[02:14:49] <jmkasunich> http://linuxcnc.org/compile_farm/emc2head_slot4.txt
looks like the vti driver wasn't added properly for the 2.4 kernel build system
I think I found it
do you have a box that is always on and can send mail?
yes, but it's not linux
does it have bash and wget?
I'm working on a script that reads the compile farm status and prepares an email if it fails
would you consider making it a cron job?
hey I bet sourceforge machines can do that
ok, I'll mail it to you when its done
do we have shell access on a sf machine?
I think so
but I bet sending mail is disallowed
I just tried sending one...
it worked fine
you shelled into sf?
damn, that would have taken me at least a half hour of searching and man-page reading
it was a lucky guess
if only I could fix dishwashers
nothing except our water supply
(I fixed our dishwasher twice ;-)
the nozzles are full of calcium "flakes"
out of our terrible water
so no matter how I painstakingly clean them, they plug immediately
try that "CLR" stuff?
today I tried vinegar, which sort of worked
About the Common Language Runtime (CLR)
aha CLR cleaner
no, I hesitate to put anything in it that's not a food product...
dishwasher soap isn't a food product
well that's true
I guess as long as it's soluble
I think its an acid that disolves lime and calcuim
never used it myself, I'd certainly read the label before buying, but it might work
I have a gallon of muriatic... I was really tempted to try it.
it would probably eat the plastic though.
thanks, I'll try it (or something like it)
were my changes to the halcmd man page ok?
(I was just thinking we need an emc man page)
I think the changes are OK
I only read the commit message, and with the markup and diff its hard to tell
it also gets installed with the latest packages
heh, for about a year, installing that one man page was the only thing "make install" did
that seems not-so-useful...
rip was all I cared about, install was an afterthought
I'm really pleased with the packaging and the easy update ability it gives the users
yeah, that is nice
as a user, I love it, debian (and now ubuntu) are very easy to add new things to
when we're using it, though, we sure have to make sure that each of our releases is better than the last...
need something, run synaptic, search, find a package, install it, done
I guess a user can just as easily go back a version though
regression testing would be nice
I just realized, that problem in 2.4... its in Testing
we didn't check the compile farm before moving the tag
if you think anyone will use TESTING on 2.4, just move the tag up
if you don't, it's probably best to leave it for next time.
well, it only works on BDI-Live anyway, the older 2.4 BDIs have other issues
I put "check farm" at the top of my notes
had to take the dog out
while I was waiting for him, I was thinking about this script
instead of reading the results in a cron job that isn't synced to the actual compiles,
could the farm ssh into the SF box and send a mail?>
that would be faster, by up to an hour
and simpler than parsing the farm results page looking for changes
sure it could
how would I ssh into SF and send a mail?
ssh firstname.lastname@example.org 'echo help me. sincerely, the farm |mail email@example.com'
err ssh firstname.lastname@example.org ...
you have to set up ssh so each of your farm machines can ssh to shell.sf.net without a password.
they can ssh to cvs without a password, hopefully the setup for shell is similar
yes, it's already done then
echo help me | mail.... seems to be missing some quotes?
echo "help me" | mail?
sure, that'll work too
echo prints all its arguments by default, so quotes are optional
you could also put a file containing your email on shell.sf.net
then ssh shell.sf.net 'mail the-list < the-file'
hmm, it wants a password, and my normal SF password doesn't work
hmm, it didn't for me
maybe you need to enable shell access in your sf profile?
let me dig in mine
I do see some shell-related stuff
on My Page >> Preferences
I see a place to set your shell
does it say Project shell server: shell.sourceforge.net?
shell is set to /bin/bash
does your local login match your sf username?
did you ssh email@example.com
my local name is john
my SF name is jmkasunich
in your local .ssh/config you can put:
john@ke-main-ubuntu:~$ ssh firstname.lastname@example.org 'echo "testing" | mail email@example.com'
The authenticity of host 'shell.sourceforge.net (22.214.171.124)' can't be established.
DSA key fingerprint is 4c:68:03:d4:5c:58:a6:1d:9d:17:13:24:14:48:ba:99.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'shell.sourceforge.net' (DSA) to the list of known hosts.
Permission denied, please try again.
then you don't have to worry about it anymore
it pulled "jmk" out of somewhere
you typed it :-)
before the @
still asked for a password
can you at least log in thougH?
thats because I don't have a ssh key for the ubuntu box
easy enough to fix
it sent the mail
yeah, but I haven'
haven't bothered - this ubuntu install is still a trial run
when I do it for real it will be on the larger disk
I thought that about mine too, but as I'm getting more things settled how I like them, it's becoming permanent.
the farm slots all have ssh keys set
gonna have to read man mail, it came thru with no subject
the hard parts are done then
mail -s "the subject" < the-file
I'm really suprised that SF lets you send mail
why? they know exactly who is sending
ok, now I need to modify the farm scripts to send mail on failure
it's not going to send every hour until something is fixed, is it?
... one from each machine?
but only for some trees (emc2, not emc or rcslib) and only some slots (slots 2 and 3 fail every time, until/unless we make the build system work with gcc 2.xx)
no, it will send after every failed compile
but it only compiles if there has been a commit in the last hour
right, I forgot that detail
I think I'll have it send to me only for the first few days, just in case
should it send to -devel or -commit?
21:25:51 < ipfw> ubuntu is kinda gay.. How do I get it to install the darn devel tools ?
everyone who commits code (and might break things) SHOULD be subscribed to commit
there are a lot more subscribed to devel
well over 100 I think, compared to maybe 20-30 on commit
I think ray has said he's not subscribed to commit
gonna take a look now
(btw, do you have the list admin passwords?)
I think so
shell.sf.net is nice - one could do useful work on a unix project even with only access to a windows machine, or a public machine
43 addys on commit, but there are quite a few that have two for one person
their SF addy (required so it will accept commit messages) and their preferred one
delivery turned off to the SF addy
ray has all delivery turned off
maybe you should use -devel then
-devel has 194 members
dunno if I want to spam all those folks
ideally you might email only the person who made the change, and include the errored compile output.
the likelyhood of ray changing any C or C++ code to cause a compile failure is pretty slim
I will include the URL of the output
do you know the committer?
forget it, that'd be too hard
and in fact there could have been multiple commits by multiple people in the last hour
besides, if it goes to the entire commit list, and one person has mail turned off, the others can yell at him
now if only it didn't have my name attached...
I think you're not going to get around that
create a SF user "emc-compile"
probably against their rules
why bother? it's your farm
yeah, but I don't want to be thought of as the guy who sends nagging messages
its not important
there is another annoying farm thing that I want to try to fix
sometimes (probably a race condition when two slots finish a compile at the same time) the index.html file gets lost
(winds up zero length on the linuxcnc server)
in a case like that, use the fact that mv is atomic
copy to some temp file (containing the slot number?) and then mv it to the final filename
ftp has a mv?
I don't think so... you'd have to use ssh
linuxcnc.org is a doze box
on the machine that creates index.html: blah > index.html.$$; mv -f index.html.$$ index.html
nobody "creates" it
jepler: I just said that...
it is there, each slot does an ftp get, sed's in its info, and ftp put
what creates index.html? software on a windows box?
the original was created by hand
jmkasunich: does the windows machine have server-side includes?
(and is sometimes edited by hand, for instance to insert a note that the farm will be down for a period)
nothing fancy that I know of
that's not very fancy...
I have ftp write access to a dir and thats about it
you should try a SSI
ftp does have 'rename', doesn't it?
RNFR + RNTO
[03:40:37] <jmkasunich> http://cvs.sourceforge.net/viewcvs.py/emc/emc_HAL/farm_scripts/run_farm?rev=1.3&view=auto
SSI format is <!--#include file="some-file.txt" -->
if you had that, each farm slot could write its own file
oh god it's microsoft IIS
(I just made that up)
yeah IFRAMEs would work
echo "get index.html old.html" >>ftp_cmds
echo "! cat old.html | sed -f sed_cmds >index.html" >>ftp_cmds
echo "put index.html" >>ftp_cmds
thats what it does today
get, sed, put
(all in one invocation of ftp)
if I could run the sed on the server... but that is just as impossible as server side includes or any other server side tricks
that'll actually work
is it MS specific?
Yeah, but will it work on IE4?
<IFRAME> is an HTML tag
State of slot 1: <IFRAME SRC="slot1.txt">
[03:45:29] <cradek> http://www.microsoft.com/windows2000/en/server/iis/default.asp?url=/windows2000/en/server/iis/htm/core/iisiref.htm
* jmkasunich googles iframe
some unknown version of IIS does actually have SSI
so you should try it first
it's the easy way that will work for all browsers
I'd rather keep things as generic as possible
do try a SSI first
the existing way is as generic as you can get
iframes may be finicky on some old browsers.
I always feel like a dinosaur when talking to you guys
SSI, IFRAMES, all kinds of stuff
whats wrong with get/sed/put
the race condition
what confuses me is how does the file get set to zero length?
I can understand losing one slots updates
I'm not sure either
slot A gets, slot B gets, both sed, slot A puts, slot B puts, slot A's changes are lost
I'd guess it's slots A & B try to put at the same time, and microsoft ftp pukes
if ssi works, each slot would put to its own file and the main page would include them, and you'd never touch the main page
right, I can see the benefit
ok, lets try it
I'm editing a test.html that I'll put up there
what needs to go in it?
the usual header/footer, plus <!--#include "otherfile.txt"--> somewhere in the body
SSI format is <!--#include file="some-file.txt" -->
then put a short message in otherfile.txt in that same directory.
the contents of otherfile.txt just appear inline?
the server replaces that comment with the contents of the file
can I embed it in a table or do other fun stuff?
no restrictions, it's a simple insertion
(right now, the word passed or failed is a link to the results file)
ok, lets try this
what happens if otherfile.txt doesn't exist?
I'm not sure
you might get an error, or the comment will be unmodified
The file containing the #include directives must use a file name extension that is mapped to the SSI interpreter; otherwise, the Web server will not process the directives. By default, the extensions .stm, .shtm, and .shtml are mapped to the interpreter
looks like you have to make it test.stm, not test.htm
not plain old html
stm, or shtml?
it's the same as html, but the extension tells the server to look for these SSI directives
do you have long file names? shtml then
I assume so , the existing one is index.html
"404 Object not found"
[03:57:18] <cradek> http://linuxcnc.org/test.shtml
the include files aren't there yet
that is next
<a href="emc2head_slot2.txt"> <!--#include "emc2head_slot2_status.txt"--></a>
not lookin' good so far
but maybe it'll work when you put a file there
on my box, the "results" column is just blank
yeah but in the source, it shows the comment
yeah, its not working
put the status file there, still nothing
I can't tell on this microsoft.com page what IIS version they are document
my (rather lame) solution to the problem was gonna be:
if length = 0, ftp put the most recent copy
lame but working is better than elegant and unworkable
(thats when I was gonna be wgeting the file anyway, to detect compile failures)
I found another page that says IIS 4.0 *does* do SSI
wait a second
I must have told you the wrong format
I told you right the first time, wrong the second time
delete the !, right?
you are missing the file=
I see the little !, but miss the big ole "file="
I left it out once or twice
I can't write a line of code right the first time, in any language, to save my life
duh, I don't want to use slot 2 for all four
ugly error message if the file is missing
but it won't be missing soon
this is a great improvement
that sed fu was yucky
dunno whether I should have the script create the individual files on the server, the put them
or have two permanent files passed.txt and failed.txt, and do put passed.txt emc2head_slot2_status.txt
(or put failed.txt emc2head_slot2_status.txt)
uh, oh, just realized something
it isn't just passed/failed
my sed fu was also changing the date
just use another file for the date?
date >file; ftp put file
theres yet another place where this could be approved...
the history file
each time, I append a line to a local file, then put the whole thing
there is an ftp append command, instead of uploading the entire file I could update the line
not sure how you'll fix that...
I obviously don't know ftp
well there sure is
and it just so happens that the line consists of the date, and PASSED/FAILED
I don't know how to use it, just did help and there it was
I never ever use (non-anonymous) ftp because of the cleartext password problem.
heh, this is a doze box, that is probably the least of the security holes
if I play it right, maybe the same file that contains the date and result (and is included into index,html) can be appended onto history.txt
11:20, gotta work tomorrow
too late to dive into this stuff
I need to do a fairly significant revision of the farm scripts, to use the includes and to mail out failure notices
too much for tonight
[Global Notice] Hi all. Just a reminder: #freenode-groups has been set up as a social channel for folks with primary space group cloaks. Work on freenode-registry, which will replace services and provide automated functionality for group registration, continues. If you don't currently have processed group registration and cloaks, please be patient. Thanks!
[Global Notice] A new network news article is up for 20 February 2006....topics: group registrations, freenode staff channels, fundraiser. Have a great morning!
[Global Notice] Hi all. Just a reminder: news article up at http://freenode.net/news.shtml
.... group registrations, staff channels, 2006-2007 fundraising .... have a great evening, and thank you for using freenode!
new emc2-axis package in the Ubuntu repository
I know it's not a supported configuration, but I am going to try emc on a kernel with CPU frequency scaling enabled
I'm going to modify the 'realtime' script to change the governor to 'performance' (top CPU speed) with 'realtime start' and the previous setting with 'realtime stop'
does anyone think I shouldn't commit a change like this?
argh. changing that would require 'sudo' or another setuid helper. :(
OK, I won't commit it