > As at log in prompt I see  <hostname> login:
  > so I tried to grep for the "login:" in .c files. but really could not
  > find the the proper location in the source code.

The login prompt isn't part of the OS kernel. It is actually generated by a user program (getty, iirc). To see it, you would have to grep through the code for getty.c, which I bet you haven't installed (since the userspace app source code is rarely installed on general machines).

  > I have a 2.6.15-01.2054_FC5 source code.
  > 
  > Am very very interested to understand this code. I understand C
  > language quite well.
  > 
  > I have 3 questions.
  > 1. Is this approach to understand the linux kernel code all right?

It is one way to do it. It might not be the best way if you don't already know what Unixish operating systems must do to start up.

  > 2. If any one can give me some hint where I can find the login related
  > code or where I should start .. I will be very glad.

Look at the source code for the getty, login, and passwd user applications. This will /not/ be in the Linux kernel source code libraries.

  > 3. Is there any documentation available which I can use as reference?

See above.

[John] - Just yesterday I found an online copy of the historical book by J. Lions, "A Commentary on the Sixth Edition Unix Operating System". The link can be found under the heading "external links" on the wikipedia page for "lions book".

Although the material is quite dated, it is still highly regarded as a reference for basic unix OS principles, so you might find it of interest.

[[Thomas]] - I still find the Unix Haters book a much "better" authority on the OS. :P

[[[Deepak]]] - Thank you all for your valuable suggestions. After listening from you people i have started building my own Linux system based LFS.

[[[John]]] - Ah, just as I suspected - you are from the dark side ... :)

[[[Marty]]] - I found the UHG

[[[[Thomas]]]] - Was it good? Did you read it? I can't help but think there was more to this email than just that sentence above, initially. :P

[[[[[Rick]]]]] - If anyone hasn't yet read this 1994 collection of Unix discontents, it's definitely worth the trouble, if only because the ranting is sometimes inventive and literate: http://research.microsoft.com/~daniel/unix-haters.html

Dennis Ritchie's comment in his "anti-forward" is worth noting: "You claim to seek progress, but you succeed mainly in whining." That's the lesser problem with the book (particularly in the parts written by LISP Machine and ITS fans whose preferred environments early Unix replaced): It's just a miscellany of sundry people's personal piques and resentments. The larger problem is that those complaints, though they were sometimes reasonable if (mostly) exaggerated at the time they were posted to the unix-haters mailing list -- late 1980s to early 1990s -- just about all seemed already quaint by the time of publication.

In fact, the key ingredient required for fixing those antique problems was open-source licensing and source code access, which was arriving in the form of the BSD family and Linux precisely as the book was being published in 1994.

Chris Browne's review of the now-publicly-available text is here: http://linuxfinances.info/info/unixhaters.html

[[[Marty]]] - I found the Unix haters guide made some valid criticisms, but I knew some of the people on the mailing list -- and found they're vociferousness humorous.

[[[[John]]]] - Yes, most of it does seem to have been written in a humorous vein, some of it very tongue-in-cheek. I haven't yet read it entirely though.

[[[Marty]]] - Their major criticism was "Unix is too complicated" which translates into Doug Gwyn's remarks (which are so true) "GUIs make simple things simple and difficult things impossible".

The criticism boils down to "I only want to do simple things, I can't learn how to do complicated things". If the GUI doesn't let me, I don't need to...so this really means "I'd rather spend a week doing writing code to do in awk/sed (back then) in 5 minutes..."

[[[[John]]]] - A lot of it was written in the early days of history of the OS family, with several mentions of Lisp Machines, etc. It doesn't seem to reflect a lot of the advances made during the past 12 years or so:

    Did users want the operating system where bugs didn't get fixed? Not
    likely. Did users want the operating system with a terrible tool set?
    Probably not. Did users want the OS without automatic command
    completion?  No. Did users really want the OS with a terrible and
    dangerous user interface?  No way. Did users want the OS without memory
    mapped files? No.  Did users want the OS that couldn't stay up more
    than a few days (sometimes hours) at a time? Nope. Did users want the
    only OS without intelligent typeahead? Indeed not. Did users want the
    cheapest workstation money could buy that supported a compiler and
    linker? Absolutely.  They were willing to make a few sacrifices.

[[[Marty]]] - BTW, the latest WSYSIWIG editor I enjoyed using was BravoX on the altos.