IRC chat logs for #ltsp on irc.libera.chat (webchat)


Channel log from 28 January 2009   (all times are UTC)

00:21
<Ryan52>
warren, http://slexy.org/raw/s21Y6ZW5uI
00:22
warren, since I can't look at the archives of anything, which lists should I actually sign up for? (I really don't want to add to my daily email PITA ... I already get more than a hundred emails a day, which is rediculous..)
00:22* Ryan52 is already on fedora-announce-list, and has been for a while..
00:24
<warren>
Ryan52: fedora-devel-announce is the minimum
00:24
Ryan52: fedora-devel-list is good if you want to follow the main discussions
00:24
Ryan52: do you do server side filtering of your mail into folders?
00:25
<Ryan52>
no. I do use colors in mutt, and I look at my INBOX via regex.
00:26
and it works well enough for me for now...once I get a sane email provider (that I can use real filtering on...not gmail's crap), I'll use folders and stuff.
00:27
so what does "main discussions" mean?
00:29
does that include useless trolling about a stupid joke posted to an announcement list? if so, then no, I'm already on Debian's equivelant, which is enough.
00:29
<warren>
Ryan52: most everything that matters
00:29
Ryan52: fedora-devel-announce has only important dev announcements, and fedora-devel-list is dev only discussion... but then there's huge flamewar threads about stupid things that nobody cares about.
00:30
I typically just set those entire threads to ignore, then it is quite readable.
00:30
<Ryan52>
okay, I'll subscribe...I can always unsubscribe.
00:30
<warren>
Ryan52: https://fedorahosted.org/k12linux/ then there's tiny lists here
00:32ball has joined #ltsp
00:46
<warren>
Ryan52: if you have ltsp-trunk in one place, and /cvs/pkgs/ltsp checked out in a different directory...
00:46
Ryan52: ln -s /path/to/ltsp/F-10/ltsp.spec /path/to/ltsp-trunk/ltsp.spec
00:46
Ryan52: cd /path/to/ltsp-trunk/
00:47
Ryan52: edit ltsp.spec to match the Version of release.conf
00:47
Ryan52: mkdst rpm
00:47
Ryan52: koji build --scratch dist-f10 ltsp-5.1.52-2.20090128.01.fc10.src.rpm
00:48
Ryan52: really quick unofficial builds of whatever is in the ltsp-trunk directory (even uncommitted)
00:48
<Ryan52>
why would I use koji for that?
00:49
<warren>
Ryan52: you can do test builds of a .src.rpm against all archs simultaneously
00:49
<Ryan52>
ah, cool.
00:49
ugh. koji has no man page :(
00:49
<warren>
then go to http://koji.fedoraproject.org/scratch/username/
00:49
Ryan52: you don't normally use the koji command directly, it is wrapped by other commands
00:49
<Ryan52>
ah.
00:51
<warren>
Ryan52: anyhow, if the build succeeds you can go to that URL to grab it
00:51
Ryan52: it it fails, koji gave you a URL to browse logs
00:52
<Ryan52>
make tag && make build
00:52
is that all you have to do to get a package uploaded to Fedora itself?
00:52
<warren>
yes
00:52
it builds whatever is tagged
00:53
<Ryan52>
wow, that's really nice.
00:53
<warren>
Ryan52: don't do that yet... will train you more tomorrow.
00:53
<Ryan52>
I know.
00:53
<warren>
Ryan52: you don't even have to remain online after make build. the server received the command. you're only watching output
00:53shamino has joined #ltsp
00:54
<Ryan52>
nice
00:55
<warren>
http://koji.fedoraproject.org/scratch/wtogami/ m
00:55
my latest scratch build succeeded
00:55
they are cleaned up after a week or so
00:57
Ryan52: how is submitting builds to debian done?
00:59
<Ryan52>
dch -r && debuild -S -sa && cd .. && cowbuilder --build PACKAGE_VERSION.dsc && debsign /var/cache/pbuilder/result/PACKAGE_VERSION_ARCH.changes && dput /var/cache/pbuilder/result/PACKAGE_VERSION_ARCH.changes
00:59ball has quit IRC
00:59
<Ryan52>
:)
01:00
does some stuff, builds the package in a chroot, asks you for your gpg passphrase twice to sign it, and transfers it via ftp.
01:02
<warren>
cowbuilder builds locally?
01:02
<Ryan52>
yes, in a chroot.
01:02
<warren>
We find it a little nutty to distribute packages that you built yourself
01:02
Ryan52: why is it called cow?
01:02
<Ryan52>
copy on right
01:02
<warren>
copy on write?
01:02
oh
01:02
<Ryan52>
ya, that one :p
01:03
that way one evil build doesn't muck up your clean chroot.
01:11
<warren>
how does it do copy on write?
01:11
aufs?
01:12
Ryan52: our buildroots are built from scratch from RPMS using the entire BuildRequires chain for every package build.
01:18
Ryan52: hmm, I need to do a major release with basically ldm-2.0.30 but I need one tiny fix. With the translation stuff incomplete I can't tag the current..
01:18
I might just patch the 2.0.30 to release
01:18
<Ryan52>
I have no clue what it does for the copy on write.
01:19* warren waiting for SSD's to drop another 30% in price before buying....
01:21
<Ryan52>
warren, ya, sorry.
01:21
<warren>
Ryan52: I appreciate that you're working on that, it was always ugly
01:22
<Ryan52>
doesn't fedora have some form of a base system? or do you have to put *everything* in BuildRequires? bash, coreutils, etc too?
01:23
<warren>
Ryan52: rpmdevtools Requires the base system that you can assume will be installed in the chroot
01:23
Ryan52: stuff like coreutils and gcc are assumed, but autotools are not
01:24
<Ryan52>
ah. ok. the cowbuilder chroots are just the base system (base, coreutils, etc) plus build-essential (dpkg-dev, make, gcc, etc)
01:24
do you actually have to put rpmdevtools in BuildRequires or is it assumed to be there?
01:25
nevermind, figured it out myself :p
01:25
<warren>
crap, it doesn't seem to be working with s/file/sha1sum/
01:26
hmm
01:28alkisg has quit IRC
01:32shrek has quit IRC
01:36
<warren>
hmm
01:36
gettext.sh
01:37
where does that come from?
01:37
<Ryan52>
gettext? :)
01:38
$ dpkg -S `which gettext.sh`
01:38
gettext-base: /usr/bin/gettext.sh
01:39
<warren>
something broke ldm on my system
01:39
related to those gettext changes
01:39
+export TEXTDOMAINDIR=/usr/share/locale
01:39
+. gettext.sh
01:40
<Ryan52>
it runs ". gettext.sh"?
01:40
<warren>
ugh
01:40* Ryan52 wonders why..
01:40
<warren>
that adds another 5MB to the package
01:40
oh wait
01:40
no
01:40
more than 5MB to chroot
01:41Egyptian[Home] has quit IRC
01:41
<warren>
Ryan52: I think stgraber might have been confused. There has to be a way to get strings out in a shell script without installing gettext.
01:42
gettext is used in development, not runtime
01:42
<Ryan52>
really?
01:42
<warren>
yes
01:42
<Ryan52>
not what I read, iirc.
01:42
<warren>
my chroot doesn't have gettext
01:42
yet it can display translated stuff from .mo files
01:42Egyptian[Home] has joined #ltsp
01:43
<Ryan52>
and your chroot is bilingual?
01:43
oh.
01:43
<warren>
yes
01:43
you use gettext to pull strings out of source to into .pot files
01:43
<Ryan52>
then why do they call the C/python functions gettext?
01:44
that's gonna confuse people :)
01:44
<warren>
Ryan52: look in fedora's initscripts, all the startup and shutdown messages can display translated
01:44
<Ryan52>
warren: http://www.gnu.org/software/automake/manual/gettext/Preparing-Shell-Scripts.html
01:44
<warren>
Ryan52: with no gettext package installed
01:44
<Ryan52>
the official docs agree with what stgraber did.
01:44
<warren>
the gettext packageis huge
01:45
ok, well there is a better way to do this.
01:45
Ryan52: fedora's initscripts package contains .mo files and can display traslated boot messages, but no gettext needed.
01:45
Ryan52: rpm -ql initscripts
01:46
Ryan52: does debian's initscripts do something similar?
01:47
<Ryan52>
echo "Error: argument '$1' not supported" >&2
01:47
log_action_msg "Will now restart"
01:47
that's Debians...
01:47
looks real i18nized...
01:48
action $"Stopping disk encryption for $dst" /sbin/cryptsetup remove "$dst"
01:48
<warren>
yes, same thing here.
01:48
<Ryan52>
where action is just doing: echo -n "$1 "
01:48* Ryan52 wonders how that's i18nized.
01:48
<warren>
I'm not sure how the heck that works..
01:48
but it does translate here
01:51
<Ryan52>
yay for insane black magic that does cool stuff! :)
01:52
gettext-base is: Priority: standard
01:52
in Debian, meaning that it's on most systems. (other than minimal chroots).
01:52
<warren>
I am not installing that in my chroot only to translate a single shell script
01:52
it is 5MB here
01:54
Ryan52: hm... action seems to only colorize?
01:55
<Ryan52>
yes, which is why I think that this boot process can't possibly be i18nized.
01:55* warren boots in french mode...
01:56
<Ryan52>
. /etc/profile.d/lang.sh
01:57
looks relatedish.
01:58
but still nothing..
01:58
<warren>
I just booted in french mode and everything was translated
01:59* warren removes gettext just to be sure
02:00
<warren>
Ryan52: initscripts doesn't have the word gettext in it anywhere
02:00
<Ryan52>
I know.
02:01
<warren>
yes
02:01
removed gettext from chroot
02:02
and it is still translated
02:02zirconiumks has quit IRC
02:02
<warren>
Ryan52: btw, l10nized is a better word
02:02
Ryan52: l10n localization
02:02
Ryan52: i18n refers to internationalization ... generally software stuff to support other languages that are not translation
02:03
<Ryan52>
ok.
02:04
already 4 emails on the devel list...geez.
02:04
<warren>
you seriously need server side filtering into boxes
02:05
ok... this gettext thing happened after 2.0.30, so I can still release with ldm-2.0.30 and the next ltsp tag
02:05
sleep....
02:06pmatulis has quit IRC
02:06pmatulis has joined #ltsp
02:22polytan has joined #ltsp
02:23
<polytan>
hi
02:23
I'm in trouble with ltsp and ubuntu
02:24
I build a client image which runs very well on core 2 duo computers
02:24
(i386)
02:25
but on the computer I'm now working (pentium D), I have nothing after the splash screen displayingh a progression bar
02:25
and I have this error message ; "INFO: task khelper: 2031 blocked for more than 120 seconds.
02:25faustino3331 has quit IRC
02:26
<polytan>
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disable this message"
02:26
but I can do nothing :(
02:27
<Ryan52>
warren, does fedora have equivalents to http://tinyurl.com/bckamp and http://tinyurl.com/b3phtf ?
02:30
<polytan>
johnny, hi, are you here ?
02:31
<johnny>
hi
02:31
maybe it's a bug..
02:31
why are you asking for me?
02:33faustino333 has joined #ltsp
02:36
<polytan>
johnny, I'm asking nithing to you about that
02:37
I've stopped to use gentoo to do ltsp
02:37
it is too young for me
02:37
but I've solved a few bugs and I think you will be interested
02:52shrek has joined #ltsp
02:54
<shrek>
Hi! Anyone one on LTSP 5 modified for NFS on Ubuntu 8.10?
02:54
facing an issue with GDM
03:11
+o
03:12
<Ryan52>
shrek, you mean ldm, right?
03:12
shrek, anyway, you have a better chance of getting help if you say what your problem actually is.
03:15chrisinajar has quit IRC
03:15
<shrek>
no
03:15
sure
03:15
<warren>
Ryan52: dude you really should sleep
03:15
Ryan52: we're building something like that, it isn't ready yet
03:15
<shrek>
sorry
03:16
actually i have a mixed environment
03:16
<warren>
Ryan52: all that data is scattered but available now
03:16
<shrek>
2 chroots
03:16
one i386 for thin-clients and one fati386 for fat-clients
03:16gate_keeper_ has joined #ltsp
03:17
<Ryan52>
warren, ok.
03:17
<warren>
Ryan52: dude, sleep
03:17
<shrek>
Ryan52: one important thing - both of them have been moved out of /opt onto /home and sym-linked back onto /opt
03:18
<Ryan52>
warren, I know, I'm gonna go to sleep soon.
03:19
<shrek>
have followed https://help.ubuntu.com/community/UbuntuLTSP/LTSPFatClients (workstation plugin and not the fatclient one) to set it up
03:19
<polytan>
Is it serious a "request_module: runaway loop modprobe binfmt-464c" ?
03:20
<shrek>
i386 works fine for thin-clients with nbd
03:20
but fati386 doesn't
03:20
hence i moved to nfs for fati386
03:20
now all goes fine until it tries to start gdm where is goes bust
03:22
it gives the following error - Server Authorization directory (daemon/ServerAuthDir ) is set to /var/lib/gdm but is not owned by user 108 and group 117.
03:22
108 and 117 are the uid and gid of gdm
03:22
in the base as well as the chroot system
03:23
"/var/lib/gdm" is also owned by gdm:gdm
03:23
don't know where to go from here
03:38
<polytan>
I've rebuild the image and it boots
03:38
yep ;)
04:15
<shrek>
is there a max size that the nbd image should be within?
04:16
my fat-client image has grown to over 1 GB and while trying to boot i get an error " Gave up waiting for root device....."
04:17Shingoshi has quit IRC
04:17Shingoshi has joined #ltsp
04:21Shingoshi has quit IRC
04:22Shingoshi has joined #ltsp
04:28
<polytan>
can we put the "rrot-path" option in pxelinux.cfg/default instead of dhcp.conf ?
04:28
like APPEND root-path=/opt/ltsp/i386 ?
04:37alkisg has joined #ltsp
04:41
<Appiah>
may I ask why you need to move it from default to dhcp.conf?
04:44
<polytan>
because on my sever, I don't only want to boot ubuntu and ltsp
04:44
I have to deal with i386, amd64, centos, opensolaris and others
04:44
<alkisg>
polytan: I thought about that some weeks ago, it would be very cool. A few lines in the initscipts would need to be modified to support it. I'd be happy to do it if the developers would be willing to consider including it upstream.
04:44
<polytan>
and the argument root-parth=/opt/ltsp/i386 is a big limitation
04:45
alkisg, actually, I just removed the line "root-path" from the dhcp and it runs
04:45
<Appiah>
multiply images on tftpserver
04:45
<polytan>
modifying the "default" it works perfectly
04:45
<Appiah>
think there was a big thread on it on the mailinglist
04:46
<alkisg>
root-path is for nfs. I'd like to also be able to pass the boot-filename, because clients use it to work out the tftp path for lts.conf.
04:46
<polytan>
http://pastebin.com/f5993f2af
04:47
and http://pastebin.com/f9387620
04:48
<alkisg>
Can you use the second one with nfs clients?
04:48
<polytan>
I just have to check for amd64 machines if it works
04:48
I will have 64bits machines tomorow
04:48
alkisg, do you mean the "win" one or the "64" ?
04:49
<alkisg>
Ah, this one: http://pastebin.com/f9387620
04:49
Does it work with e.g. debian clients, that use nfs?
04:49
(I mean debian server, of course! :) :) )
04:49
<polytan>
I don't know why this will not work
04:50
it is quite generic options for dhcp, I believe :)
04:50
<alkisg>
Because I think nfs uses root-path to locate the files
04:50
So if you don't specify root-path, they won't be able to boot
04:50
<polytan>
don't you think it is in the initrd ?
04:50
I'm nearly sure that we can specify the nfs option using tftp and not dhcp
04:50
<alkisg>
root-path? I don't think it's hardcoded...
04:51
<polytan>
04:51
who gives the root-path parameter ?
04:51
<alkisg>
But with what distro?
04:51
<polytan>
I don't really know I have to admit ...
04:51
I'm on ubuntu
04:51
<alkisg>
Ubuntu doesn't use it at all, unless you switch from nbd to nfs
04:52
For example, now (=with IPAPPEND 3) your ubuntu clients won't be able to get lts.conf, because they don't know the path (they usually get it from boot-filename)
04:53
<polytan>
ok
04:54
<alkisg>
So in order to use IPAPPEND 3 and be able to use an lts.conf file, you either have to (1) put it in /var/lib/tftpboot/, or (2) patch the initscripts
04:54
(ltsp_nbd, specifically)
05:08kalib has joined #ltsp
05:10
<kalib>
Hi guys... My ltsp is working fine.. But I'm having a problem with serial mouses on the stations. What can I do?
05:18
<alkisg>
kalib, distro?
05:20
<kalib>
edubuntu
05:21
ubuntu 8.10 + edubuntu patch
05:21
<alkisg>
Did you declare them as serial in lts.conf ?
05:21
At least in 7.10 they weren't autodetected...
05:22
I changed mice after that, I don't know if that's still the case.
05:22
!lts.conf
05:22
<ltspbot>
alkisg: "lts.conf" is http://wiki.ltsp.org/twiki/bin/view/Ltsp/LtsConf
05:25Guaraldo has joined #ltsp
05:26
<kalib>
alkisg, [labinfo2-19]
05:26
X_MOUSE_PROTOCOL = "Microsoft"
05:26
X_MOUSE_DEVICE = "/dev/ttyS0"
05:26
X_MOUSE_RESOLUTION = 400
05:26
X_MOUSE_BUTTONS = 2
05:26
X_MOUSE_EMULATE3BTN = Y
05:27
<alkisg>
Ah, and it still doesn't work? Are you sure that the clients does indeed get the settings from lts.conf ?
05:27
E.g. you could run getltscfg -a from a terminal running locally on the client
05:28
<kalib>
alkisg, I'm not sure if they're getting the conf file.. I'll try your command...
05:28
;]
05:28
<alkisg>
(Also, I think that my mice didn't work with the "Microsoft" protocol, but I can't remember if I left it empty or if I specified "Logitech")
05:29
<kalib>
good tip.. I'll try it
05:29
;]
05:30
just a minute..
05:30
<Guaraldo>
kalib: :-/
05:31
<kalib>
Guaraldo, o.O ?
05:31
<Guaraldo>
kalib: :-D
05:39neoshrek has joined #ltsp
05:42polytan has quit IRC
05:43
<redspike>
Hi all, i have problems with the fonts on my clients, some terminal applications will not start, and i have read that i need to add :unscale in the font apth in xorg.conf on 75 and 100dpi fonts, where i add that so the clients gets that settings?
05:44neoshrek has quit IRC
05:49polytan has joined #ltsp
05:53dirigeant has joined #ltsp
05:57shrek has quit IRC
06:07hanthana_ has joined #ltsp
06:10
<kalib>
alkisg, alkisg still there?
06:10
<alkisg>
yeap
06:10
<kalib>
When I tried to run your command... I got a message asking me to install ltsp-client-core
06:11
<alkisg>
kalib: ah, you should logon to the client *locally* to issue this command.
06:12
This is a little difficult, you have to e.g. put SCREEN_02=shell in lts.conf or unlock the root account in the chroot...
06:12hanthana has quit IRC
06:13
<alkisg>
There's a how-to in the ubuntu-wiki to see how to run commands locally
06:13
<kalib>
I c.. let me try. ;]
06:14
but... I the terminal is not reading the lts.conf, why should I put this in lts.conf? o.O
06:14
but... If the terminal is not reading the lts.conf, why should I put this in lts.conf? o.O
06:14
<alkisg>
But you'll make sure that they don't! :)
06:14
...so try the second option, to unlock the root account: https://help.ubuntu.com/community/UbuntuLTSP/ClientTroubleshooting
06:15elisboa has joined #ltsp
06:21plamengr has joined #ltsp
06:21
<kalib>
I'll check the link
06:23chupacabra has joined #ltsp
06:23
<chupacabra>
anybody awake?
06:27hanthana_ is now known as hanthana
06:44
<Appiah>
alot awake
06:44
alot of people are*
06:44
O_o
06:48pasmen has joined #ltsp
06:50
<chupacabra>
could someone try to hit topstexas.com a client that seems to have lost dsl.
06:50
I'm trying to not have to go there right away.
06:52
<laga>
doesnt load here
06:52
<chupacabra>
dern. thanks.
06:53
rules out my network.
06:53
I'm going in....
06:54BrunoXLambert has joined #ltsp
07:00polytan has quit IRC
07:09polytan has joined #ltsp
07:14mistik1 has quit IRC
07:16Egyptian[Home] has quit IRC
07:18mistik1 has joined #ltsp
07:18Egyptian[Home] has joined #ltsp
07:21
<polytan>
johnny, do you want my modifications ?
07:26cliebow has joined #ltsp
07:38Gadi_eeepc has joined #ltsp
08:05
<alkisg>
Can someone please clarify something for me? Would it be possible for a (specifically designed for this) application to connect and display video to e.g. 10 different X servers simultaneously? Or I didn't understand at all how X works? :)
08:11
<Gadi_eeepc>
alkisg: I believe you would need an Xproxy (like VNC)
08:12
<ogra>
VLC can work in a broadcast mode afaik
08:12
<alkisg>
Gadi_eeepc: not to display the same thing, just to be able to connect to more than 1 server
08:12
<Gadi_eeepc>
a single Xclient can connect to only one Xserver or Xproxy
08:13
in the case of VLC, the VLC clients eaach connect to an Xserver
08:13
and each receive streams from the VLC server
08:13
but, the VLC client is the Xclient
08:14
make sense?
08:14
<alkisg>
So a C program can't declare more than one "Xclient structures/callback functions/whatever" and be an Xclient for more than one server, right? :(
08:14
Pitty, it would be awesome if it could
08:14
<Gadi_eeepc>
no, but it can connect to seveeral client apps each of which connects to a different Xserver
08:15
<alkisg>
Ah, you're right. That won't be hard!
08:15
<Gadi_eeepc>
or it could connect to a single Xproxy that connects to multiple Xservers
08:15
so, you can handle it in the app or in the X environment
08:15
up to you
08:16oh207_ has quit IRC
08:16
<alkisg>
I'm not sure I got the Xproxy thing, how can an Xproxy connect to multiple Xservers? (ok - let me read up on "Xproxy" and I'll ask again if I have any questions)! Thanks!!!
08:16oh207_ has joined #ltsp
08:17
<Gadi_eeepc>
alkisg: have you used VNC before?
08:17
<alkisg>
Yes, sure
08:17
<Gadi_eeepc>
so, you can share the same Xproxy amongst multiple Xserver connections
08:17
because the vncviewer (Xclient to the Xserver) connects to the Xproxy running on the server
08:18
and the Xproxy allow for multiple concurrent connections
08:18
so, it can reflect the session to many viewers at once
08:18
<alkisg>
Ah, ok, that's the opposite of what I was looking for.
08:18
Gadi_eeepc: a last question, please. If I copy the xauthority file to the server, is it possible to use it to run an application on the server and display the output to the client? (ignore any security concerns, I just want to see if I got this right)
08:18
(the xauthority from a TC, I mean)
08:19
<Gadi_eeepc>
alkisg: it should be on the server already in ~/.Xauthority
08:19
<alkisg>
Really? I thought that was local to the client
08:19
<Gadi_eeepc>
hmm...
08:19
I think it should bee there
08:19
*be
08:20
<alkisg>
I mean this: User A sits on client B. User C sits on server D. Can it use the xauthority of user A to display an application in the client B?
08:20
<Gadi_eeepc>
ah
08:20
well, you can accept remote connections with xhost
08:21
<alkisg>
OK, so it can be done, but maybe the xauthority file from user A is useless to the user C, got it. :)
08:21
Thanks a lot, Gadi!
08:21
<Gadi_eeepc>
right xhost can be used to control access to the Xdisplay
08:22
without transferring files
08:22
you can allow access from hostname/user
08:22* alkisg needs to think if he can use this in a secure manner, for the student to display arbitrary apps on the TCs..
08:22
<alkisg>
*teacher
08:23
<Gadi_eeepc>
you should be able to allow just the teacher to write to any display
08:23
something like: xhost +teacher
08:23
where teacher is the teacher's username
08:23
<alkisg>
Perfect!!!
08:23
<Gadi_eeepc>
that command should run in every user session
08:23
upon login
08:24
<alkisg>
Not in the chroot, one time for all clients?
08:24
<Gadi_eeepc>
I would do it on the server, so you can change it easily
08:24
<alkisg>
OK
08:24
<Gadi_eeepc>
Xclients are network transparent, so you can run xhost from either side
08:25
<alkisg>
wow... I wouln't have figured this out
08:26
<Gadi_eeepc>
when I started my first job out of school, I worked on an HPUX machine - my mentor had me type "xhost +" in a terminal window. When I got back from getting some coffee, he had run a screensaver that melted my display
08:26
:P
08:26
be careful with X access
08:26
:)
08:26
<alkisg>
Heh!! lol!
08:27
OK, it's strictly on local network, with 10-12 year old students knowing only a little openoffice... :)
08:27
I don't expect to see any screens melting!
08:38pmatulis has quit IRC
08:39pmatulis has joined #ltsp
08:44alkisg has quit IRC
08:44polytan has quit IRC
08:45npman has joined #ltsp
08:47polytan has joined #ltsp
08:49polytan has quit IRC
08:50
<sbalneav>
Morning all
08:51polytan has joined #ltsp
08:59F-GT has quit IRC
09:01litlebuda has joined #ltsp
09:04kalib is now known as off-kalib
09:05otavio has quit IRC
09:06otavio has joined #ltsp
09:12chrisinajar has joined #ltsp
09:13F-GT has joined #ltsp
09:21pasmen has quit IRC
09:24spectra has joined #ltsp
09:27rjune has quit IRC
09:36rjune has joined #ltsp
09:42etyack has joined #ltsp
09:46Gadi_eeepc has quit IRC
09:56Gadi_eeepc has joined #ltsp
10:00etyack_ has joined #ltsp
10:01gate_keeper_ has quit IRC
10:02etyack has quit IRC
10:05etyack_ has quit IRC
10:06etyack has joined #ltsp
10:12npman has quit IRC
10:18dirigeant has quit IRC
10:18
<polytan>
how to have a "local" graphical term ?
10:21evilx has joined #LTSP
10:22
<sbalneav>
polytan: You mean, launch an X term on the local terminal?
10:23
<polytan>
yes
10:23
<sbalneav>
Are you wanting it launched as the user currently logged in? If so, you'll want localapps.
10:24
<polytan>
ok
10:27staffencasa has joined #ltsp
10:27cliebow has quit IRC
10:32japerry has quit IRC
10:33
<polytan>
good bye
10:33polytan has quit IRC
10:43vvinet has joined #ltsp
10:44cliebow has joined #ltsp
10:46japerry has joined #ltsp
10:46runout has joined #ltsp
10:50
<mistik1>
morning #ltsp
10:51
<warren>
stgraber: we can fully translate those scripts without needing to pull in gettext into the client chroot
10:51
stgraber: gettext is ONLY needed for building
10:52
<ogra>
we do translate ltsp-build-client already
10:52
doing that with the scripts should just be a copy/paste job from there
10:55likuidkewl has joined #ltsp
10:55likuidkewl is now known as dmaran
10:56
<dmaran>
Real quick we are looking to run a script at login and logoff any views on the "best place" for it to be called from in ltsp5 Ubx64
10:57
<warren>
stgraber: I figured it out, it is bash itself doing localization without gettext involved, pulling from .mo files based upon LANG and TEXTDOMAIN
10:58hanthana has left #ltsp
10:59
<warren>
ogra: copy/paste job from where?
10:59
ogra: is that because dash doesn't support the native localization like bash?
10:59
<ogra>
warren, ??
11:00
<warren>
ogra: how do you translate your ubuntu bootup messages?
11:00
<ogra>
thats been added by vagrant years ago
11:00
ltsp-build-client never runs under dash ... unless an admin creates a script that calls it or so
11:01
<warren>
oh
11:01
<sbalneav>
dmaran: at logon, usually in the Xsession chain.
11:01
<warren>
stgraber: I think we should write a ldm-dialog function that extracts a translated string instead of pull in 5MB of extra dependencies to get it from a shell script.
11:01
<sbalneav>
dmaran: /etc/X11/Xsession.d or whatever your host is.
11:01
<warren>
ogra: does dash support native localization?
11:02
<ogra>
warren, user shells are still bash ... only /bin/sh points to dash
11:02
<sbalneav>
logoff, there's currently no good way to do that that I know of.
11:02
<ogra>
warren, so only if you are not logged in dash will be used (i.e. for system scripts)
11:03
else your environment has SHELL=/bin/bash
11:04cliebow has quit IRC
11:04runout has left #ltsp
11:05
<cyberorg>
warren, were you able to fix the hang on shutdown? i get the same here :(
11:05
<warren>
cyberorg: with nbd?
11:06
cyberorg: I talked with our OS people, they said, "Stop using nbd" =(
11:07
<cyberorg>
i get it with both nfs and nbd
11:08
<warren>
cyberorg: oh, nfs is fine here
11:08
cyberorg: only nbd
11:09
<cyberorg>
i am thinking of patching greeter to call custom shutdown and restart script, that closes most stuff normally and then call poweroff -fp
11:09
bypassing the systems init scripts that is hanging
11:10off-kalib is now known as kalib
11:12
<warren>
why not write a replacement shutdown init script?
11:13
<ogra>
just dont call the nbd scrip
11:13
t
11:13
<cyberorg>
ogra, i removed all the K scripts the client still hangs
11:14
we are still on sysvinit btw :)
11:14
<ogra>
my initial idea was to have RC0_WHITELIST and RC6_WHITELIST additionally to the other RC whitelists
11:14
what do you see on screen if the client hangs ?
11:15
<cyberorg>
ogra, one of the init scripts with done at the end
11:15
<warren>
cyberorg: I think I'm just abandoning nbd
11:15
<ogra>
no "system halted" message ?
11:15
<cyberorg>
there wouldnt be anything once syslog is shutdown
11:15
ogra, no
11:15
<warren>
cyberorg: there are other problems, like 2.6.29 kernel is totally incompatible
11:16nicoAMG has joined #ltsp
11:16
<ogra>
cyberorg, then th kernel never recieved the shutdown command
11:17
warren, to what ?
11:17
<warren>
cyberorg: squashfs incompatible with any released version of mksquashfs
11:17
oops
11:18
ogra: ^^
11:18
<ogra>
ah
11:18
yeah, you will likely need a new version
11:18
which doesnt really bother me in ubuntu
11:19
(2.6.29 is still one release ahead ... 9.04 will use .28)
11:19
<warren>
cyberorg: hangs on every client?
11:19
cyberorg: curious that it hangs on nfs, that means you are not having the same issue as me
11:20
<cyberorg>
warren, on the laptop i am testing, have to try one more
11:25* ogra sighs about evdev ...
11:26sepski has joined #ltsp
11:26
<mistik1>
Hey guys, can i ask your opinions on something as you understand it...
11:27
Is GPL v3 incompatible with GPL v2 ?
11:27
This is not a legal request just asking an opinion of friends
11:27
<dmaran>
sbalneav: what about using /etc/gdm/WhatEver or is this broken in LTSP
11:27
<ogra>
dmaran, ltsp doesnt use gdm at all
11:28
<dmaran>
crap forgot about that
11:29japerry has quit IRC
11:29
<dmaran>
@ogra: anything comparable in ldm?
11:29
<ogra>
only in very recent upstream
11:29
<dmaran>
ugh
11:29
<ogra>
so ubuntu or fedora development releases should have it
11:30
<dmaran>
This is for a lts box, so not really an option
11:31
<sbalneav>
dmaran: I guess we could better help you if we knew what you were wanting to run. Perhaps there's a better/other way of doing it?>
11:32
<dmaran>
The idea is to run asimple rsync at logon/out
11:32
no biggie I though
11:32
t
11:36dirigeant has joined #ltsp
11:38npman has joined #ltsp
11:38
<sbalneav>
This is for what? Updating a remote home drive or something?
11:38
perhaps something in .logout in the users' home dir?
11:39
Certainly, the /etc/X11/Xsession.d can be used for the rsync at login.
11:39
<dmaran>
sbalneav: this is fired before the desktop loads correct?
11:39
<sbalneav>
(make sure you rsync ... || true in the script, as Xsession scripts are usually .'d
11:40
dmaran: Should be.
11:40
<dmaran>
Ok
11:58F-GT has quit IRC
12:00Lns has joined #ltsp
12:00Egyptian[Home]1 has joined #ltsp
12:02alkisg has joined #ltsp
12:03
<_UsUrPeR_>
what version did the X_VIRTUAL and X_RANDR_MODE_0 release in?
12:03
err version of ltsp for F10
12:04
<warren>
_UsUrPeR_: http://kojipkgs.fedoraproject.org/packages/ltsp/
12:04
_UsUrPeR_: http://kojipkgs.fedoraproject.org/packages/ldm/
12:04
sbalneav: ogra: stgraber: I want to tag ltsp-trunk
12:05
<sbalneav>
fine by me
12:06
<ogra>
sure sure ...
12:12F-GT has joined #ltsp
12:14Egyptian[Home] has quit IRC
12:17otavio has quit IRC
12:18otavio has joined #ltsp
12:26vagrantc has joined #ltsp
12:30plamengr has quit IRC
12:40
<_UsUrPeR_>
ok, so I have updated to 5.1.51 in F10. It doesn't seem like the "X_VIRTUAL" command is working properly. X_VIRTUAL = "3360 1050" should work, correct?
12:41
<stgraber>
warren: a sec, just checking if I have something to push
12:45vvinet has quit IRC
12:45etyack has quit IRC
12:48
<stgraber>
well, looks like you released anyway :)
12:51shrek has joined #ltsp
12:54din_os has joined #ltsp
12:55otavio has quit IRC
12:55otavio has joined #ltsp
12:56npman has quit IRC
12:56
<din_os>
exo ena problimataki
12:56
se ena kainourio mixanima, dokimazo ton client
12:57
den 8elw dhcp opote me to config orizo mono to next-server, filename kai root-path
12:57
kanw ena autoboot meta
12:58
kai parolo pou katebazei ton kernel otan ton trexei mou dinei stoixeia
12:58
<Gadi_eeepc>
_UsUrPeR_: check the xorg log file and /var/run/ltsp-xorg.conf on the client
12:58
<din_os>
IP: 192.168.2.5:192.168.2.1: ....etc
12:58
<_UsUrPeR_>
Gadi_eeepc: k. How do you like the eeepc?
12:58
<din_os>
to proto ip to pairnei swsta apo ton router tou spitiou
12:59
<Gadi_eeepc>
love it - especially when I havee to juggle 2 kids on a snow day :P
12:59
<_UsUrPeR_>
:)
12:59
<din_os>
to deutero einai la8os einai to ip tou router enw 8a eprepe na einai to ip tou server (.2.20)
12:59
<_UsUrPeR_>
I liked the size/battery life, but the keyboard/mouse was something that it seemed like I would have to get used to
12:59* Gadi_eeepc has small fingers
13:00
<_UsUrPeR_>
heh
13:00
what language do you suppose that is?
13:00
<Gadi_eeepc>
greek?
13:00
<laga>
portuguese?
13:00
<_UsUrPeR_>
I find it interesting that there's "8"s interspersed
13:01
<din_os>
nomizo oti o kernel ksanapsaxnei to ip tou server kai enw to eixa katebasei kanonika me tftp, meta briskei to .2.1 anti gia .2.20
13:02
<laga>
is he actually talking to someone?`
13:02
<din_os>
oups sorry
13:02
ok here goes again :P
13:03
<Gadi_eeepc>
din_os: the universal translator is broken - what language is that
13:03
?
13:03
<din_os>
i have a tiny problem
13:04
i load up the 8.04 client on a new system
13:04
i don't want dhcp, so i hit ctrl-b to set it manually
13:05
i use config to set next-server filename and root
13:05
then i hit autoboot
13:05
this works on other systems
13:05
but
13:05
<Gadi_eeepc>
ctrl-b?
13:06
<din_os>
ermm.. gPXE
13:06
<Gadi_eeepc>
ah
13:06
<din_os>
maybe i am in the wrong forum :)
13:06
but i think it's a kernel error and not gPXE so...
13:07
well never mind
13:08
<Gadi_eeepc>
no, keep going
13:08
<din_os>
sure?
13:08
ok
13:08
<Gadi_eeepc>
sure - I dont know much about gPXE myself, but maybe someone here can help
13:08
<din_os>
tftp finds the kernel allright
13:09
and after fetching it executes it
13:09
but then the kernel seems to display the IP details as : 192.168.2.5:192.168.2.1....etc
13:10
usually the first ip is the clients ip and the second the ltsp server's
13:11vagrantc has quit IRC
13:11
<din_os>
but it seems to forget the variable next-server and instead sets my home router's ip (.2.1) instead of .2.20
13:12kalib has left #ltsp
13:12
<din_os>
of course the kernel get's stuck trying to find pxelinux.conf
13:12
that is all... :) sorry for the greek
13:13
<Gadi_eeepc>
is it your router responding too the dhcp request?
13:13
*to
13:13
<_UsUrPeR_>
din_os: is the client on the same network as two DHCP-providing servers? (i.e. router and gpxe server)
13:13vagrantc has joined #ltsp
13:15
<din_os>
Gadi: i think the router might be doing something, but it didn't stop the initial tftp fetch
13:15
UsUrPer: no only my phillips dsl router provides dhcp
13:16
<Gadi_eeepc>
so, the router has the appropriate next-server statement?
13:16
<din_os>
yup, it uses it to imgfetch
13:16
it does so successfully
13:16
imgexec also runs the fetched kernel
13:17
i guess it doesn't pass the next-server argument? ??
13:17
<Gadi_eeepc>
well, next-server is sent down by dhcp, so if you ctrl-b and do not allow dhcp, that may be why
13:17
<din_os>
nono
13:17
I set next-server myself
13:17
<Gadi_eeepc>
on the client?
13:17
<din_os>
the router has no capability to send this so it just sends an ip
13:17
<Gadi_eeepc>
ah
13:18
<din_os>
yup i use the set ip <> command
13:18
sorry
13:18
<Gadi_eeepc>
hmm.. sounds like a gPXE issue
13:18
<din_os>
the set next-server command
13:18
well if gPXE is meant to pass the next-server via args then yes
13:19
<Gadi_eeepc>
yeah
13:19
<din_os>
if not, then the kernel cannot get it right
13:19
<Gadi_eeepc>
I dont know much about gPXE
13:19
you do add an nbdroot kernel arg to the pxeconfig, tho, right?
13:20
<din_os>
hmm that i do not know, how do i check
13:21
'/var/..../pxelinux.conf/default ???
13:23
<Gadi_eeepc>
yup
13:23
<din_os>
i don't think that would matter, because it gets stuck trying to load pxelinux.conf
13:23
<ogra>
right
13:23
it dies before
13:23
<warren>
stgraber: go ahead and tag whenever you want
13:24
Ryan52: I'll wait to build this next release when you're back, so you can learn how to do it
13:25
<din_os>
i don't know, maybe the router does something to it, everything works with different router with dhcp enabled and all that, or maybe the network card on the pc is too new??? who knows
13:25* ogra curses Xorg upstream ... ALL LIARS !
13:25
<din_os>
let me set a simpler problem
13:26
on the working clients the resolution is 1280x1024 or something
13:26
even though i set it 1024x768 on lts.cfg
13:27
what gives?
13:29
<Gadi_eeepc>
you mean lts.conf
13:29
:)
13:30
<din_os>
ah yes
13:30
<Gadi_eeepc>
and: /var/lib/tftpboot/ltsp/i386/lts.conf
13:30
<din_os>
yup
13:30
<Gadi_eeepc>
and, you have have a section like: [default]
13:30
before the config variables
13:30
<_UsUrPeR_>
warren/gadi et all: building a new client w/LTSP 5.1.51 results in the error /etc/sysconfig/ltspdist: line 18: syntax error near unexpected token `)'
13:31
it turns out that line 18 is missing two semi-colons
13:31
<Gadi_eeepc>
sounds like a fedora thing
13:31
<_UsUrPeR_>
it appears to build a client properly once I put them in
13:31
<ogra>
yeah, us a stable distro
13:31
*use
13:31
<_UsUrPeR_>
gadi_eeepc: correct: F10
13:31* ogra hides quickly
13:31
<warren>
_UsUrPeR_: huh
13:32
_UsUrPeR_: i'm trying
13:32* Gadi_eeepc wishes ogra put a default lts.conf file in the tftpdir :P
13:32
<din_os>
Gadi yes, and i use X_MODE_0=1024x768 to set it
13:32* Gadi_eeepc doesnt know how many folks don't know what format to use
13:33
<din_os>
but also enable configure_x=true, maybe that kind of confuses it?
13:33
<Gadi_eeepc>
din_os: try adding: CONFIGURE_X=True
13:33
<ogra>
Gadi_eeepc, ask the package maintainer to do that, not me :P
13:33
Gadi_eeepc, shouldnt be needed in hardy
13:33
<din_os>
well, i'll try removing it?
13:33
<_UsUrPeR_>
!pastebot
13:33
<ltspbot>
_UsUrPeR_: "pastebot" is The LTSP pastebot is at http://pastebot.ltsp.org. Please paste all text longer than a line or two to the pastebot, as it helps to reduce traffic in the channel. A link to the content will be pasted in the channel.
13:34
<ogra>
hardy was just gutsy with bugfixes and very few new features
13:34
so most gutsy settings apply there
13:34
<ltsppbot>
"_UsUrPeR_" pasted "line 18 problem just above line fc11) missing ;;" (32 lines) at http://pastebot.ltsp.org/219
13:35
<ogra>
ouch
13:35
<warren>
stgraber: translations in shell scripts without .gettext.sh
13:35
1. set TEXTDOMAIN to the text domain you're expecting to look up translations in.
13:35
2. make sure your LANG locale is set before the shell runs
13:35
3. prefix your echo text with $
13:35
<Gadi_eeepc>
poor case statement
13:35
<_UsUrPeR_>
lol
13:36
<Gadi_eeepc>
din_os: if u want, paste your lts.conf
13:36
<warren>
_UsUrPeR_: oops
13:36
<Gadi_eeepc>
any syntax error can render it useless
13:36
Gadi_eeepc: another test is to get a shell on the client and run: getltscfg -a
13:37
<ogra>
ogra, yeah yeah ... the old talk to yourself syndrome
13:37
<warren>
ogra: introspection
13:37
<din_os>
thanks, i'll try some stuff myself first, i think it doesn't load it at all using tftp so i might just paste it in its original position and rebuild the image
13:37
<Gadi_eeepc>
Gadi_eeepc: why is ogra talking to himself?
13:38* Gadi_eeepc shrugs
13:38
<ogra>
ogra, because thats the last person that will still listen to him, even if nobody else does :)
13:38
<din_os>
my case is kind of weird because ltsp server is not also dhcp but a dhcp exists on the network only to provide ips
13:39
i think many submodules try to find either tftp server or ltsp server thinking it is the dhcp server by default
13:40
<johnny>
you don't want 2 dhcp servers
13:40
just 1
13:40
<din_os>
i only have one
13:40
<johnny>
oh
13:40
<din_os>
but is not the ltsp
13:40
<johnny>
we rarely make such assumptions
13:40
that the dhcp server== the ltsp server
13:40
many people run in the configuration you describe
13:41
<din_os>
ok
13:42pscheie has quit IRC
13:45
<johnny>
yay.. my box is coming back alive.. turns out i had a wrong egrep somehow
13:48bobby_C has joined #ltsp
13:52pscheie has joined #ltsp
13:57shrek has quit IRC
13:58
<kc8pxy>
johnny: congrats :)
13:58vvinet has joined #ltsp
14:02
<warren>
stgraber: you want to tag?
14:04
Ryan52: err, I have to go ahead with a build because I broke something important prior
14:04
stgraber: I'm tagging
14:05
stgraber: ogra: Gadi_eeepc: vagrantc: anything else for ltsp-trunk before taggin?
14:06
<ogra>
not from me ... (though stgraber is more important here anyway)
14:10
<Gadi_eeepc>
ditto
14:10
:D
14:11alekibango has joined #ltsp
14:11takis_rs has joined #ltsp
14:15spectra has quit IRC
14:15etyack has joined #ltsp
14:17
<elisboa>
where do I find information about serving via VNC?
14:22
<dmaran>
@all any ideas why sometimes dhcp will work but not tftp 'read error time out' or dhcp will just sit there. This is a test setup I am trying to run and it not cooperating at all ltsp5 install Xubuntu 8.04.1 medium
14:27warren has quit IRC
14:28
<Gadi_eeepc>
din_os: did you use the short form of the filename in gPXE?
14:29
din_os: Beware: If you want to use the next-server as the DHCP information contains, you must either use the short form (without tftp: prefix), or use three slashes in a row,
14:30
<din_os>
now that is interesting
14:30
i'll try with three
14:32takis_rs has left #ltsp
14:32
<alkisg>
Gadi_eeepc: din_os puts the information in gpxe manually, with the embedded config program. So in the second dhcp request done from ipconfig the initramfs, there isn't any boot filename field nor next-server field.
14:34
din_os: if you can make your router send an next-server field, do so, otherwise you'll have to use the IPAPPEND 3 directive in pxelinux.cfg/default.
14:52dmaran has left #ltsp
14:53warren has joined #ltsp
14:56warren has quit IRC
14:56warren has joined #ltsp
15:01Eghie has joined #ltsp
15:03otavio_ has joined #ltsp
15:10otavio has quit IRC
15:20Eghie has quit IRC
15:23pasmen has joined #ltsp
15:25din_os has quit IRC
15:26etyack has quit IRC
15:30yanu has quit IRC
15:30warren has quit IRC
15:30warren has joined #ltsp
15:33pasmen has quit IRC
15:44vvinet has quit IRC
15:45alkisg has quit IRC
15:49bobby_C has quit IRC
15:53yanu has joined #ltsp
15:57runout has joined #ltsp
16:04Guaraldo has quit IRC
16:10BrunoXLambert has quit IRC
16:14nicoAMG has quit IRC
16:30din_os has joined #ltsp
16:33elisboa has quit IRC
16:33elisboa has joined #ltsp
16:36Shingoshi has quit IRC
16:39Shingoshi has joined #ltsp
16:50
<ogra>
wohoo ... signed PPAs
16:58
<stgraber>
yeah, all of mine now have their GPG key, that's really cool
17:06runout has left #ltsp
17:13evilx has quit IRC
17:14din_os has left #ltsp
17:17
<ogra>
yup
17:17
Gadi_eeepc, note that the urls have changed ...
17:22sepski has quit IRC
17:42japerry has joined #ltsp
18:00nothingman has joined #ltsp
18:05
<nothingman>
hi, all
18:05
wondering how confused I am about the ltsp-* commands in edubuntu
18:07
ltsp-build-client makes a complete chroot directory containing all the installed packages -- binaries and config files, mainly
18:07
ltsp-update-kernels
18:08
ltsp-update-kernels makes a kernel for any such chroot with the appropriate architecture?
18:08
and ltsp-update-image makes a compressed file of the chroot that is treated as an nbd? so, everything but the kernel?
18:09
and also excluding anything that would be nfs-mounted, of course, such as /home from the original root
18:20
or maybe you're all just wondering why I haven't read the man pages, but I'm still confused, because there's no image being created for my fat client image
18:25
<Ryan52>
warren, okay, back...that's okay, you can teach me later today.
18:25litlebuda has quit IRC
18:26
<Ryan52>
vagrantc: that (your response to that bug) reminds me...does gdm make .xsession take precedence over .dmrc? cause ldm doesn't..
18:43alekibango has quit IRC
18:48staffencasa_ has joined #ltsp
18:52
<warren>
Ryan52: we have nothing ready to build into fedora at the moment
18:55
<Ryan52>
ok.
18:58japerry has quit IRC
19:01staffencasa has quit IRC
19:01staffencasa_ has quit IRC
19:03
<Ryan52>
$ TEXTDOMAIN=iso_639; LANGUAGE=fr_CA.UTF-8; echo $"French"
19:03
French
19:03
warren, what did I do wrong there?
19:05
oh, nevermind, you have to export it..
19:05
<stgraber>
Ryan52: isn't $"" deprecated ?
19:06
<Ryan52>
11:36 < warren> stgraber: translations in shell scripts without .gettext.sh
19:06
11:36 < warren> 1. set TEXTDOMAIN to the text domain you're expecting to look up translations in.
19:06
11:36 < warren> 2. make sure your LANG locale is set before the shell runs
19:06
11:36 < warren> 3. prefix your echo text with $
19:06
stgraber, no clue.
19:06
<stgraber>
Ryan52: IIRC xgettext told me that $"" was deprecated and shouldn't be used
19:07
<Ryan52>
warren, that only works for me after I ". gettext.sh"...
19:07
<nothingman>
wondering how confused I am about the ltsp-* commands in edubuntu
19:07
ltsp-build-client makes a complete chroot directory containing all the installed packages -- binaries and config files, mainly
19:07
ltsp-update-kernels makes a kernel for any such chroot with the appropriate architecture?
19:07
and ltsp-update-image makes a compressed file of the chroot that is treated as an nbd? so, everything but the kernel?
19:08
and also excluding anything that would be nfs-mounted, of course, such as /home from the original root
19:09
<stgraber>
stgraber@sahal:~$ xgettext -L shell I00-nbd-checkupdate.bad
19:09
I00-nbd-checkupdate.bad:39: warning: the syntax $"..." is deprecated due to security reasons; use eval_gettext instead
19:09
Ryan52, warren: ^
19:09
<Ryan52>
stgraber, and it doesn't even work like warren said anyway...
19:09
so ya, I'm convinced ;)
19:11
<stgraber>
Ryan52: as I said, I'm open to anything easier/lighter than what I did so far but as far as I could see the way to gettextize a script is to source gettext.sh set the domain and path to the locales and then use gettext instead of echo and eval_gettext when your string contains a variable
19:11
http://www.gnu.org/software/automake/manual/gettext/Preparing-Shell-Scripts.html
19:11
basically
19:11
<Ryan52>
yes, I'm not argueing with you...
19:12
<warren>
Ryan52: you cannot set the locale within the same shell process that gets translated.
19:12
Ryan52: translation happens once only when the interpreter runs at the beginning
19:13
Ryan52: if this is too difficult, then we should write a C binary that pulls translated strings out of the .mo. I am NOT installing gettext into the chroot.
19:13pmatulis has quit IRC
19:13
<Ryan52>
I see.
19:13pmatulis has joined #ltsp
19:15
<warren>
stgraber: isn't $"" what your own distro's initscripts do?
19:15
<Ryan52>
warren, ya, I'm more than fine with that. trivial to do..
19:15
<warren>
stgraber: your own distro translates its bootup without gettext.sh
19:15
but I think a small binary to pull translations from .mo is better
19:16
<Ryan52>
it seems really stupid for us to write a small binary, imo...
19:16
<warren>
what would you prefer?
19:17
<Ryan52>
why is your gettext package so big?
19:17* Ryan52 wonders how to get that how big Debian's is
19:17
<warren>
Ryan52: because gettext is a DEVELOPMENT TOOL, not necessary during runtime.
19:17
your own distro's initscripts are translated without it
19:18
<Ryan52>
yes warren, we get that you think it's a development tool, however the official docs that we keep pointing to make it seem differently. (which is correct, I don't know..)
19:18
<nothingman>
am I just asking a stupid question, or does no one here have an answer?
19:18
<Ryan52>
warren, who ever said that? I said Debian's initscripts don't use it, but I never said they were translated.
19:19
<warren>
Ryan52: ok, our alternative re-run sh in cases where we want it to pull translations
19:19
Ryan52: I can run my entire system just fine ripping out gettext entirely, because it is a development tool.
19:20dirigeant has quit IRC
19:20
<warren>
anything in runtime that depends on it should be fixed.
19:20
<Ryan52>
ltsp-server's Debian package does.
19:20
but ya, gettext-base does have a really low reverse dependancies list in debian..
19:21
<warren>
because it is entirely not necessary to use it
19:21try2free has joined #ltsp
19:21
<Ryan52>
warren, stgraber: so what's the problem with $""? anybody know?
19:22
hrm. so how do the C binaries get gettext? are they statically linked to it? 0.o
19:22
<warren>
Ryan52: they don't use gettext
19:22
Ryan52: gettext's primary purpose is to pull strings out of source code to put it into .pot files
19:22
<Ryan52>
they use the function called gettext...is that in the standard library?
19:22* warren checks
19:23
<warren>
/usr/lib64/libasprintf.so.0 and /usr/lib64/libgettextpo.so.0
19:24
hmm
19:24
it seems most apps don't use that either....
19:24
<Ryan52>
that comes from the gettext-base package in debian.
19:24
s/that/those/
19:24
<warren>
in fact, nothing uses that
19:24
<Ryan52>
heh
19:25
*shrug*. I don't care how it works, as long as it does :)
19:25
<warren>
I'll ask our eng-i18n engineers, brb
19:25* stgraber checks to see if our boot process is actually translated, AFAIk it's not supposed to be :)
19:26
<stgraber>
root@mxapp-126-1-1:/var/log# /etc/init.d/gdm restart * Stopping GNOME Display Manager... [ OK ] * Starting GNOME Display Manager...
19:26
that's with LANG=fr_CA.UTF-8 and all langpacks
19:28
Depends: libc6 (>= 2.8~20080505), libgcc1 (>= 1:4.1.1), libstdc++6 (>= 4.1.1)
19:28
Filename: pool/main/g/gettext/gettext-base_0.17-3ubuntu2_i386.deb
19:28
Size: 73350
19:28
what's making it so big in fedora ?
19:29
<Ryan52>
I'm guessing that they don't have it split into gettext-base
19:29
since warren keeps calling it "gettext".
19:29
<warren>
"usually we use a macro to deal with it like _(). if you use glib, it's defined in glib/gi18n.h or glib/gi18n-lib.h"
19:29
<stgraber>
right, both are quite different
19:30
<warren>
glib can do it. We already provide tiny support binaries in there. Just add it there.
19:31
<stgraber>
warren: what's the package that contains /usr/bin/gettext in fedora and what else is in it ?
19:31
<warren>
stgraber: http://fpaste.org/paste/2449
19:32
stgraber: we don't split it because it is NOT a runtime package, development only
19:32
<Ryan52>
ours is split
19:32
http://pastebin.com/f594f26fd
19:32
<warren>
our initscripts are translated without it installed
19:32
<Ryan52>
that's all we have.
19:32
but that's a security problem, according to what stgraber posted.
19:32
<warren>
where?
19:33
<stgraber>
not according to me but according tox gettext :)
19:33
*xgettext
19:33
<Ryan52>
17:10 < stgraber> I00-nbd-checkupdate.bad:39: warning: the syntax $"..." is deprecated due to security reasons; use eval_gettext instead
19:33
<stgraber>
stgraber@sahal:~$ xgettext -L shell I00-nbd-checkupdate.bad
19:33
I00-nbd-checkupdate.bad:39: warning: the syntax $"..." is deprecated due to security reasons; use eval_gettext instead
19:34
<warren>
asking someone...
19:34
<Ryan52>
if you want, I'll just write the little c binary...
19:35
<warren>
that would be preferable
19:35* Ryan52 enjoys reinventing wheels
19:35
<stgraber>
warren: anyway, if $"" is working for your init scripts it's because something at some points defines $""
19:35
<warren>
stgraber: no
19:35
stgraber: it is a built-in of the shell
19:35
<Ryan52>
stgraber: no, his works fine...even I agree with that one :)
19:35
<stgraber>
Ryan52: hmm .. but that's not dash I guess ? :)
19:36
<Ryan52>
oh, I only tested interactively...one sec.
19:37
ahah, I see.
19:37
isn't this /bin/bash anyway?
19:37
I thought that that's how our rc.d scripts were ran.
19:38
<warren>
Keep in mind that LANG must be set *before* the shell interpret
19:38
interpreter
19:38
<Ryan52>
$ dash test.sh
19:38
$French
19:38
$ bash test.sh
19:38
Français
19:38
<warren>
what shell are we using for the rc.d scripts?
19:38try2free has left #ltsp
19:38
<Ryan52>
iirc, bash.
19:39
#! /bin/sh
19:41
<stgraber>
Ryan52: does that only work with $"" or also with gettext "" ?
19:42
I'd appreciate xgettext not complaining every-time I update the .pot telling me it's deprecated :)
19:42
<Ryan52>
I'd appreciate if we didn't use things that have "security" reasons for being deprecated..
19:43* Ryan52 wonders if they are relevant or not
19:43
<warren>
hmm, we don't use xgettext anywhere here.
19:43
or gettext.sh
19:44
<Ryan52>
xgettext generates the .pot and .po files (or one of them) from the source files.
19:44
<stgraber>
warren: xgettext is the thing used to extract the strings from the scripts
19:44* warren reads why xgettext thinks that is insecure
19:45
<Ryan52>
I doubt it's relevance to our stuff, but still..
19:45
<warren>
Could we please have a tiny binary or ldm-dialog handle this, since glib already does it?
19:45
<Ryan52>
yes, of course.
19:46
I offered a page of scrollback ago :)
19:46
<stgraber>
what we need though is something that's then parsable using xgettext
19:46
otherwise it'll be a pain to generate the .pot :)
19:47
<Ryan52>
stgraber, we can define a shell function called eval_gettext :)
19:47
<vagrantc>
gah.
19:47
<warren>
Ryan52: that calls our binary instead
19:47
<Ryan52>
warren, exactly
19:47
<vagrantc>
what the world needs is a whole new revolution in wheel design!
19:48
we should at least integrate it with dbus or pulseaudio or something, just to make sure we have lots of bugs in it.
19:48
<stgraber>
Ryan52: well, you can probably do that mini-binary and then we can re-implement gettext.sh :)
19:48
let's call it /usr/share/ltsp/gettext.sh and make it export the same functions as /usr/bin/gettext.sh
19:48
:)
19:49
<vagrantc>
gettext.sh is *only* useful as a runtime function.
19:49
that is it's only purpose.
19:49
<Ryan52>
vagrantc: it's only for development, says warren :)
19:49
<warren>
vagrantc: we don't use it anywhere
19:49
<vagrantc>
warren: yes, well, it's used all over the place in debian.
19:50
<Ryan52>
$ apt-cache rdepends gettext-base | wc -l
19:50
30
19:50
...
19:50
out of the how many packges that Debian has?
19:51
<vagrantc>
warren: we've been using this in ltsp-server since 2006 ... why is this a crisis now?
19:51
<Ryan52>
cause it's in the chroot.
19:51
<stgraber>
Ryan52: 30 + all the packages that assumed it'd be here as a base package :)
19:51
<Ryan52>
stgraber, heh, ya.
19:51
warren likes to make me reinvent wheels to keep the chroot small.
19:52
<vagrantc>
that's not really an assumtion they're suppose to make ... as gettext-base is only priority standard.
19:52
<Ryan52>
:)
19:52
vagrantc: you mean "Priority: Standard" isn't required? :p
19:52
<warren>
If you rather not use glib's native ability to handle this, then I'm going to add code that simply leaves it untranslated.
19:52
in my case
19:52
<vagrantc>
Ryan52: nope.
19:52* Ryan52 was joking..
19:52
<warren>
currently it errors out due to a syntax error
19:53
<Ryan52>
warren: so put it back to eval_gettext
19:53
<warren>
perhaps one reason why we don't use gettext.sh anywhere is because we stopped writing text interfaces for anything many years ago.
19:53
<Ryan52>
but anyway, I'm fine with making a new binary.
19:55
<stgraber>
Ryan52: I'm fine with a new binary as long as you provide a gettext.sh script with it and tweak the environment in fedora to use that gettext.sh (so I can still use the one I have in the chroot and don't have to maintain both one binary and one more script :))
19:56
<warren>
Ryan52: I'm OK with that, or I'm also fine with not bothering to translate anything in shell in Fedora.
19:56
stgraber: where is debian's gettext.sh installed?
19:56
<stgraber>
warren: /usr/bin/
19:57* warren writes a patch for review...
19:57
<warren>
hm
19:57
) | ldm-dialog --progress --auto-close "`eval_gettext "A new version of the system is available, rebooting in 10s."`"
19:57
with the ``'s you're already running another process
19:58
<stgraber>
yeah, that's eval_gettext which is defined in gettext.sh
19:58
<Ryan52>
stgraber, so only Fedora would be using my code? 0.o
19:59
<stgraber>
Ryan52: well, why using a clone of something I already have in my chroot and that'll provide less features and won't be as supported as the main one ?
19:59
<Ryan52>
why? I don't see how that matters..
19:59
this is rediculous.
19:59* Ryan52 == done
19:59
<stgraber>
I'll have to ship both :) so why using a clone when you have the original anyway ?
19:59* warren writes a eval_gettext entirely in bash
20:00
<stgraber>
though it perfectly makes sense for fedora
20:00
<warren>
stgraber: because the original requires 5MB in the chroot
20:00
<Ryan52>
warren, but only Fedora is allowed to use your code.
20:00
which is stupid.
20:00
imho.
20:00
<warren>
allowed?
20:00
<Ryan52>
because stgraber is being stubborn...
20:01
<warren>
I don't think his input there makes sense.
20:01
<vagrantc>
warren: why not use /usr/share/ltsp/ltsp-vendor-functions .... ?
20:01
<warren>
vagrantc: does that already exist?
20:01
<Ryan52>
stgraber, why does an ltsp chroot have gettext-base anyway?
20:01
<vagrantc>
warren: in ltsp-common-functions, it sources that if present.
20:02
<stgraber>
Ryan52: well, give me one reason why I should use your implementation of eval_gettext ? Maybe I'm missing something but in my case gettext-base is there by default ...
20:02
Ryan52: hang on a sec, I'll check what's depending on it
20:02
<warren>
I'm thinking if implementing eval_gettext in shell is possible
20:03
<Ryan52>
stgraber: for simplicity so that there don't have to be differences between Fedora and Ubuntu for no reaosn..
20:04
<vagrantc>
eval_gettext *is* implemented in shell.
20:04
<warren>
vagrantc: I mean the part that pulls the string out of .mo
20:04
<Ryan52>
eval_gettext () { gettext "$1" | (export PATH `envsubst --variables "$1"`; envsubst "$1")
20:04
}
20:04
it's running gettext, tho.
20:05
warren, anyway, why would you prefer to do it in shell than c?
20:05
it's gonna be much more trivial in c..
20:05
and either way it's gonna be Fedora specific, I guess..
20:05
<warren>
Wouldn't this be better to have entirely in C?
20:05
without the weird bash loop
20:05
<stgraber>
Ryan52: hmm, ok. Just checked on a clean chroot, gettext-base isn't there by default, so I'm saving 73k by using your script
20:05* vagrantc notices that gettext is written in C
20:05
<Ryan52>
warren, what's "this"?
20:06
<warren>
Ryan52: ldm-dialog --nbd-update-countdown
20:06
<Ryan52>
stgraber, see? that's a lot! :p
20:06
o noez.
20:06
<stgraber>
warren: no please no !!!
20:06
<vagrantc>
why not just ship a copy of gettext with ltsp? in fact, why not ship a copy of every binary we use as part of the ltsp codebase?
20:06
<stgraber>
warren: I have 3 more scripts using ldm-dialog that I need to be translated :)
20:06
<warren>
stgraber: in the chroot?
20:06
<Ryan52>
stgraber: then just let me write the binary and we'll be done with it.
20:06
<stgraber>
warren: yeah
20:06
<warren>
I don't care outside the chroot.
20:07
<stgraber>
warren: session killing, user limitation, ...
20:07
<Ryan52>
it's *really* trivial in to do in c.
20:07
<warren>
I'm in favor of Ryan52 writing a tiny binary that uses libraries that we already use in the chroot.
20:07
<Ryan52>
stgraber, I don't think that it'll hurt you that much to have to "maintain" another binary.
20:07
(note the maintain in quotes, I don't see how it affects you at all..)
20:08
vagrantc, do you have a position on this?
20:08
<vagrantc>
i think it's silly that we should re-implement code that already exists.
20:08
<Ryan52>
one that isn't sarcastic, preferably :)
20:08
it's like 10 lines of code..
20:09
<warren>
vagrantc: you don't care that this is asking me to add 5MB to my chroot?
20:09
<stgraber>
Ryan52: just make sure it behaves as eval_gettext so it doesn't break when sent shell variables and make it so we can call it using eval_gettext otherwise the .pot generator just won't work
20:09
<vagrantc>
warren: i care some, but fedora chooses to make larger binary packages rather than splitting them up. that's a distro-specific decision.
20:09
<warren>
vagrantc: we don't care to split it up because we don't use gettext.sh anywhere during runtime
20:10
<vagrantc>
Ryan52: i'm just hoping this 10 lines of code actually handles all cases well... gettext is much wider deployed and tested and seasoned and weathered.
20:11
<warren>
vagrantc: I don't care if you don't use this tiny binary
20:11
vagrantc: it is my problem if it breaks.
20:12
My alternative that i'm perfectly happy to live with is to not bother translating it. I'm glad that he's willing to write a tiny binary though.
20:12
<vagrantc>
are all the things to be translated part of ldm-dialog?
20:12
i.e. things called with ldm-dialog?
20:12
<Ryan52>
running ldm-dialog...not part of it.
20:12
yes.
20:13
<vagrantc>
maybe ldm-dialog should just support translations.
20:13
<Ryan52>
your point?
20:13
<stgraber>
vagrantc: we can't do that
20:13
<Ryan52>
no, because if you use "hi there $a" it doesn't work.
20:13
and that makes it really hard for xgettext
20:13
<vagrantc>
hm.
20:13
<Ryan52>
possibly even s/really hard/impossible/
20:14
<stgraber>
yeah, I had some issue with xgettext recently, this thing is really hard to make to work :)
20:14
<warren>
I might point out that we seemed so concerned about performance in the past, but using gettext.sh in this manner in shell scripts is really slow.
20:14
<stgraber>
don't expect something tweakable ... we'll likely to have to drop the autotools for po handling just because of it :)
20:14
<warren>
`` runs another process, then it runs gettext (another process), which exits back to the original
20:15
<Ryan52>
no, it does't do subshell..
20:15
(if that's what you mean)
20:15
<warren>
`` doesn't run another process?
20:15
<Ryan52>
not another sh process.
20:16
<warren>
We could just write ldm-dialog --do-something-specific for each case where we need a particular dialog. Then the strings are nicely in C which are easy for autotools to handle.
20:16
<Ryan52>
eww.
20:16
<warren>
and it is faster
20:16
it runs fewer procesess
20:16
<vagrantc>
warren: seems liek it's called from about 4 places, and not in your typical login scenario. performance isn't probably a huge issue.
20:16
<Ryan52>
then we would also need --something-specific-option-1=blah
20:16
<warren>
vagrantc: well, the nbd-update check itself slows down showing the login screen by a noticeable amount
20:16
<Ryan52>
and other mess.
20:17
but stgraber needs it elsewhere, apparently.
20:17
<vagrantc>
warren: i'm not particularly happy about that at all, yes.
20:18
i'm inclined to conditionalize the nbd-update-check script in an lts.conf value.
20:18
<warren>
what happens if it runs backgrounded only after ldm greeter is up?
20:19
oh
20:19
hm
20:19
nevermind
20:19
<stgraber>
warren: there would probably be a problem due to lack of window manager, other than that it should be fine
20:20
<warren>
stgraber: no, it is possible someone might begin logging in and only then the pop-up happens
20:20
reboot could happen during login
20:20
<stgraber>
warren: he can still click cancel
20:20
<warren>
but it is confusing
20:20
<stgraber>
indeed
20:21
<vagrantc>
nbd, a faster route to troubles. :)
20:21
<Ryan52>
is ldm GPLv3+?
20:21
<warren>
nbd is plenty useful without that auto-reboot thing
20:21
<vagrantc>
Ryan52: not yet.
20:21
<Ryan52>
well, this program can be :)
20:21
<warren>
and I'm wary of the explosion of new lts.conf options
20:21* Ryan52 will steal the variable substition stuff from gettext.
20:22* vagrantc watches the lines of code increment...
20:22
<Ryan52>
:P
20:22
<stgraber>
Ryan52: can't you just copy/paste gettext.c and call it ltsp_gettext ? :)
20:22
<warren>
stgraber: it needs libs I think
20:23
<stgraber>
ons
20:23
Depends: libc6 (>= 2.8~20080505), libgcc1 (>= 1:4.1.1), libstdc++6 (>= 4.1.1)
20:23
any of that isn't standard in fedora ?
20:23
<warren>
gettext needs libstdc++?!
20:24
<vagrantc>
required on debian systems.
20:24
<Ryan52>
vagrantc, I'll just delete all of the newlines :P
20:24
<vagrantc>
Ryan52: ah, good thinking!
20:24
<Ryan52>
warren, stgraber: anyway, if we're going that route, then just get the fedora packagers to split it.
20:24
<vagrantc>
warren: i'm guessing this is another one of those debian splits things into smaller packages issues.
20:24
<warren>
Ryan52: no chance that will happen
20:25
<Ryan52>
if we are going to actually copy it all, then that's beyond stupid.
20:25
<warren>
Ryan52: they'll tell me "don't use it" because nothing else does, meanwhile they will bitch that they have to edit hundreds of other packages to comply with the new gettext package split names.
20:25
<vagrantc>
Ryan52: if we're not copying it all, what are we missing?
20:26
<Ryan52>
vagrantc, dunno...do I look like I understand l10n? :)
20:26
I was just gonna implement basic eval_gettext and nothing else.
20:26
I doubt we need anything else.
20:26
do we?
20:26
<warren>
I'm fine with that
20:26
) | ldm-dialog --progress --auto-close "`eval_gettext "A new version of the system is available, rebooting in 10s."`"
20:27
<vagrantc>
don't really know enough to know what we're missing.
20:27
<warren>
seriously, this doesn't make other people want t pouke?
20:27
to puke
20:27
<vagrantc>
warren: you don't have to include that plugin in fedora.
20:28
<stgraber>
I00-nbd-checkupdate, S15-userLoginCheck, S20-restrictUser and S99-ltsp-cluster are all using ldm-dialog
20:28* vagrantc might not include it in debian ...
20:28
<stgraber>
so it's not only the nbd-checkupdate
20:29
<vagrantc>
ah, so all the other scripts are broken :)
20:29
<Ryan52>
or silent..
20:29
<vagrantc>
since they call eval_gettext without making it available...
20:29
<Ryan52>
what?
20:29* Ryan52 confused
20:29
<stgraber>
vagrantc: the first thing it does is looking for NFS, so in your case it won't take so much extra time
20:29
I just haven't gettextized the others yet
20:30
<vagrantc>
stgraber: well, they currently call eval_gettext ...
20:30
<stgraber>
vagrantc: do they have the header set ?
20:30
<vagrantc>
but don't '. gettext.sh'
20:30
<Ryan52>
hrm.
20:30* Ryan52 sees a problem with that..
20:30
<stgraber>
ok, that's what I thought :)
20:30
<warren>
stgraber: S99-ltsp-cluster is this script meant to exit?
20:31
stgraber: I thought we arn't supposed to exit from these scripts
20:31
<stgraber>
oh, that's a good question :)
20:31
what happens if we exit from a rc.d script ?
20:32
<warren>
stgraber: then it stops execution of the entire string of scripts
20:32
stgraber: since they are all sourced and executed in sequence
20:32
<stgraber>
good thing it's currently the last one then :), what is it supposed to do instead ? a giant if or can I return or something from it instead of exit ?
20:33* vagrantc is struggling with getting a useful response out of http://bugs.debian.org/512505
20:33
<vagrantc>
they claim that it doesn't store the default session, but also claim that they can't login.
20:34
<warren>
stgraber: you can't exit or return
20:34
stgraber: you need conditional blocks
20:37
<stgraber>
warren: return seems to work
20:38
<warren>
stgraber: without disrupting the chain of scriptS?
20:38
<stgraber>
warren: yes
20:39
I wrote 3 scripts: test.sh sourcing test1.sh and test2.sh, test1.sh containing an echo, return and another echo. Then test2.sh contained another echo
20:39
I saw the first echo of test1.sh and the echo of test2.sh
20:39
so the execution of test1.sh stopped with the return and test2.sh was still executed
20:40
<warren>
ok
20:40
<stgraber>
fixing the rc.d scripts now.
20:41
<warren>
I was under the impression that . means "include"
20:41
return is normally for a {} block
20:41
<vagrantc>
. == source in shell
20:41
<Ryan52>
it does.
20:41
<stgraber>
I'm also completing the gettext on these, in your case it means they'll now all break (well, these using ldm-dialog) but at least it'll be consistent and we'll be able to fix them all at once without leaving some untranslated
20:41
warren: fine with that ?
20:41
<warren>
stgraber: the first script will gettext.sh once?
20:42
stgraber: I'll test Ryan52's tiny binary when it becomes ready
20:42
<stgraber>
hmm, in fact, shouldn't we source gettext.sh in one of the common scripts ?
20:42
<Ryan52>
okay, nevermind, this feels too stupid...I'm too tempted to just copy all of the pieces from gettext into ldm-trunk, which is stupid. why not just switch to #!/bin/bash and suck it up about the warnings? :p
20:42
<stgraber>
ltsp-common-functions or similar ?
20:43
<Ryan52>
I can rewrite it, but seeing as how I'm looking at code that does it, that feels even stupider...:)
20:43
can we just switch the shebang to bash, use $"", and suck it up about the warnings?
20:44
<stgraber>
Ryan52: well, I don't really like the idea of using something that's marked as deprecated :) because these thing tend to disappear in the next release or so :)
20:44
<Ryan52>
then we'll see what happens to Fedora's init scripts and copy them :)
20:44
<warren>
stgraber: unlikely, given that our initscripts use it
20:45* vagrantc suspects this is a years-long stalemate weather to finally deprecate it or remove the deprecation
20:45
<stgraber>
Ryan52: can you confirm that TEXTDOMAIN and TEXTDOMAINDIR are correctly working with bash ?
20:46
<Ryan52>
...
20:46
how else would it be finding the translation?
20:46
<warren>
I would like to note that using bash might not actually be slower, because it calls ZERO external processes.
20:47
<Ryan52>
but ya, setting it to something nonexistant puts back the English output.
20:47
so confirmed.
20:47
<vagrantc>
though would be slower in most of the scenarios where that code is never even called
20:47
<stgraber>
warren: well, bash itself is slower, that's why Debian and Ubuntu are using dash so ...
20:47
<vagrantc>
debian hasn't yet switched to dash, but it's working towards it...
20:47
<Ryan52>
Debian doesn't use dash by default.
20:48
<stgraber>
Ryan52: oh, really ? I thought they did the switch as well.
20:48
<Ryan52>
nope.
20:48
tho all #!/bin/sh have to work with dash.
20:48
<warren>
Ryan52: our initscripts have TEXTDOMAIN and TEXTDOMAINDIR
20:48
<vagrantc>
stgraber: it's been a release goal to make it possible, and i think nearly all bugs have been fixed with that.
20:48
<Ryan52>
warren, I know...I was saying it was working fine.
20:48
<warren>
If you folks are against a tiny binary or using bash for this, then I'm just going to write a blank eval_gettext.
20:49
because I really don't care about this.
20:49* vagrantc wonders if we still need to have ldm-trunk/share/ldm-script.in anymore ...
20:49
<stgraber>
http://www.gnu.org/software/automake/manual/gettext/bash.html
20:50
<Ryan52>
ah, I see.
20:51
<warren>
Are you against bash?
20:52
<Ryan52>
it's a security hole...
20:52
<warren>
how the hell is it a security hole?
20:52
<Ryan52>
warren, did you read that page?
20:52
it might help.
20:52
:)
20:52
<vagrantc>
particularly point 2
20:53
<kc8pxy>
johnny: more success :) i have my ltsp working with my local but seperate dhcp/dns server, and it's the same dhcp software i use at the production site :)
20:53
<stgraber>
how easily can one get a translation in fedora ? if it's easy, it basically means anyone can run any command as root on translated systems
20:54
<warren>
point #1 is not an issue for us
20:54
<stgraber>
if I understand that page right
20:54
<warren>
point #2 is an issue if we are really not paying attention
20:54
stgraber: if we're THAT paranoid, then I would add a script that greps for that in the .po files and errors out
20:55
<kc8pxy>
the ltsp nfs and tftp are on one machine, and the dnsmasq is playing dhcp on the gateway :)
20:56
how do i dynamically adjust the xorg display size?
20:57
<Ryan52>
warren, how can you write a blank eval_gettext?
20:57
<stgraber>
warren: can we do something like checking for gettext.sh, if not found then define eval_gettext as an alias of $"" ?
20:57
<Ryan52>
it needs to parse the variables..
20:57
that makes sense.
20:57
and is easy to do.
20:57
<stgraber>
then it won't affect Debian and Ubuntu as we'll just ship the 53k package and fedora will use the insecure bash thingy
20:58
<warren>
stgraber: I don't think bash can do that.
20:58
<Ryan52>
warren, why not?
20:58
<stgraber>
eval_gettext() {
20:58
echo $"$*"
20:58
}
20:58
<Ryan52>
if test -f /usr/bin/gettext.sh; then
20:58
<stgraber>
?
20:58
<Ryan52>
oh, crap, stgraber got there first :P
20:58
but ya, that's what I was gonna say :)
20:59
<stgraber>
that defined in some generic place like ltsp-common-functions so we don't have to do it for every single rc.d script
20:59
<kc8pxy>
arbitrary code execution?
21:00
<stgraber>
kc8pxy: well, you have to get that code in a .po first, not that arbitrary
21:00
<warren>
stgraber: hm, that's pretty clever if it works.
21:00
<stgraber>
though in Ubuntu's case with Launchpad and our language packs, yes
21:00
<warren>
kc8pxy: arbitrary as in you would have had to build it into the code...
21:00
<stgraber>
warren: well, that's with considering $"" as some kind of function but it may work :)
21:01
<kc8pxy>
warren: so you would have to bypass dm5 and those checksums,
21:01
md5
21:01
<stgraber>
warren: you could translate the string on LP, then wait for the next langpack and get your code, but it'd take a few months and would very likely be reviewed :)
21:01
<warren>
stgraber: how do you conditionally define a function?
21:02
<vagrantc>
if -x gettext.sh ; the . gettext.sh ; else eval_gettext() { ...
21:02
though you'll need to verify that the shell you're running supports $""
21:03
<stgraber>
warren: http://pastebin.com/f1ede08be
21:04
<warren>
vagrantc: if it doesn't support $"" the worst thing that happens is untranslated tsrings
21:04
<stgraber>
warren: I don't have any langpack installed but it seems to do what it's supposed to do
21:05
warren: as in, not having gettext.sh, I get "test" as output and not "$test" or an error message
21:05
warren: no, what will happen will be "$your string here"
21:05
I tried it with dash
21:05
<Ryan52>
stgraber, http://slexy.org/view/s20TaU94Yh
21:06
since there isn't any docs on that, and the error message lies, and I don't speak python, how do I actually do that?
21:06
(do that == use a config file)
21:06
(yes, off topic here, sorry.)
21:07japerry has joined #ltsp
21:08
<stgraber>
warren: http://pastebin.com/f5cf6b4f2
21:08
that's for the version with bash detection
21:09
so it detects if you have gettext.sh, if not it detects if you're using something that works with built-in translation
21:09
<Ryan52>
echo $"$*" shouldn't work.
21:09
err, I meant the echo "$*" one
21:09
in dash
21:09
because it won't substitute variables.
21:10
<stgraber>
Ryan52: right, that shouldn't be echo actually but eval or something like that :)
21:10
<Ryan52>
yes.
21:10
<stgraber>
Ryan52: that was more of a proof-of-concept that we can call that built-in and detect if the shell supports translation
21:11
<Ryan52>
yes, I know ... I just wanted to make sure that it didn't get left like that and included.
21:11
<stgraber>
(it's actually my ~/test.sh that's usually my test-crazy-things-that-shouldnt-work script)
21:12
Ryan52: do you have some time to do the real implementation ? I have to prepare a school for a ltsp-cluster demo tomorrow :)
21:13TheProf has joined #ltsp
21:13
<Ryan52>
yours should work fine, with the change I already said..
21:13
so I guess you just want me to test it then?
21:13
<warren>
I seriously confused.
21:14
<TheProf>
Hello. Hope everyone is well. Sorry if this is a very basic question, but how do I know what the IP / WS number of a client is?
21:14
<stgraber>
Ryan52: what release of pastebinit is that btw ? I'm not the one who implemented it and I don't use it actually. Though someone told be he was working on improving it and actually making it useful :)
21:14
<warren>
Ryan52: could you add what you think works to the code and I'll test it tomorrow?
21:14
I was under the impression that you couldn't conditionally define a function.
21:15
<stgraber>
Ryan52: I'm actually looking for someone to find the right place to put it (probably in ltsp-common-functions or ldm's equivalent (the rc.d launcher ??)) so I can just commit my modified rc.d scripts and have them working :)
21:15
<Ryan52>
stgraber, uhhh...then how did you get involved in it? 0.o
21:15
stgraber, it's 0.11-1 from debian experimental.
21:15
<vagrantc>
warren: don't know where you got that impression from...
21:15
<stgraber>
Ryan52: I was talking of the .xml feature :)
21:15
<Ryan52>
oh, ok :p
21:15
<warren>
vagrantc: what impression?
21:16
<vagrantc>
19:15 < warren> I was under the impression that you couldn't conditionally define a function.
21:16
<warren>
vagrantc: sh seems to ignore scope for lots of things
21:16
<stgraber>
Ryan52: I'm the initial coder of pastebinit but haven't really played with all the .xml thingy yet (though I did some for italc so can probably just use some well-tested functions from it)
21:17
<Ryan52>
I just want it to default to slexy.org so I can destroy my stupid slexy.rb script.
21:18* Ryan52 was so excited to see that The_PHP_Jedi's feature request got included, and then realized that it wasn't (apparently) usable :(
21:19
<TheProf>
I'm just stuck trying to get a printer to work off a client. I can print perfectly when it's plugged into the server directly, but not when it's on a client so I'm worried I have the wrong client specified!
21:20
<Ryan52>
eval_gettext "User ${LDM_USERNAME} is not allowed to log into this workstation."
21:20
stgraber: how does that work?
21:21
bash will substitute the variable in before calling eval_gettext, won't it?
21:21
which means that gettext won't find anything matching that, right?
21:21
we need to use single quotes..
21:23
<warren>
Ryan52: stgraber: Especially those strings that contain other strings... it is approaching stupid to try to support it from shell scripts. It would actually be cleaner to write special modes of ldm-dialog.
21:23
<Ryan52>
what do you mean by "that contain other strings"?
21:23
<warren>
Ryan52: "User ${LDM_USERNAME} is not allowed to log into this workstation."
21:24
<Ryan52>
eval_gettext takes that into account, by getting the translation, then doing the variable processing
21:24
gettext doesn't. it just gets the direct translation.
21:24
so it's not stupid, it's intended behavior.
21:24
<warren>
the .po file has a place to leave the variable?
21:24
because the variable wont always be the second word
21:24
<Ryan52>
I think it'll just have "User ${LDM_USERNAME} is not allowed to log into this workstation." in the .po file.
21:24
<warren>
for example
21:25
<Ryan52>
tho stgraber is the xgettext guy now :)
21:25
<warren>
Ryan52: stgraber: is one of you adding the conditional gettext.sh?
21:26
<Ryan52>
I am.
21:26
<stgraber>
Ryan52: hmm, current bzr branch actually seems to work with the .xml but paste.debian.net doesn't work now ...
21:26
<warren>
thanks
21:26
<Ryan52>
stgraber, with what .xml file?
21:26
the one in the error message?
21:26
<stgraber>
Ryan52: the one you pastebined
21:27
<Ryan52>
okay, thanks.
21:28
<stgraber>
Ryan52: http://slexy.org/view/s214k31Qzv
21:28
Ryan52: I'd have to check what changed on paste.debian.net though as something broke there ...
21:28
<Ryan52>
looks like revision 61 fixed it...planning on making a new release soon?
21:29
<vagrantc>
ltsp-trunk/po/ltsp.pot doesn't have any variable names, only %s substitutions
21:29
<Ryan52>
vagrantc: that's cause it's c code..
21:29
or does it have the "is not allowed to log into this workstation." already?
21:29
<stgraber>
Ryan52: so in whatever script you put that eval_gettext function, can you make sure you also have both TEXTDOMAIN and TEXTDDOMAINDIR exported ?
21:29
Ryan52: then I'll just drop these from the rc.d scripts and commit my chnages (also drop the evil exits)
21:29
<vagrantc>
Ryan52: we used printf for those strings...
21:30
<warren>
stgraber: can you make sure you also have both TEXTDOMAIN and TEXTDDOMAINDIR exported ? huh? where is it now?
21:30
stgraber: not?
21:30
<Ryan52>
vagrantc, ...in shell?
21:30
<vagrantc>
Ryan52: it's all in shell code, but using printf
21:30
<Ryan52>
o.
21:30
weird.
21:30
why didn't we think of that? :p
21:30
<vagrantc>
Ryan52: ltsp-trunk/server/plugins/ltsp-build-client/common/001-set-arch, for example.
21:30
<stgraber>
warren: currently it's in all rc.d scripts but that should be in a common script
21:31
<Ryan52>
if we do it that way, the simple binary is actually really, really simple...no stolen code or anything. :)
21:31
but still, ya.
21:31
stgraber, okay, but where does it get TEXTDOMAIN from?
21:31
<warren>
where does that printf come from?
21:31
<Ryan52>
or what should TEXTDOMAIN be?
21:31
<warren>
Ryan52: if you look in ltsp-trunk/server/plugins/ltsp-build-client/common/001-set-arch it still uses eval_gettext
21:31
<vagrantc>
warren: coreutils
21:32
<stgraber>
Ryan52: ldm probably or ldmrc if we want a different domain
21:32
<vagrantc>
warren: /usr/bin/printf
21:32
<Ryan52>
warren: it could just be using gettext tho.
21:32
<warren>
Ryan52: what do you mean?
21:32
<Ryan52>
it doesn't need eval_gettext...it just needs gettext.
21:33
<warren>
Ryan52: which is in the same package as gettext.sh here...
21:33
<Ryan52>
my point is that if we used all printf's in the shell scripts, then the simple binary idea would actually be 10 lines, no stolen code or nothin.
21:33
<warren>
to replace which part?
21:33
<Ryan52>
what do you mean which part?
21:34
<warren>
if you use coreutils printf
21:34
<stgraber>
hmm, what's the problem with my shell script ? why do we need a binary again ? :)
21:34
<Ryan52>
and a simple binary we create.
21:34
<warren>
Ryan52: ltsp-trunk/server/plugins/ltsp-build-client/common/001-set-arch uses printf with eval_gettext
21:34
<Ryan52>
then we need nothing.
21:34
stgraber, it requires either bash or gettext.
21:34
:)
21:34
<warren>
requiring either bash or gettext is fine
21:35
<Ryan52>
I know.
21:35
<warren>
Just pick something and stick to it!
21:35
=)
21:35
<Ryan52>
I wasn't suggesting we go back.
21:35
<warren>
I don't understand what you mean by printf though
21:35
<stgraber>
+1 for the script thingy
21:35
<Ryan52>
yes, we have all agreed to use the script.
21:35
<stgraber>
if we feel the need, we can still write a eval_gettext in C afterwards :)
21:36
<Ryan52>
19:32 < stgraber> Ryan52: ldm probably or ldmrc if we want a different domain
21:36
so which one?
21:36
you're currently putting it into ldm with the xgettext stuff, right?
21:36
<stgraber>
I found it hard to merge the .pot generated from the C code and the one from the rc.d/ so ldmrc would be easier for me :)
21:36
<Ryan52>
okay, then we'll go with that.
21:36
<stgraber>
no, I haven't managed to make xgettext to merge them yet
21:37
<Ryan52>
now why do we need to export TEXTDDOMAINDIR?
21:37
doesn't it have reasonable defaults?
21:37
(I didn't set it in any of my tests, and they all worked fine..)
21:37
<TheProf>
Maybe I'll start with a more basic question -- do you need to be logged into a client to get the printer attached to that client to print?
21:38
<johnny>
yes
21:38
<Ryan52>
okay, code works, now where do I put it?
21:39
<TheProf>
johnny: so it won't just print if it is booted up and waiting at the login prompt?
21:39
<johnny>
well.. technically.. you could point to the jetpipe by port on that ip..
21:39
unless it only listens over the ssh connection that is..
21:39
<TheProf>
johnny: I was following the instructions given here: http://wiki.ltsp.org/twiki/bin/view/Ltsp/Printers
21:39pscheie has quit IRC
21:39pscheie has joined #ltsp
21:40
<TheProf>
I attempted to point the server to the printer using the jetdirect feature
21:40
<johnny>
oh.. then it should work
21:40
but.. you would have known that.. if you would have tried to login and do it before asking the question
21:40
does it even work from a different machine even if logged in?
21:40* johnny imagines not
21:40
<stgraber>
Ryan52: do we use ldm-script.in ?>
21:40
<TheProf>
johnny: that's the part that's got me stumped. I had it working when I plug the printer directly into the server no problem.
21:41
<stgraber>
Ryan52: if so, it may be the right place
21:41
<johnny>
sure.. but i bet your problem has nothing to do with being logged in
21:41
check bug reports perhaps
21:41
<TheProf>
I was grasping at straws :)
21:41
<warren>
I'm incredibly exhausted.
21:41* johnny realises he needs to step away for a few
21:41* warren sleep
21:41
<TheProf>
All the instructions on the wiki are straightforward hence my confusion. I'll check the cups log now
21:42
<stgraber>
Ryan52: about TEXTDOMAINDIR, I have no idea, it was part of the gettext example :)
21:43
<Ryan52>
pushed
21:44
or pushing..
21:44
there.
21:45
<stgraber>
I guess that for po/ I'll just revert to using autotools (as it does some magic with the C code) and will put a script to generate the template and update the .po
21:46
any problem with that ?
21:46
<Ryan52>
not from me.
21:49
<warren>
stgraber: wait, you're relying on autotools to read the source files and write .pot?
21:49
<TheProf>
johnny: so as far as I can understand the cups error log, nothing's wrong. it is listing all the jobs as completed
21:49
<warren>
stgraber: I think some (most?) projects separate that entirely...
21:50
<TheProf>
But nothing's coming out of the printer - it's sitting there staring at me waiting.
21:50
<stgraber>
warren: it's how it was done in ldm
21:50
<warren>
stgraber: they run their own script separate from autotools to do gettextize and spit out the .pot updates
21:50
oh
21:50
unless you mean the part where it takes .po and compiles .mo
21:50
<stgraber>
warren: I don't like it but that Makefile does some magic I don't really understand and don't want to check now
21:51
<warren>
source -> .pot shouldn't be part of the usual autotools build
21:51
<stgraber>
warren: we have: make ldm.pot, make update-po and make all done by autotools at the moment
21:51
<warren>
.po -> .mo should
21:51
<stgraber>
warren: so I basically can't change that Makefile because it's all generated by the autogen and the configure ...
21:51
that's why I'll just write an external script that'll do the job for ldm rc.d for now
21:51
<warren>
grr
21:52
ok
21:59
<TheProf>
So I just tried it again and it shows, after a long wait, that the job is complete!
21:59
But nothing is coming out, and the error log shows no errors.
22:00
<stgraber>
ok, just pushed the fixed rc.d scripts
22:00
next task on my list is to write that script for po/
22:00
<Ryan52>
which scripting languages would people be okay with me adding to the build dependancies? for generating the translations for language and territory, I don't want to bother doing it in C...I can handle perl or ruby. I would do shell, but it can't parse xml...I would be fine with using xslt, tho, if you prefer me to use that.
22:01
(xslt + bash as opposed to perl or ruby)
22:01
s/bash/sh/
22:02japerry has quit IRC
22:02japerry has joined #ltsp
22:03phoenix has joined #ltsp
22:03nothingman has quit IRC
22:04nothingman has joined #ltsp
22:05CaScAdE^FarAway has joined #ltsp
22:05
<Ryan52>
I guess it doesn't even need to do it on build...we can just generate it and put it in bzr.
22:06
<stgraber>
Ryan52: python is already a dependency
22:07
(for jetpipe)
22:07
<Ryan52>
yes, but my brain doesn't speak python :)
22:07* Ryan52 will figure it out
22:07
<TheProf>
Sooo...is there something else I should do to trouble-shoot this? :S
22:15
I went through the mailing list archives and found this: "...on K12Linux (Fedora) we forgot to include pyserial in the chroot, which jetpipe requires, resulting in symptoms similar to yours. Adding it fixed the problem." The problem was the system reporting completed printing but nothing comes out. But I have no idea how to add pyserial.
22:17
<Ryan52>
no clue if that's your problem, but in Debian/Ubuntu, it looks like those come from the python-serial package.
22:18CaScAdE^1arAway has quit IRC
22:19
<TheProf>
Ryan52: OK at least I am a bit reassured that it's not something totally unique here.
22:27phoenix has quit IRC
22:28
<Ryan52>
warren, hrm. it's still gonna be 3 megs of stuff with the translations (for including the language name and territory translated into everything).
22:28
warren, so by regenerating it we aren't helping much.
22:29
<warren>
Ryan52: huh!?
22:29
<Ryan52>
warren, what do you think about pushing only the translated name and territory over the ldminfod?
22:29
one sec...is that in bytes or kilobytes? :p
22:29
<warren>
can't possibly be 3MB
22:29nothingman has quit IRC
22:30
<Ryan52>
warren, http://pastebin.com/f3a9a4654 http://pastebin.com/f72cb27a7
22:30TheProf has left #ltsp
22:30
<Ryan52>
that's the current sizes of the stuff from the package.
22:30
all we're gonna be doing is mucking with it..
22:30
<warren>
is that bytes or kilobytes?
22:31
oh
22:31
hm
22:31
<Ryan52>
kilobytes, I think.
22:31
<warren>
KB
22:31
<Ryan52>
my idea is to push it over ldminfo, but only in the language itself.
22:31
would that still be too much?
22:31
<warren>
273 * <how many bytes per language?>
22:32
on average
22:32
<vagrantc>
i tried with 256, and it was about 16Kb
22:32
<Ryan52>
language-with-name:en_US:English (United States)
22:33
<warren>
This is entirely to avoid parsing the xml
22:33
on the client side
22:34
<Ryan52>
yes.
22:34
<warren>
Ryan52: I guess that's OK... but let's come up with a more reasonable way to filter languages
22:34
<Ryan52>
well, that much data into the chroot is bad.
22:34
<warren>
locale -a is ureasonable
22:34
<Ryan52>
Fedora is unreasonable for that one, imho :)
22:34
<warren>
Fedora is doing the standard glibc behavior
22:34
<Ryan52>
pfft. standard.
22:34
<warren>
Debian must have mucked with it.
22:35
<Ryan52>
anyway, is there a reasonable way to filter languages?
22:35
is there some way to know what the user actually wants?
22:36
or do you mean adding settings to ldminfod somehow, or something complicated like that?
22:36* vagrantc wonders how many systems actually support more than a small handfull of langauges ...
22:37
<vagrantc>
in practice, enabling languages as needed seems much more sane.
22:37
<stgraber>
vagrantc: most of ours have two languages (so that's 10 or so locales) never had more on one box yet
22:38
<vagrantc>
stgraber: right.
22:39
there are some countries with 6 or more languages in common use...
22:39
<warren>
vagrantc: you mean we should define languages we want to have in the menu in a config file read by ldminfod?
22:40
<vagrantc>
warren: that's one possible way to do it ... though i don't really understand why 'locale -a' should ever output such an absurd number of locales.
22:41
<warren>
vagrantc: it was never meant to be used in this manner.
22:41
<vagrantc>
warren: maybe debian has patched 'locale -a' to only display available locales, but i can't see that as a bad thing.
22:42
<Ryan52>
vagrantc, but it's not standard! :p
22:42
<vagrantc>
warren: actually, i suspect debian hasn't patched locale, but rather generates desired locales at install time.
22:43
<warren>
oh, that might be it
22:43
vagrantc: we default to install all, you can exclude if you want, but nobody bothers.
22:43
<vagrantc>
which seems reasonable to me...
22:44
<warren>
still, it seems reasonable that a deployment will want to choose the languages they display
22:44
<vagrantc>
exactly.
22:46
debian typically only generates the locales the admin asked for, so that mechanism is typical practice on a debian system ...
22:47
<Ryan52>
so anyway...still no solution.
22:47
<vagrantc>
so an optional ldminfod-specific configuration file?
22:48
<Ryan52>
oh yay, I get to figure out how to make a python script parse a config file.
22:48
<warren>
273 languages isn't *THAT* much data over the wire
22:48
but how long does it take ldminfod to generate all that?
22:49
<vagrantc>
Ryan52: are you working on patches to ldminfod to get more distinct locale names?
22:49
Ryan52: i still get no distinction between es_AR, es_MX and es_US, for example ...
22:49
language-with-name:Spanish; Castilian:es_AR.UTF-8
22:49
language-with-name:Spanish; Castilian:es_MX.UTF-8
22:50
<Ryan52>
yes, I know. it will now add territory.
22:50
<vagrantc>
Ryan52: now, as in patches you have yet to commit?
22:50
<Ryan52>
and not require iso-codes in the chroot.
22:50
yes, not yet finished.
22:50
<vagrantc>
Ryan52: cool. :)
22:56hanthana has joined #ltsp
23:01
<Ryan52>
warren: real 0m0.501s
23:01
on my fedora box.
23:02
<warren>
half a second?
23:02
<Ryan52>
yep.
23:02
11951 characters.
23:02
<warren>
So it has to touch a lot of files every time ldminfod is read
23:03
Ryan52: half second is 2nd and subsequent runs after all files are already in memory?
23:03
<Ryan52>
0m0.318s
23:04
<warren>
I suppose it has to read a lot fewer files when we limit it with a config file.
23:05
<Ryan52>
0m0.159s on my Debian system, which has 3 languages..
23:05
wheras before I started to touch stuff it took 0m0.040s
23:05
<vagrantc>
no, all the languages and countries are defined in two files...
23:05
so limiting it won't speed it up
23:06
well, i suppose the output will be reduced some...
23:06
<Ryan52>
translating.
23:06
it's doing a gettext for each one.
23:06
each contry and each language.
23:06
<vagrantc>
oh, this must be a different implementation...
23:07* Ryan52 is on attempt number 3! :p
23:08
<Ryan52>
how do you do comments in python?
23:08
<vagrantc>
Ryan52: #
23:08
there are other options...
23:08
<Ryan52>
emacs doens't make it colored.
23:08
hrm.
23:09
<vagrantc>
you have to make sure your whitespace doesn't get borked when you comment stuff out ...
23:09
<Ryan52>
yes, I know.
23:09
okay, so is this pushable?
23:10* vagrantc has no idea what it even is...
23:10
<Ryan52>
0.o
23:10
<vagrantc>
Ryan52: new ldminfod variant of the week?
23:10
<Ryan52>
Traceback (most recent call last):
23:10
File "./ldminfod", line 227, in <module>
23:10
print lang_with_name(lang)
23:10
UnicodeEncodeError: 'ascii' codec can't encode character u'\xe5' in position 23: ordinal not in range(128)
23:10
what's that mean?
23:11
<warren>
Ryan52: you need a different data type for utf-8
23:11
<Ryan52>
this is why I shouldn't touch python :)
23:11
what do you mean data type?
23:11
<warren>
hm
23:11
<vagrantc>
integer, string, ..
23:11
<Ryan52>
language = gettext.dgettext("iso_639", langmap.get(langcode, ''))
23:11
I never said any type..
23:12
was I supposed to? :p
23:12
<vagrantc>
i guess so.
23:12
<warren>
Ryan52: you could paste that in #fedora-devel and the error message, lots of python devs there.
23:12
<Ryan52>
okay, so how do I do that?
23:12
<stgraber>
Ryan52: hang on a sec
23:13
Ryan52: try adding .decode('UTF-8') or .encode() I never know which one of these that's
23:13
<vagrantc>
Ryan52: prepend a "u" to your strings to indicate a unicode string ... maybe
23:13
language = gettext.dgettext(u"iso_639", langmap.get(langcode, ''))
23:13
maybe
23:14
or whichever string you expect to come out as unicode
23:15
<Ryan52>
Traceback (most recent call last):
23:15
File "./ldminfod", line 227, in <module>
23:15
print lang_with_name(lang).decode('UTF-8')
23:15
same thing.
23:16
<vagrantc>
encode
23:16
<Ryan52>
Traceback (most recent call last):
23:16
File "./ldminfod", line 226, in <module>
23:16
print lang_with_name(lang).encode()
23:18
and it gets quiet...
23:18
<vagrantc>
http://evanjones.ca/python-utf8.html
23:19
i'm wondering if you wouldn't want to encode the string as you inserted it?
23:20
<Ryan52>
do you want to try?
23:20
<stgraber>
Ryan52: .encode('UTF-8')
23:20
<vagrantc>
Ryan52: where's the code?
23:20
<Ryan52>
you need a Fedora system to test on, tho :p
23:21
wait, it's working.
23:21
no, it's not :p
23:22
vagrantc, do you have a Fedora vm you can test with? or can you enable all of the locales somehow?
23:22
(which would probably take too long..)
23:22
<vagrantc>
Ryan52: i could enable some unicode locales
23:22
some locales that depend on unicode characters, that is
23:23
<Ryan52>
I have it printing this on my Debian system fine: language-with-name:日本語 (日本):ja_JP.UTF-8
23:23
so I think unicode works..
23:23
<warren>
https://bugzilla.redhat.com/show_bug.cgi?id=480995
23:23
This bug makes no sense at all.
23:24
Ryan52: hmm, if ldm itself is in english mode will it print "Japanese" before that?
23:24
<Ryan52>
no.
23:24
is that a problem?
23:25
there's no way for it to get information about every single language over ldminfo without way more data going back and forth.
23:25
and does it even matter?
23:25
I'm guessing that if you can't read the language name, you probably don't want to log in with that language..
23:25
<vagrantc>
language selections are actually reasonable places to have break default locale settings, as a native speaker might recognize the native spelling of their language rather than the default locale's language
23:26
<warren>
Ryan52: yes, that is a problem.
23:27
hmm
23:27
<Ryan52>
well, then it's impossible.
23:27
if we can't have it all over the wire.
23:27
and you won't allow it all in the chroot.
23:27
then there is no way for ldm to know it in all those languages.
23:27
<vagrantc>
"
23:27* vagrantc notes that in the chroot *is* in fact, over the wire
23:28
<Ryan52>
-.-
23:28
<warren>
No, I was in favor of only LANG codes over the wire like the previous ldminfod and parsing stuff in the chroot, even if it is 8MB.
23:28
<Ryan52>
really? I could swear you had a problem with that yesterday..
23:29
so you can add 8mb for language names but not 5mb for gettext? 0.o
23:29
whatever.
23:29
in that case, current trunk is good, right?
23:29
<warren>
Ryan52: you have: language-with-name:日本語 (日本):ja_JP.UTF-8 working?
23:29
<Ryan52>
not really.
23:29
if you ignore the errors, then it might be working..
23:29
(can't tell, really)
23:29alkisg has joined #ltsp
23:29
<warren>
Ryan52: I was unhappy about trunk because it needs iso-codes on both sides which I found to be highly redundant.
23:30
<Ryan52>
ah, right.
23:30
xml parsing in with glib...hrm.
23:30
<warren>
Ryan52: I suppose give it a try with only native language display and we'll see how it looks
23:30
<Ryan52>
with which way? over the wire or on the chroot side?
23:30
<warren>
I would prefer sending it all over the wire and avoiding 8MB in the chroot.
23:31
<Ryan52>
ok.
23:31* vagrantc too
23:31
<Ryan52>
vagrantc, if I give you access to my Fedora system can you look at it?
23:31
<vagrantc>
Ryan52: sure.
23:38warren has quit IRC
23:38
<alkisg>
Would anyone be interested in allowing some network information to be passed in the kernel command line? E.g. kernel vmlinuz ro root-path=/opt/ltsp/i386 ? This would help avoiding a second dhcp request while clients boot (with pxelinux's IPAPPEND 3 or static IPs or whatever), and it would only require a couple of lines to change...
23:41
<vagrantc>
that's some weird output...
23:42
Ryan52: my favorite is language-with-name:Project-Id-Version: iso_639 3.2
23:43
<Ryan52>
0.o
23:43
oh, right, I remember that.
23:43
one sec. :p
23:43
or are you doin stuff?
23:43
<vagrantc>
i'm looking at my own copy
23:44
<Ryan52>
well, it needs to not do the dgettext if it gets back an empty string from the map thing.
23:48
<alkisg>
If, when the clients started, they automatically connected with ssh to the server as a restricted user, and this ldm_user only had ldm as a shell, wouldn't it be much easier? E.g. there wouldn't be any need for the client to have language packs installed...
23:48
<vagrantc>
alkisg: tricky to implement cleanly...
23:49
<alkisg>
I think the benefits would justify some extra work...
23:49
anyway :)
23:50
<vagrantc>
i'd be wary of it from a security standpoint ... restricted users that control everybody's login have a tendency to get attacked.
23:52
<alkisg>
If it only run ldm, and as a restricted user, how valneruble could it be?
23:52
*vulnerable
23:52
And it's not like it would be a passwordless restricted user...
23:53
It could only be accessed by the TCs
23:54
<vagrantc>
Ryan52: i'm not seeing the traceback anymore... did you fix that?
23:55
<Ryan52>
0.o
23:55
<vagrantc>
i definitely did see it when i first ran it
23:55
<Ryan52>
[ryan52@newbie ~]$ /ldminfod >/dev/null
23:55
Traceback (most recent call last):
23:55
File "/ldminfod", line 226, in <module>
23:55
print lang_with_name(lang)
23:56
<vagrantc>
when i run it without a redirect, i get no traceback...
23:57
<Ryan52>
oh, ok then...
23:58
so then it's fine?
23:58
piping into less I get nothing other than language stuff
23:58warren has joined #ltsp
23:58
<Ryan52>
$ /ldminfod | grep -v language
23:58
same thing.
23:59
<vagrantc>
piping i get the traceback
23:59* Ryan52 doesn't like python