[00:01] <nbd> [mbm]: what would be a good dummy ip for /etc/hosts when ppp is not up yet?
[00:01] <nbd> [mbm]: 1.1.1.1?
[00:02] <[mbm]> hmm
[00:02] <nbd> afaik this ip is not routed anywhere
[00:03] <nbd> and people are not likely to use it for their lan :)
[00:03] <[mbm]> the 1.x is an iana reserved net
[00:04] <nbd> how about 0.0.0.1?
[00:04] <[mbm]> likewise reserved but that'll probably work fine
[00:05] <groz_> pick something in the 172 private range that nobody uses
[00:05] <groz_> that way, if it leaks, it will stop at the first properly configure upstream router
[00:05] <[mbm]> I doubt that anyone is going to route 0/8
[00:13] <CIA-17> nbd * r4285 /branches/whiterussian/openwrt/package/ (ppp/files/ifup.pppoe pptp/files/ifup.pptp): fix dial on demand support (#436)
[00:15] <[mbm]> nbd: tested?
[00:15] <nbd> yeah
[00:15] <nbd> well, not fully
[00:15] <nbd> but the relevant part
[00:16] <nbd> i set the config for dial on demand and created a /tmp/resolv.conf
[00:16] <nbd> when i did a nameserver lookup, it started its connection attempts
[00:16] <[mbm]> :)
[00:43] <[mbm]> nbd: think your module patch earlier actually broke a few things
[00:45] <nbd> which one?
[00:45] <groz_> weree there more since last nite ?
[00:45] <[mbm]> still poking around to figure out the exact cause
[00:46] <nbd> what's the symptom?
[00:46] <groz_> what is the symptom ?
[00:46] <groz_> gmta
[00:46] <[mbm]> but it seems it's trying to build kmod-ip6tables_2.6.17-aruba-1_mips.ipk even though I never selected any of the ip6tables crap
[00:46] <[mbm]> here's the relevent line -
[00:46] <[mbm]> cp: cannot stat `.../openwrt/build_mips/linux-2.6-aruba/modules/lib/modules/2.6.17/kernel/net/ipv6/netfilter/ip*.ko': No such file or directory
[00:46] <[mbm]> make[4]: *** [.../openwrt/bin/packages/kmod-ip6tables_2.6.17-aruba-1_mips.ipk] Error 1
[00:46] <nbd> something missing ... fixing that...
[00:48] <[mbm]> ah, you forgot the KCONFIG
[00:48] <CIA-17> nbd * r4286 /branches/buildroot-ng/openwrt/package/kernel/modules.mk: disable kmod-ip6tables if netfilter ipv6 support is disabled
[00:49] <nbd> yes
[00:49] <[mbm]> umm yeah, that
[00:49] Action: [mbm] was just about to commit it
[00:49] <nbd> :)
[00:52] <groz_> nbd, that new wl driver you stuck in the other day, have you done any real testing at all with it ?
[00:52] <groz_> the prior one I found 2 issues, one was the google thing, the other was a centrino wouldn't associate with it
[00:52] <groz_> i'm just building now with the new one, wondering if you've actually done anything with it yet ?
[00:53] Action: [mbm] thinks we need to make the svn server actually attempt to compile code before accepting it
[01:02] <{Nico}> it seems base-files is not installed anymore in recent -ng builds
[01:02] <nbd> groz_: haven't done much with it. just basic testing
[01:03] <groz_> ok, well i'm gonna give it a big smoke run here then shortly
[01:03] <[mbm]> {Nico}: that would explain the ticket about the broken wgt634u
[01:03] <groz_> if it goes ok on this first unit, i8'll put it on 10 more and start a regressino running
[01:04] <[mbm]> hmm image I built 30 minutes ago is missing base-files
[01:04] <{Nico}> [mbm]: yeah, and my (same) problems with uml-2.6
[01:04] Action: [mbm] is going to start revoking nbd's commit privledges if this keeps up
[01:06] <[mbm]> hmmm
[01:06] <[mbm]> target is aruba-2.6 but it's building brcm-2.4 packages
[01:06] <[mbm]> something is wrong with the stamp system
[01:08] <[mbm]> hrm.. .kernel.mk is wrong
[01:12] <[mbm]> wiped bin/package/base-files-* and .kernel.mk
[01:12] <[mbm]> built properly after
[01:14] <{Nico}> [mbm]: yeah, it seems there's now an svn rev number in the package version
[01:15] <[mbm]> yeah I did that a few revisions back just to force it to rebuild base-files
[01:15] <{Nico}> the build process does not fail if a package to be installed is not found
[01:15] <[mbm]> hmm
[01:16] <crazy_imp> {Nico}: took a look at my changes at the download.pl?
[01:19] <{Nico}> crazy_imp: yep, but i haven't tested it yet
[01:19] <crazy_imp> ack
[01:21] <[mbm]> {Nico}: did you actually find an example of it working even if a package failed to install?
[01:24] <{Nico}> [mbm]: http://pastebin.ca/100245
[01:25] <[mbm]> mumble .. stupid thing
[01:25] <[mbm]> that's the problem with base packages .. it doesn't compile
[01:26] <[mbm]> so some of the rules don't apply
[01:26] <{Nico}> look at the rev numbers
[01:26] <[mbm]> that's exactly the problem
[01:26] <[mbm]> I know what caused that
[01:28] <{Nico}> svn info are called from different directories?
[01:29] <[mbm]> no, what happens is that the ipkg control files are generated
[01:29] <[mbm]> and once generated they're pretty much ignored
[01:29] <[mbm]> so you had built base-files when you were running 4129
[01:30] <[mbm]> but when it updated to 4273 nothing forced it to rebuild the ipkg control file
[01:30] <[mbm]> so it rebuilt base-files as 4129 again
[01:30] <[mbm]> there's the secondary error that when it failed to find the package it didn't bomb
[01:32] <{Nico}> changing the dependency on .prepared to .built should be enough for 1)
[01:32] <{Nico}> err
[01:33] <[mbm]> not quite
[01:33] <[mbm]> note that the calls to timestamp.pl ignore the ipkg directory
[01:33] <[mbm]> (-x ipkg)
[01:33] <[mbm]> that's part of the problem
[01:33] <[mbm]> and I could have sworn I wrote a patch
[01:33] <[mbm]> must not have commited it
[01:34] <groz_> you are just saving it so that all the other dists dont get it out first
[01:34] <[mbm]> heh
[01:42] <groz_> mbm, i take it you are working with the aruba target mostly these days no ?
[01:43] <[mbm]> groz_: good guess
[01:43] <groz_> ok, now the dumb question
[01:43] <groz_> does aruba make any cheap low end devices ?
[01:44] <[mbm]> they tried
[01:44] <{Nico}> groz_: seems more like a wish than a question :)
[01:44] <[mbm]> they ended up with something called an ap40
[01:44] <[mbm]> which is more like cheap quality than cheap price
[01:44] <groz_> lol, so they did half of the project, made poor quality stuff
[01:44] <groz_> didn't do the other half, low price
[01:45] <groz_> is what you are saying
[01:45] <[mbm]> well, the hardware is good
[01:45] <[mbm]> it's an atheros board
[01:45] <groz_> do they still sell it ?
[01:45] <[mbm]> but they put it in a really cheap case
[01:45] <[mbm]> with an integrated antenna
[01:45] <[mbm]> and oddly no power connector
[01:45] <[mbm]> -- requires poe
[01:46] <groz_> ewwww
[01:46] <groz_> that's back assedwards
[01:46] <[mbm]> needless to say I haven't bothered with an ap40 port
[01:46] <groz_> hmm it doesn't even show in the products page of the website
[01:46] <[mbm]> think it got pulled
[01:48] <{Nico}> [mbm]: can i change ipkg to break on "file not found" error?
[01:48] <[mbm]> {Nico}: I'm still looking through the source to see why it didn't
[01:48] <[mbm]> groz_: btw, the ap41 suffers almost exactly the same fate
[01:49] <groz_> i already saw, poe only
[01:49] <groz_> useless for what i do
[01:49] <[mbm]> same cheap case too
[01:49] <groz_> our philosophy is, power plugs already exist everywhere
[01:50] <groz_> for the data side of the equation, 'what part of the word wireless is not clear'
[01:50] <{Nico}> [mbm]: basically, it's: [ -f $pkg ] then pkg_install else echo "File not found" even if pkg_install handles such error
[01:50] <[mbm]> I don't know what the case is actually made of but it feels like that old fragile bakelite stuff that radios in the 60s were made out of
[01:51] <[mbm]> {Nico}: ouch, it's in the ipkg code? fix that
[01:52] <{Nico}> [mbm]: in all branches?
[01:52] <[mbm]> yeah, probably should
[01:53] <{Nico}> ok
[01:53] <[mbm]> an 'exit 1' should do it
[01:55] <{Nico}> ipkg_install_file: ERROR: File /home/users/nthill/OpenWrt/branches/buildroot-ng/openwrt/bin/packages/base-files-brcm-2.4_8-4273_mipsel.ipk not found
[02:03] <CIA-17> nico * r4287 / (3 files in 3 dirs): let ipkg fail when a package file to be installed is not found
[02:17] <CIA-17> nico * r4288 /packages/net/freeradius/ (. Config.in Makefile ipkg/): port freeradius to buildroot-ng
[02:19] <[mbm]> hmm .. damn this is annoying
[02:20] <[mbm]> the reason why I put $(REV) in the base-files is so that the /etc/banner would always be correct
[02:20] <[mbm]> without having to force it to rebuild base-files each time
[02:21] <[mbm]> but it's too stupid to know to rebuild the ipkg control file
[02:21] <groz_> put $(REV) into a tempfile, make base dependant on that
[02:22] <[mbm]> from it's point of view, none of the timestamps have changed so there's nothing to be done
[02:22] <groz_> rewrite it if the contents are wrong
[02:22] <groz_> hmm, it'll still be circular and wont resolve right
[02:24] <[mbm]> yeah, the bigest error is that it knows what version it's trying to build but it has no idea what version is in the ipkg control file
[02:25] <{Nico}> groz_: uml-2.6 works
[02:25] <groz_> i expect it should nico
[02:25] <groz_> i'm going to add a few more lines with 'network quick start' to the readme
[02:25] <groz_> eventually
[02:26] <groz_> for now, think it's safe to assume, folks building and trying uml
[02:26] <groz_> either know what they are doing, or, know how to read up at it's home site
[02:26] <[mbm]> not an assumption I'd make
[02:27] <groz_> well, then they wont brick any routers when they screw it up
[02:27] <groz_> they can kill -9 the uml
[02:27] <groz_> and try again
[02:28] <[mbm]> true
[02:28] <groz_> and actually thinking about it
[02:28] <groz_> may not be a bad idea for total newbs
[02:28] <groz_> reccomend they learn the concepts with a uml first
[02:28] <groz_> before they risk hardware
[02:29] <groz_> particularily true when i get the colinux build running correctly
[02:30] <{Nico}> groz_: this is a really nice addition, especially for package maintainers
[02:30] <groz_> i see maintaining and testing
[02:30] <groz_> it'll get a lot of use
[02:30] <groz_> and i also think, it'll get more folks looking at working on packages
[02:31] <groz_> if they can do the work/experiments without using thier one and only router
[02:31] <groz_> which they need for internet connection
[02:31] <[mbm]> hmm found one more obscure case in the above error -- make doesn't stop if it fails to build the ipkg
[02:32] <CIA-17> nico * r4289 /packages/libs/freetype/ (. Config.in Makefile ipkg/): port freetype to buildroot-ng
[02:34] <groz_> Nico, are you using the uml for anything yet ?
[02:42] <CIA-17> nico * r4290 /packages/libs/libao/ (. Config.in Makefile ipkg/): port libao to buildroot-ng
[02:43] <{Nico}> groz_: nope, focused on -ng porting but i plan to test -ng packages on it
[02:55] <CIA-17> nico * r4291 /packages/libs/libaudiofile/ (. Config.in Makefile ipkg/): port libaudiofile to buildroot-ng
[03:01] <CIA-17> mbm * r4292 /branches/buildroot-ng/openwrt/include/package.mk: add more ipkg validation checks
[03:01] <[mbm]> that should hopefully put an end to all those ipkg errors
[03:02] <groz_> question for you folks
[03:02] <groz_> i have a patch here for bridge code, that enables the ability to change the ethernet hardware address of the bridge
[03:02] <groz_> any objections to adding that to kernel patches ?
[03:03] <groz_> instead of bridge taking arbitrarily the last adapter inserted
[03:03] <groz_> you can nail it down to one of the adapters in there
[03:03] <groz_> and force it to keep using that one
[03:03] <[mbm]> thought the bridge took the first mac, but anywaysm I see your point
[03:03] <groz_> actually ya, it does
[03:03] <groz_> but if you remove the first
[03:03] <groz_> then it's got the second
[03:03] <groz_> with this patch in
[03:03] <groz_> you can pick one, nail it to that
[03:04] <groz_> then if you are removing and inserting adapters regularily
[03:04] <[mbm]> one of the things that has bothered me is the fact that the macs under openwrt are usually completely different than the ones stamped on the device
[03:04] <groz_> it's not always chainging
[03:04] <[mbm]> which I think I mostly fixed in rc5
[03:05] <groz_> yah, but using this example
[03:05] <groz_> iv'e got a box here on the desk, it's wired to a couple of switches
[03:05] <groz_> one on the 4 port, one on the uplink
[03:05] <groz_> they are running different ip segments
[03:05] <groz_> but upstream they do join again
[03:05] <[mbm]> no, I'm not diagreeing by any means, I think we should add the patch
[03:05] <groz_> so if the 4 port and the uplink are using same mac
[03:05] <groz_> it's a pain
[03:06] <groz_> this adds the option
[03:06] <groz_> doesn't force it
[03:06] <[mbm]> just pointing out that macs have always been an issue
[03:06] <groz_> just make 'ifconfig hw ether xxxxxx' work
[03:06] <groz_> yup, it's always been a pain/hassle with these boxes
[03:06] <groz_> and it gets much worse when you have a spider of wds connections between a half dozen of them
[03:07] <[mbm]> yay, the ipkg patches work (had to commit them to bump the svn revision so I could actually test them properly)
[03:07] <groz_> ok, i'll add that patch in on my next commit
[03:08] <groz_> it's actually a nice clean one, last time i checked, it applies equally well to 2.4 and 2.6
[03:08] <groz_> just adds one more ioctl to the bridge code
[03:08] <[mbm]> btw a side effect of the ipkg patch above is that after you svn update to that version it will be forced to rebuild packages that you've already built
[03:09] <[mbm]> (just a warning for those people that do all their work in the build_mips directory)
[03:09] <groz_> that's not necessarily a _bad_ side effect
[03:09] <[mbm]> right
[03:09] <groz_> i've got all my work soft linked in
[03:09] <groz_> and linked into 4 different build trees
[03:09] <groz_> soon to be 6
[03:09] <CIA-17> nico * r4293 /packages/libs/libart/ (. Config.in Makefile ipkg/): port libart to buildroot-ng
[03:09] <[mbm]> heh, same here
[03:10] <groz_> i haven't played much with 2.6 on the wrt yet
[03:10] <groz_> but it's coming real soon now
[03:10] <[mbm]> 2.6 on the wrt isn't much to look at
[03:10] <groz_> and i'm gonna build a colinux target for my own use
[03:10] <[mbm]> diag doesn't work and wireless doesn't work
[03:10] <groz_> dunno if you guys want it committed in or not
[03:10] <[mbm]> (plus there's no nvram which means that after flashing it'll get a /etc/config which will force it to 192.168.1.1)
[03:11] <[mbm]> but hey, it has a spiffy 2.6.x kernel, :P
[03:11] <groz_> well, i'm planning on taking a good look at the open driver
[03:11] <groz_> and seeing if i cant make a few things work with it
[03:11] <[mbm]> yeah, I plan on fixing the upgrade problems caused by the lack of persistant config storage
[03:11] <groz_> i'm getting very tired of fighting one bad bug in the broadcom drivers
[03:12] <groz_> the wds connections when bridged
[03:12] <[mbm]> heh
[03:12] <groz_> often send packets thru with the wrong ethernet macs in the headers
[03:12] <groz_> and it's persistent across EVERY revision of the drivers
[03:12] <[mbm]> 'd expect that from client mode but not wds
[03:12] <groz_> i've got some test case code here that exercises it, and counts errors
[03:12] <[mbm]> figures
[03:13] <groz_> by sending raw ethernet frames with payload sying who really sent it
[03:13] <groz_> and then comparing that to the output of the driver
[03:13] <groz_> so far, the best driver
[03:13] <groz_> is the one in the gl linksys release
[03:13] <groz_> lowest error counts overall
[03:13] <groz_> and by the time the day is over
[03:13] <groz_> i'll be able to tell how this latest wl does
[03:13] <[mbm]> nbd just changed the wl version about a day ago
[03:13] <groz_> yah, that's why i'm building up some wrts here
[03:13] <groz_> to run this test on them
[03:13] <[mbm]> ah
[03:14] <groz_> just cleaning up my scripts to work in this new setup
[03:14] <groz_> then i'll flash 10 of them
[03:14] <groz_> go for dinner
[03:14] <groz_> come back and see the results
[03:14] Action: groz_ not got high hopes this bug cured
[03:14] <CIA-17> nico * r4294 /packages/libs/libcli/ (. Config.in Makefile ipkg/): port libcli to buildroot-ng
[03:15] <[mbm]> hmm yeah, I don't think it is either
[03:15] <groz_> the long and the short of this one
[03:15] <[mbm]> companies really aren't in any rush to fix bugs
[03:15] <groz_> if you have 15 units, with an average of 3 wds each, all interconnected to bridge one big mesh
[03:15] <groz_> after 15 units, expect about 5% of packets
[03:15] <groz_> to route incorrectly thru the bridge
[03:15] <groz_> it goes up exponentially from there
[03:15] <groz_> as you add units to the mesh
[03:16] Action: [mbm] has a tv with an upgradable firmware and a few nasty firmware bugs .. you'd think that 3 revisions later they'd have fixed them (yes, I did my darnest to report them) .. nope
[03:16] <groz_> well, my _hope_ is the bug is in wl.o, and not the actual device firmware
[03:17] <groz_> if that's the case, i'll be HUGELY inspired to get working in the internals of the open source driver
[03:17] <groz_> and get this FIXED
[03:17] <[mbm]> every time I try to report the bugs I get the same crap about how I somehow caused the error by not following instructions correctly (which inclues giving me a new copy of the manuals) or that the bug will be fixed in the next rev
[03:18] <[mbm]> which leads me to believe that it'll never get fixed
[03:18] <groz_> mbm, there is a reason a project like this gains steam after a while
[03:18] <[mbm]> :)
[03:18] <groz_> if your tv was linux and open
[03:18] <groz_> those bugs would be long fixed
[03:18] <groz_> example, local computer stores here in vancouver
[03:18] <groz_> they dont carry linksy v5 anymore
[03:18] <groz_> all carry gl
[03:18] <CIA-17> nico * r4295 /packages/libs/rrdtool-1.0.x/ (. Config.in Makefile ipkg/): port rrdtool-1.0.x to buildroot-ng
[03:18] <groz_> and, have cd's with various firmwares for the taking when you buy one
[03:19] <groz_> the sales staff will tell you, it's the router of choice for home use
[03:19] <[mbm]> hell, if they didn't encrypt the tv firmware I'd have probably taken a crack at it
[03:19] <[mbm]> but no, those bastards put high grade encryption on the damned thing
[03:19] <groz_> which tv is it ?
[03:19] <[mbm]> philips
[03:19] <groz_> hmm, i have a contact in phillips
[03:20] <groz_> very well placed too, but he retired last year
[03:20] <groz_> an uncle
[03:20] <[mbm]> it actually is encrypted -- service manual even tells you that
[03:20] <[mbm]> and it's not just some trivial xor crap either
[03:21] <[mbm]> :/
[03:22] <groz_> had a very interesting talk about encryption/drm etc when i was out flying the other day
[03:22] <groz_> the person chartering the plane, a gal with quite a selection of albums out
[03:22] <groz_> a couple of them are platinum
[03:26] <CIA-17> nico * r4296 /packages/ (17 files in 17 dirs): change SECTION and CATEGORY to put packages in the right place, change PKG_BUILDDEPs to use package-name instead of source-name (seems to work best with WR sdk).
[05:03] <CIA-17> nico * r4297 /packages/net/ez-ipupdate/ (. Makefile): port ez-ipupdate to buildroot-ng
[05:05] <CIA-17> nico * r4298 /packages/net/ez-ipupdate/ (Config.in ipkg/): oops, forgot to remove unneeded files for -ng from my last commit
[05:19] <CIA-17> nico * r4299 /packages/libs/ (8 files in 8 dirs): remove ref to host include and lib dirs from pkgconfig files
[05:22] <CIA-17> nico * r4300 /packages/net/nsca/ (. Makefile): add send-nsca to buildroot-ng
[10:08] <florian__> ola
[10:10] <groz_> good morning
[10:15] <groz> not complaining, it's getting late, i had a very rude early start this morning
[10:15] <groz> damn phone
[10:16] <groz> so just fiddling with a little 'fun stuff' before i call it quits
[10:16] <florian__> lol
[10:17] <groz> well, i guess, that is if you call setting up network tunnels for a test network 'fun'
[10:17] <groz> I'm scripting the bring up of a few umls with openwrt, then set up some tunnels between them over pretend networks
[10:18] <groz> building it up so that it can all be driven by a script set, bring them up from fresh build, install the packages, install configs, then start bringing up and testing tunnels
[10:20] <florian__> has anyone ever backported a netfilter module from 2.6 to 2.4 ?
[10:20] <groz> i have no clue
[10:20] <groz> what do you want to backport ?
[10:23] <Kaloz> http://ars.userfriendly.org/cartoons/?id=20060726
[10:23] <Kaloz> :D
[10:24] <florian__> SIP connection tracking
[10:25] <groz> I thought i've seen that in 2.4 at some point
[10:27] <florian__> I merged it into whiterussian, but its licence was unclear
[10:27] <Kaloz> how could it be unclear?
[10:27] <Kaloz> it's gpl
[10:27] <florian__> I think you know better than me because you told me I had to remove it, don't you remember ?
[10:28] <florian__> the 2.6 sip connection tracking is GPL, the 2.4 one was taken apart
[10:28] <Kaloz> remove in that way :p add gpl header
[10:28] <Kaloz> iirc this happened with some of those in the netfilter svn as well
[10:28] Action: [mbm] mumbles .. is today "bad advice day" on the forum?
[10:28] <Kaloz> all of those are derivative works
[10:29] <florian__> I don't understand a word of what are you writing :p
[10:29] <Kaloz> florian__: gpl, search for derivative work
[10:30] <florian__> Kaloz: ok so you mean that some netfilter modules were integrated as GPL simply because the source was available or ?
[10:30] <Kaloz> The GPL implies that a derivative work is one that is linked, statically or dynamically, with the original work.
[10:31] <florian__> right, so ?
[10:31] <Kaloz> this is the reason why legally the broadcom wl.o is violating the gpl, as it doesn't have the "gpl'ed glue" like the ati, nvidia drivers and madwifi does
[10:31] <Kaloz> but that's the grey zone
[10:31] <Kaloz> with netfilet/iptables
[10:32] <Kaloz> additional modules are stricktly go into gpl
[10:32] <Kaloz> no matter what crap does a company write into the headers
[10:32] <murb> Kaloz: if you take a fundamtalist approach then the madwifi stuff is equally as illegal as the wl.o
[10:33] <florian__> Kaloz: so what's the point with the SIP connection tracking module I merged few months ago ? you mean that adding the GPL header back would be sufficient for it to be integrated back again ?
[10:34] <murb> florian__: which sip conntracking module that there have been atleast 2
[10:34] <Kaloz> florian__: remove the propietary header if exist, and add a gpl header that states where did it came from and why it's gpl
[10:34] <murb> ?
[10:34] <murb> florian__: you'd probably be on pretty safe ground if you took the patch posted to netfilter-devel
[10:34] <florian__> murb: was it a 2.4 or 2.6 module ?
[10:35] <murb> florian__: 2.6
[10:35] <murb> there was an early implementation from some italian guys iirc.
[10:35] <murb> for 2.4
[10:35] <murb> s/early/earlier/
[10:35] <florian__> ok
[10:36] <murb> the 2.6 on has been merged into 2.6.18 iirc.
[10:36] <florian__> that's why I say I would rather backport the 2.6 module, rather than using a 2.4 one that was posted before
[10:36] <florian__> the 2.6 is mature and stable
[10:36] <murb> florian__: problem is netfilter has changed a *lot* since then.
[10:36] <florian__> well, playing with existing connection tracking module seems to be understandable
[10:37] <florian__> and there are few macros, and types that changed regarding module development
[10:39] <florian__> {Nico}: did you test the au1000 port ?
[10:40] <florian__> well I don't know whether it's worth actually
[10:48] <[mbm]> florian__: if code was released as GPL and later modified, it's still GPL code and any headers added that claim otherwise can be removed
[10:49] <florian__> ok, maybe I should check then
[10:50] <florian__> https://dev.openwrt.org/browser/trunk/openwrt/target/linux/generic-2.4/patches/613-netfilter_nat_sip.patch?rev=3291
[10:50] <groz> actually mbm, the _original author_ can take something, and for example, go gpl->bsd license with a later revision
[10:50] <groz> and if you wanna stay technically pure, you go back to the last gpl release
[10:50] <groz> and fork from there
[10:51] <[mbm]> groz: yeah, the original other maintained copyright and can do whatever
[10:51] <groz> but, if joe smith takes the code
[10:51] <groz> they cannot do the same
[10:51] <[mbm]> right, in this case it sounds like broadcom has just blindly added useless headers
[10:52] <groz> oh, didn't you know, that's an automated process
[10:52] <groz> take code, strip comments, add new copyrights, feed it thru a beautifier to change th e'style'
[10:52] <groz> and call it 'my original work'
[10:52] <[mbm]> I didn't expect them to open every file by hand and type out the new header
[10:54] <[mbm]> Kaloz: we need better spam filters on openwrt-devel@
[10:56] <Kaloz> [mbm]: greylisting + razor + rbl + spamassasin is there...
[10:56] <[mbm]> yeah, but I'm still getting blatent spam :/
[10:56] <groz> kaloz, are you feeding the filter with a big batch of spam every day to train it ?
[10:57] <Kaloz> groz: no, but that's pretty useless if you have razor
[10:58] <groz> i set up a couple of common names, jane and frank
[10:58] <Kaloz> but we can give it a try later :)
[10:58] <groz> anything arriving for them, goes strait into the filters as 'training mateiral'
[10:58] <groz> it works pretty slick
[10:58] <groz> they get LOTS of spam
[11:08] <groz> mbm, whats your thoughts on this, building images for regression testing things
[11:08] <groz> should it start with a 'stock bare' image, and scripts intall packages
[11:08] <groz> or should we build images with a list pre-installed for the regression ?
[11:09] <groz> i'm hedging toward working with a stock bare starter image
[11:09] <groz> and then script the installation of packages as needed for testing
[11:09] <[mbm]> bare, then you catch dependancy issues
[11:16] <groz> I need a proper regression system for a couple of projects, so looking at some ideas here, and looking to roll this all into one setup
[11:16] <groz> thinking of re-working some of the existing scripts, so they actually populate database tables with results
[11:17] <groz> for a lot of things testing various packages and so on
[11:17] <groz> that coud be scripted all on one machine, thru umls, and then post results into a database
[11:17] <groz> which we could build some report pages into the trac system from
[11:18] <groz> so if we went to an automated scripting test on things, we end up with one page in the wiki to look at, which formats the results of the last test run
[11:18] <groz> you think that's headed down the right track, or, overkill ?
[11:18] <[mbm]> I'd like to get it so an svn checkin automatically triggers snapshots and testing
[11:18] <groz> well some of the regressions will take many hours to run
[11:18] <groz> example, exercise the dhcp server is a test that has to run until leases expire
[11:19] <groz> but that would only really need to be done if dnsmaq is touched
[11:20] <[mbm]> I'd be happy with automatically attempt to build, report compile errors and do some preliminary tests each checkin, perform detailed test every day or two
[11:21] <groz> and automated testing in umls is pretty strait forward, but, it's not so strait forward to do it on specific target hardware
[11:21] <murb> how close are any of the qemu-system-$arch to target hardware?
[11:22] <groz> hehe, for my own stuff, i'll be happy with full regression once a week
[11:22] <murb> there is now a qemu mtd emulated device as well.
[11:22] <groz> i've never played with qemu murb
[11:22] <groz> its on the list of things i'd like to play with, haven't gotten to it yet
[11:22] <[mbm]> ditto
[11:23] <groz> i got uml and colinux going here from necessity
[11:23] <[mbm]> although I really doubt qemu emulates all the interfaces and platforms nor do I think it will any time soon
[11:23] <groz> no, i think the best we could hope for short term from qemu
[11:24] <groz> is to run against a pretned mips processor
[11:24] <[mbm]> really my main concern is does openwrt even build
[11:24] <groz> not sure what that would even buy us
[11:24] <groz> yah, building is good
[11:24] <groz> no need to test beyond that, if it doesn't build
[11:24] <[mbm]> since we frequently have build errors caused by typos and omissions
[11:25] Action: [mbm] always makes sure to test before commiting .. not sure about the other devs
[11:25] <Kaloz> [mbm]: i'll take a look at hardening spam filtering when i'm back from the uk
[11:26] <[mbm]> Kaloz: yeah, no rush
[11:26] Action: Kaloz does as well, even on minimal changes :p
[11:26] <groz> well, the way I normally do it, but, haven't been at this one long enough to say here with certainty
[11:26] <Kaloz> if i commit a typo, those are because i commit the wrong patch :P but most of the time i take care
[11:26] <groz> but i normally do make clean make before I commit
[11:27] <[mbm]> I clear out any files involved but rarely will I go to the clean/distclean levels
[11:27] <groz> i use it as an excuse to go make coffee or food
[11:28] <groz> and now that computers getting so damn fast
[11:28] <groz> if you dont do a full clean
[11:28] <groz> you never get a decent break
[11:28] <[mbm]> heh, yeah, that's about the only time I do a distclean is when I want to take a coffee break at the office
[11:28] Action: groz pines for the good old days where full project build took 6 hours
[11:28] <groz> the first REAL big project i did for a customer
[11:29] <Kaloz> groz: select "compile all packages"
[11:29] <groz> after 9 months of work, the build took almost 8 hours
[11:29] <Kaloz> and you are set for half a day at least
[11:29] <groz> no kaloz
[11:29] <groz> i bought a new puter a few weeks ago
[11:29] <[mbm]> heh, well if you wipe the dl directory it usually will take 6+ hours .. the default set of mirrors are slow
[11:29] <groz> LOL mbm
[11:29] <murb> groz: and now you have no time for coffee?
[11:29] <groz> things build to fast now murb, ya
[11:30] <groz> make clean is no longer a full hour break
[11:30] <groz> altho now that we have a good swack of packages into buildroot-ng
[11:30] <groz> that's probly getting there
[11:30] <[mbm]> of course buildroot-ng supports a parallel build so if needed you could do a full distclean rebuild very fast
[11:30] <murb> someone needs to package openoffice that'll fix up your lunch breaks
[11:31] <groz> umm, these are embedded headless devices
[11:31] <[mbm]> murb: we need XFree86 first ;)
[11:31] <[mbm]> (and that aone is a nice coffee break)
[11:31] <murb> [mbm]: good point!
[11:31] <groz> i dont think openoffice will fit in 16 megs of ram
[11:31] Action: murb hands groz a soldering iron.
[11:32] <groz> well, we have till april
[11:32] <groz> we do need an Xfree package for april 1
[11:32] <[mbm]> groz: well, the first step is mounting the wrt on a moving platform that slowly sweeps scanlines .. next is to pulse the dmz light in perfect sync for 8bit green ..
[11:33] <groz> nahhh
[11:33] <groz> i got a thing beside my desk here
[11:33] <groz> an led display
[11:33] <murb> [mbm]: usb vga adaptors would be cheating
[11:33] <groz> i can feed it thru a serial port
[11:33] <[mbm]> wait! let's do tempest
[11:33] <[mbm]> there's a damned radio on the device
[11:34] <[mbm]> ;)
[11:34] Action: murb wonders if kdrive supports any such devices.
[11:59] <florian__> I think we can close the qt3 and konqueror tickets, don't you think so ?
[12:00] <[mbm]> think so
[12:47] <nbd> .
[12:49] Action: [mbm] having fun in #openwrt
[12:51] <nbd> :)
[13:27] <florian__> interesting :)
[13:28] <florian__> nbd: I think you made a mistake while commiting this : https://dev.openwrt.org/changeset/4285
[13:28] <florian__> should not the nameserver address be 127.0.0.1 ?
[13:30] <nbd> no error
[13:30] <nbd> this is intentional
[13:30] <florian__> ah, ok
[13:34] Action: crazy_imp is confused -> mr_pump# ping 0.0.0.1
[13:35] <crazy_imp> connect: Invalid argument
[13:35] <[mbm]> crazy_imp: that's the point
[13:37] <nbd> hmm... maybe i need to change it after all
[13:38] <nbd> for some reason it doesn't work for me anymore
[13:38] <nbd> the first test might have accidentally succeeded
[13:40] <florian__> nbd: what did you meant by 0.0.0.1 in fact ?
[13:41] Action: florian__ is just curious ;)
[13:41] <[mbm]> just a dummy address that nobody will ever use and will never route anywhere
[13:42] <florian__> ok
[13:42] <[mbm]> but will trigger enough network activity to cause ppp to dial
[13:42] <crazy_imp> if i traceroute 1.1.1.1 the routers of my isp are playing ping-pong with it :D
[13:42] <CIA-17> nbd * r4301 /branches/whiterussian/openwrt/package/ (ppp/files/ifup.pppoe pptp/files/ifup.pptp): change to a slightly less invalid ip address
[13:42] <florian__> ok, why not ping a multicast address instead ? maybe it won't trigger ppp to dial ?
[13:44] <florian__> is it ok to use the debian patches for ipsec in whiterussian, in order to use ipsec-tools with 2.4 kernel ?
[13:44] <nbd> i guess so
[13:44] <florian__> ok
[13:54] <nbd> lol. we finally got linked by heise
[13:55] <nbd> (about some planned fon stuff)
[13:55] <nbd> http://www.heise.de/newsticker/meldung/75948 (german)
[13:57] <[mbm]> about damned time :)
[14:01] <murb> hmm one of my neighbours has a FON open AP thing.
[14:08] <florian__> how far is your master mode patch integrated in the official branch ?
[14:10] <NDL> I hadn't followed up for some time already the situation there, but so far it looks like the patch is integrated to the wireless-dev tree with some additional bugfixes by Michael.
[14:10] <florian__> ok, it's just great
[14:10] <florian__> what about the hostapd side ?
[14:13] <NDL> florian__: at the time the patch was discussed on the mailing list (several weeks ago), it was mentioned that the hostapd was also adjusted, so it can be used together with the patch. However, as things are changing quite rapidly there, I'm not 100% sure it's still the case ;-)
[14:13] <NDL> Unless there were re-designs in dscape or hostapd, I would assume they should work in AP mode OK, though.
[14:14] <florian__> great
[14:14] <NDL> (I had no time to work on it further, so that's why all my info is a bit outdated already, sorry ...)
[14:15] <florian__> you don't have to be sorry, you made a great work !
[14:15] <NDL> Thanks! :-P
[14:16] <NDL> There's still the work left - I mean PS support - but that's currently beyond my abilities (mainly - in terms of time).
[14:19] <florian__> PS ?
[14:19] <florian__> power saving ?
[14:27] <florian__> db90h: is it technically possible to have a firmware image that embedds the vx killer, or do you always need 2 steps to flash to openwrt ?
[14:28] <NDL> florian__: yes, power saving (I mean, AP support for PS-enabled STAs).
[14:33] <db90h> florian__: the short answer is two steps are required
[14:40] <florian__> NDL: ok, this seems to be documented in the specs in fact
[14:41] <florian__> db90h: ok thanks
[14:41] <db90h> note that in my recommended procedure there is a third step (done first), the vxworks_prep firmware.. that's totally optional and just to make the process safer and easier for end users
[14:42] <db90h> but, yea, I wish there were a way to do it in one step.. but with a CFE compatible firmware image it's just not.. the best you can do is make it appear to look like one step through some abstraction (like a nice little utility that does all the steps for the user)
[14:43] <NDL> florian__: there are some things that are still missing in specs related to PS support in AP mode (but they can be figured out from the hardware behavior ... probably)
[14:54] <florian__> NDL: ok
[14:55] <florian__> I would rather have client power saving capabilities because I found them really useful tool, especially on my ibook ;)
[14:55] <florian__> and also WDS would be a great thing if implemented in bcm43xx
[15:02] <NDL> florian__: from the discussions on the bcm-specs IRC, I understood that they worked on the client PS issue for quite some time, but haven't succeeded so far :-(
[15:02] <florian__> what is their channel ?
[15:02] <florian__> #bcm43xx ?
[15:02] <florian__> #bcm-dev
[15:06] <NDL> florian__: it's #bcm-dev on freenode for development (not really much activity, though ;-) and #bcm-specs on sipsolutions.net for specs discussion (more active)
[15:14] <db90h> how is the RAM allocated to /tmp managed? if more application memory is needed, is its size reduced?
[15:16] <db90h> or maybe its allocated on-demand?
[17:36] <{Nico}> nbd: ping
[17:36] <nbd> pong
[17:37] <{Nico}> i have a problem when building -ng packages with the WR sdk
[17:37] <nbd> what problem?
[17:37] <nbd> and what package?
[17:37] <{Nico}> Makefile that depends on CONFIG_PACKAGE_* to change configure options
[17:38] <nbd> hmm...
[17:38] <{Nico}> freeradius for example
[17:38] <nbd> don't know how to solve that right now
[17:38] <{Nico}> can i add something like
[17:38] <{Nico}> ifeq(SDK,1)
[17:38] <nbd> yeah
[17:39] <{Nico}> define CONFIG_PACKAGE_foo=m
[17:39] <{Nico}> endif
[17:39] <nbd> huh?
[17:39] <nbd> that looks weird
[17:40] <{Nico}> nbd: that would fake the package Makefile as if all optionnal options were selected
[17:40] <{Nico}> s/optionnal options/config options/
[17:41] <nbd> either use CONFIG_PACKAGE_foo=m or define
[17:41] <nbd> but define in this case doesn't make sense
[17:42] <{Nico}> nbd: err, soryr
[17:42] <{Nico}> ifeq(SDK,1)
[17:42] <{Nico}> CONFIG_PACKAGE_foo=m
[17:42] <{Nico}> endif
[17:42] <nbd> yep
[17:43] <nbd> well, almost
[17:43] <nbd> $(SDK)
[17:44] <{Nico}> yeah, i planned to put that somewhere in the BuildPackage template in ./include/package.mk
[17:45] <nbd> why? doesn't it already build all packages?
[17:45] <nbd> i thought this problem only affected extra menuconfig options
[17:46] <{Nico}> those options are not defined in Build/Configure or Build/Compile
[17:47] <{Nico}> s/in/during/
[17:48] <{Nico}> nbd: wait, maybe my makefile is wrong
[20:50] <florian__> re
[23:34] <florian__> I have some news for the wrt54gx2 / ROME bootloader
[23:34] <florian__> it looks like it only accepts gzip'ed kernel
[23:34] <florian__> anyway, I did not yet succeeded in making it accept a gziped kernel
[23:34] <florian__> I think it embedds a gzip decompressor
[23:35] <florian__> following the makefile creation in the wpnt834 dont't seem to make valid images :(
[00:00] --- Thu Jul 27 2006
[00:01] <nbd> [mbm]: 1.1.1.1?
[00:02] <[mbm]> hmm
[00:02] <nbd> afaik this ip is not routed anywhere
[00:03] <nbd> and people are not likely to use it for their lan :)
[00:03] <[mbm]> the 1.x is an iana reserved net
[00:04] <nbd> how about 0.0.0.1?
[00:04] <[mbm]> likewise reserved but that'll probably work fine
[00:05] <groz_> pick something in the 172 private range that nobody uses
[00:05] <groz_> that way, if it leaks, it will stop at the first properly configure upstream router
[00:05] <[mbm]> I doubt that anyone is going to route 0/8
[00:13] <CIA-17> nbd * r4285 /branches/whiterussian/openwrt/package/ (ppp/files/ifup.pppoe pptp/files/ifup.pptp): fix dial on demand support (#436)
[00:15] <[mbm]> nbd: tested?
[00:15] <nbd> yeah
[00:15] <nbd> well, not fully
[00:15] <nbd> but the relevant part
[00:16] <nbd> i set the config for dial on demand and created a /tmp/resolv.conf
[00:16] <nbd> when i did a nameserver lookup, it started its connection attempts
[00:16] <[mbm]> :)
[00:43] <[mbm]> nbd: think your module patch earlier actually broke a few things
[00:45] <nbd> which one?
[00:45] <groz_> weree there more since last nite ?
[00:45] <[mbm]> still poking around to figure out the exact cause
[00:46] <nbd> what's the symptom?
[00:46] <groz_> what is the symptom ?
[00:46] <groz_> gmta
[00:46] <[mbm]> but it seems it's trying to build kmod-ip6tables_2.6.17-aruba-1_mips.ipk even though I never selected any of the ip6tables crap
[00:46] <[mbm]> here's the relevent line -
[00:46] <[mbm]> cp: cannot stat `.../openwrt/build_mips/linux-2.6-aruba/modules/lib/modules/2.6.17/kernel/net/ipv6/netfilter/ip*.ko': No such file or directory
[00:46] <[mbm]> make[4]: *** [.../openwrt/bin/packages/kmod-ip6tables_2.6.17-aruba-1_mips.ipk] Error 1
[00:46] <nbd> something missing ... fixing that...
[00:48] <[mbm]> ah, you forgot the KCONFIG
[00:48] <CIA-17> nbd * r4286 /branches/buildroot-ng/openwrt/package/kernel/modules.mk: disable kmod-ip6tables if netfilter ipv6 support is disabled
[00:49] <nbd> yes
[00:49] <[mbm]> umm yeah, that
[00:49] Action: [mbm] was just about to commit it
[00:49] <nbd> :)
[00:52] <groz_> nbd, that new wl driver you stuck in the other day, have you done any real testing at all with it ?
[00:52] <groz_> the prior one I found 2 issues, one was the google thing, the other was a centrino wouldn't associate with it
[00:52] <groz_> i'm just building now with the new one, wondering if you've actually done anything with it yet ?
[00:53] Action: [mbm] thinks we need to make the svn server actually attempt to compile code before accepting it
[01:02] <{Nico}> it seems base-files is not installed anymore in recent -ng builds
[01:02] <nbd> groz_: haven't done much with it. just basic testing
[01:03] <groz_> ok, well i'm gonna give it a big smoke run here then shortly
[01:03] <[mbm]> {Nico}: that would explain the ticket about the broken wgt634u
[01:03] <groz_> if it goes ok on this first unit, i8'll put it on 10 more and start a regressino running
[01:04] <[mbm]> hmm image I built 30 minutes ago is missing base-files
[01:04] <{Nico}> [mbm]: yeah, and my (same) problems with uml-2.6
[01:04] Action: [mbm] is going to start revoking nbd's commit privledges if this keeps up
[01:06] <[mbm]> hmmm
[01:06] <[mbm]> target is aruba-2.6 but it's building brcm-2.4 packages
[01:06] <[mbm]> something is wrong with the stamp system
[01:08] <[mbm]> hrm.. .kernel.mk is wrong
[01:12] <[mbm]> wiped bin/package/base-files-* and .kernel.mk
[01:12] <[mbm]> built properly after
[01:14] <{Nico}> [mbm]: yeah, it seems there's now an svn rev number in the package version
[01:15] <[mbm]> yeah I did that a few revisions back just to force it to rebuild base-files
[01:15] <{Nico}> the build process does not fail if a package to be installed is not found
[01:15] <[mbm]> hmm
[01:16] <crazy_imp> {Nico}: took a look at my changes at the download.pl?
[01:19] <{Nico}> crazy_imp: yep, but i haven't tested it yet
[01:19] <crazy_imp> ack
[01:21] <[mbm]> {Nico}: did you actually find an example of it working even if a package failed to install?
[01:24] <{Nico}> [mbm]: http://pastebin.ca/100245
[01:25] <[mbm]> mumble .. stupid thing
[01:25] <[mbm]> that's the problem with base packages .. it doesn't compile
[01:26] <[mbm]> so some of the rules don't apply
[01:26] <{Nico}> look at the rev numbers
[01:26] <[mbm]> that's exactly the problem
[01:26] <[mbm]> I know what caused that
[01:28] <{Nico}> svn info are called from different directories?
[01:29] <[mbm]> no, what happens is that the ipkg control files are generated
[01:29] <[mbm]> and once generated they're pretty much ignored
[01:29] <[mbm]> so you had built base-files when you were running 4129
[01:30] <[mbm]> but when it updated to 4273 nothing forced it to rebuild the ipkg control file
[01:30] <[mbm]> so it rebuilt base-files as 4129 again
[01:30] <[mbm]> there's the secondary error that when it failed to find the package it didn't bomb
[01:32] <{Nico}> changing the dependency on .prepared to .built should be enough for 1)
[01:32] <{Nico}> err
[01:33] <[mbm]> not quite
[01:33] <[mbm]> note that the calls to timestamp.pl ignore the ipkg directory
[01:33] <[mbm]> (-x ipkg)
[01:33] <[mbm]> that's part of the problem
[01:33] <[mbm]> and I could have sworn I wrote a patch
[01:33] <[mbm]> must not have commited it
[01:34] <groz_> you are just saving it so that all the other dists dont get it out first
[01:34] <[mbm]> heh
[01:42] <groz_> mbm, i take it you are working with the aruba target mostly these days no ?
[01:43] <[mbm]> groz_: good guess
[01:43] <groz_> ok, now the dumb question
[01:43] <groz_> does aruba make any cheap low end devices ?
[01:44] <[mbm]> they tried
[01:44] <{Nico}> groz_: seems more like a wish than a question :)
[01:44] <[mbm]> they ended up with something called an ap40
[01:44] <[mbm]> which is more like cheap quality than cheap price
[01:44] <groz_> lol, so they did half of the project, made poor quality stuff
[01:44] <groz_> didn't do the other half, low price
[01:45] <groz_> is what you are saying
[01:45] <[mbm]> well, the hardware is good
[01:45] <[mbm]> it's an atheros board
[01:45] <groz_> do they still sell it ?
[01:45] <[mbm]> but they put it in a really cheap case
[01:45] <[mbm]> with an integrated antenna
[01:45] <[mbm]> and oddly no power connector
[01:45] <[mbm]> -- requires poe
[01:46] <groz_> ewwww
[01:46] <groz_> that's back assedwards
[01:46] <[mbm]> needless to say I haven't bothered with an ap40 port
[01:46] <groz_> hmm it doesn't even show in the products page of the website
[01:46] <[mbm]> think it got pulled
[01:48] <{Nico}> [mbm]: can i change ipkg to break on "file not found" error?
[01:48] <[mbm]> {Nico}: I'm still looking through the source to see why it didn't
[01:48] <[mbm]> groz_: btw, the ap41 suffers almost exactly the same fate
[01:49] <groz_> i already saw, poe only
[01:49] <groz_> useless for what i do
[01:49] <[mbm]> same cheap case too
[01:49] <groz_> our philosophy is, power plugs already exist everywhere
[01:50] <groz_> for the data side of the equation, 'what part of the word wireless is not clear'
[01:50] <{Nico}> [mbm]: basically, it's: [ -f $pkg ] then pkg_install else echo "File not found" even if pkg_install handles such error
[01:50] <[mbm]> I don't know what the case is actually made of but it feels like that old fragile bakelite stuff that radios in the 60s were made out of
[01:51] <[mbm]> {Nico}: ouch, it's in the ipkg code? fix that
[01:52] <{Nico}> [mbm]: in all branches?
[01:52] <[mbm]> yeah, probably should
[01:53] <{Nico}> ok
[01:53] <[mbm]> an 'exit 1' should do it
[01:55] <{Nico}> ipkg_install_file: ERROR: File /home/users/nthill/OpenWrt/branches/buildroot-ng/openwrt/bin/packages/base-files-brcm-2.4_8-4273_mipsel.ipk not found
[02:03] <CIA-17> nico * r4287 / (3 files in 3 dirs): let ipkg fail when a package file to be installed is not found
[02:17] <CIA-17> nico * r4288 /packages/net/freeradius/ (. Config.in Makefile ipkg/): port freeradius to buildroot-ng
[02:19] <[mbm]> hmm .. damn this is annoying
[02:20] <[mbm]> the reason why I put $(REV) in the base-files is so that the /etc/banner would always be correct
[02:20] <[mbm]> without having to force it to rebuild base-files each time
[02:21] <[mbm]> but it's too stupid to know to rebuild the ipkg control file
[02:21] <groz_> put $(REV) into a tempfile, make base dependant on that
[02:22] <[mbm]> from it's point of view, none of the timestamps have changed so there's nothing to be done
[02:22] <groz_> rewrite it if the contents are wrong
[02:22] <groz_> hmm, it'll still be circular and wont resolve right
[02:24] <[mbm]> yeah, the bigest error is that it knows what version it's trying to build but it has no idea what version is in the ipkg control file
[02:25] <{Nico}> groz_: uml-2.6 works
[02:25] <groz_> i expect it should nico
[02:25] <groz_> i'm going to add a few more lines with 'network quick start' to the readme
[02:25] <groz_> eventually
[02:26] <groz_> for now, think it's safe to assume, folks building and trying uml
[02:26] <groz_> either know what they are doing, or, know how to read up at it's home site
[02:26] <[mbm]> not an assumption I'd make
[02:27] <groz_> well, then they wont brick any routers when they screw it up
[02:27] <groz_> they can kill -9 the uml
[02:27] <groz_> and try again
[02:28] <[mbm]> true
[02:28] <groz_> and actually thinking about it
[02:28] <groz_> may not be a bad idea for total newbs
[02:28] <groz_> reccomend they learn the concepts with a uml first
[02:28] <groz_> before they risk hardware
[02:29] <groz_> particularily true when i get the colinux build running correctly
[02:30] <{Nico}> groz_: this is a really nice addition, especially for package maintainers
[02:30] <groz_> i see maintaining and testing
[02:30] <groz_> it'll get a lot of use
[02:30] <groz_> and i also think, it'll get more folks looking at working on packages
[02:31] <groz_> if they can do the work/experiments without using thier one and only router
[02:31] <groz_> which they need for internet connection
[02:31] <[mbm]> hmm found one more obscure case in the above error -- make doesn't stop if it fails to build the ipkg
[02:32] <CIA-17> nico * r4289 /packages/libs/freetype/ (. Config.in Makefile ipkg/): port freetype to buildroot-ng
[02:34] <groz_> Nico, are you using the uml for anything yet ?
[02:42] <CIA-17> nico * r4290 /packages/libs/libao/ (. Config.in Makefile ipkg/): port libao to buildroot-ng
[02:43] <{Nico}> groz_: nope, focused on -ng porting but i plan to test -ng packages on it
[02:55] <CIA-17> nico * r4291 /packages/libs/libaudiofile/ (. Config.in Makefile ipkg/): port libaudiofile to buildroot-ng
[03:01] <CIA-17> mbm * r4292 /branches/buildroot-ng/openwrt/include/package.mk: add more ipkg validation checks
[03:01] <[mbm]> that should hopefully put an end to all those ipkg errors
[03:02] <groz_> question for you folks
[03:02] <groz_> i have a patch here for bridge code, that enables the ability to change the ethernet hardware address of the bridge
[03:02] <groz_> any objections to adding that to kernel patches ?
[03:03] <groz_> instead of bridge taking arbitrarily the last adapter inserted
[03:03] <groz_> you can nail it down to one of the adapters in there
[03:03] <groz_> and force it to keep using that one
[03:03] <[mbm]> thought the bridge took the first mac, but anywaysm I see your point
[03:03] <groz_> actually ya, it does
[03:03] <groz_> but if you remove the first
[03:03] <groz_> then it's got the second
[03:03] <groz_> with this patch in
[03:03] <groz_> you can pick one, nail it to that
[03:04] <groz_> then if you are removing and inserting adapters regularily
[03:04] <[mbm]> one of the things that has bothered me is the fact that the macs under openwrt are usually completely different than the ones stamped on the device
[03:04] <groz_> it's not always chainging
[03:04] <[mbm]> which I think I mostly fixed in rc5
[03:05] <groz_> yah, but using this example
[03:05] <groz_> iv'e got a box here on the desk, it's wired to a couple of switches
[03:05] <groz_> one on the 4 port, one on the uplink
[03:05] <groz_> they are running different ip segments
[03:05] <groz_> but upstream they do join again
[03:05] <[mbm]> no, I'm not diagreeing by any means, I think we should add the patch
[03:05] <groz_> so if the 4 port and the uplink are using same mac
[03:05] <groz_> it's a pain
[03:06] <groz_> this adds the option
[03:06] <groz_> doesn't force it
[03:06] <[mbm]> just pointing out that macs have always been an issue
[03:06] <groz_> just make 'ifconfig hw ether xxxxxx' work
[03:06] <groz_> yup, it's always been a pain/hassle with these boxes
[03:06] <groz_> and it gets much worse when you have a spider of wds connections between a half dozen of them
[03:07] <[mbm]> yay, the ipkg patches work (had to commit them to bump the svn revision so I could actually test them properly)
[03:07] <groz_> ok, i'll add that patch in on my next commit
[03:08] <groz_> it's actually a nice clean one, last time i checked, it applies equally well to 2.4 and 2.6
[03:08] <groz_> just adds one more ioctl to the bridge code
[03:08] <[mbm]> btw a side effect of the ipkg patch above is that after you svn update to that version it will be forced to rebuild packages that you've already built
[03:09] <[mbm]> (just a warning for those people that do all their work in the build_mips directory)
[03:09] <groz_> that's not necessarily a _bad_ side effect
[03:09] <[mbm]> right
[03:09] <groz_> i've got all my work soft linked in
[03:09] <groz_> and linked into 4 different build trees
[03:09] <groz_> soon to be 6
[03:09] <CIA-17> nico * r4293 /packages/libs/libart/ (. Config.in Makefile ipkg/): port libart to buildroot-ng
[03:09] <[mbm]> heh, same here
[03:10] <groz_> i haven't played much with 2.6 on the wrt yet
[03:10] <groz_> but it's coming real soon now
[03:10] <[mbm]> 2.6 on the wrt isn't much to look at
[03:10] <groz_> and i'm gonna build a colinux target for my own use
[03:10] <[mbm]> diag doesn't work and wireless doesn't work
[03:10] <groz_> dunno if you guys want it committed in or not
[03:10] <[mbm]> (plus there's no nvram which means that after flashing it'll get a /etc/config which will force it to 192.168.1.1)
[03:11] <[mbm]> but hey, it has a spiffy 2.6.x kernel, :P
[03:11] <groz_> well, i'm planning on taking a good look at the open driver
[03:11] <groz_> and seeing if i cant make a few things work with it
[03:11] <[mbm]> yeah, I plan on fixing the upgrade problems caused by the lack of persistant config storage
[03:11] <groz_> i'm getting very tired of fighting one bad bug in the broadcom drivers
[03:12] <groz_> the wds connections when bridged
[03:12] <[mbm]> heh
[03:12] <groz_> often send packets thru with the wrong ethernet macs in the headers
[03:12] <groz_> and it's persistent across EVERY revision of the drivers
[03:12] <[mbm]> 'd expect that from client mode but not wds
[03:12] <groz_> i've got some test case code here that exercises it, and counts errors
[03:12] <[mbm]> figures
[03:13] <groz_> by sending raw ethernet frames with payload sying who really sent it
[03:13] <groz_> and then comparing that to the output of the driver
[03:13] <groz_> so far, the best driver
[03:13] <groz_> is the one in the gl linksys release
[03:13] <groz_> lowest error counts overall
[03:13] <groz_> and by the time the day is over
[03:13] <groz_> i'll be able to tell how this latest wl does
[03:13] <[mbm]> nbd just changed the wl version about a day ago
[03:13] <groz_> yah, that's why i'm building up some wrts here
[03:13] <groz_> to run this test on them
[03:13] <[mbm]> ah
[03:14] <groz_> just cleaning up my scripts to work in this new setup
[03:14] <groz_> then i'll flash 10 of them
[03:14] <groz_> go for dinner
[03:14] <groz_> come back and see the results
[03:14] Action: groz_ not got high hopes this bug cured
[03:14] <CIA-17> nico * r4294 /packages/libs/libcli/ (. Config.in Makefile ipkg/): port libcli to buildroot-ng
[03:15] <[mbm]> hmm yeah, I don't think it is either
[03:15] <groz_> the long and the short of this one
[03:15] <[mbm]> companies really aren't in any rush to fix bugs
[03:15] <groz_> if you have 15 units, with an average of 3 wds each, all interconnected to bridge one big mesh
[03:15] <groz_> after 15 units, expect about 5% of packets
[03:15] <groz_> to route incorrectly thru the bridge
[03:15] <groz_> it goes up exponentially from there
[03:15] <groz_> as you add units to the mesh
[03:16] Action: [mbm] has a tv with an upgradable firmware and a few nasty firmware bugs .. you'd think that 3 revisions later they'd have fixed them (yes, I did my darnest to report them) .. nope
[03:16] <groz_> well, my _hope_ is the bug is in wl.o, and not the actual device firmware
[03:17] <groz_> if that's the case, i'll be HUGELY inspired to get working in the internals of the open source driver
[03:17] <groz_> and get this FIXED
[03:17] <[mbm]> every time I try to report the bugs I get the same crap about how I somehow caused the error by not following instructions correctly (which inclues giving me a new copy of the manuals) or that the bug will be fixed in the next rev
[03:18] <[mbm]> which leads me to believe that it'll never get fixed
[03:18] <groz_> mbm, there is a reason a project like this gains steam after a while
[03:18] <[mbm]> :)
[03:18] <groz_> if your tv was linux and open
[03:18] <groz_> those bugs would be long fixed
[03:18] <groz_> example, local computer stores here in vancouver
[03:18] <groz_> they dont carry linksy v5 anymore
[03:18] <groz_> all carry gl
[03:18] <CIA-17> nico * r4295 /packages/libs/rrdtool-1.0.x/ (. Config.in Makefile ipkg/): port rrdtool-1.0.x to buildroot-ng
[03:18] <groz_> and, have cd's with various firmwares for the taking when you buy one
[03:19] <groz_> the sales staff will tell you, it's the router of choice for home use
[03:19] <[mbm]> hell, if they didn't encrypt the tv firmware I'd have probably taken a crack at it
[03:19] <[mbm]> but no, those bastards put high grade encryption on the damned thing
[03:19] <groz_> which tv is it ?
[03:19] <[mbm]> philips
[03:19] <groz_> hmm, i have a contact in phillips
[03:20] <groz_> very well placed too, but he retired last year
[03:20] <groz_> an uncle
[03:20] <[mbm]> it actually is encrypted -- service manual even tells you that
[03:20] <[mbm]> and it's not just some trivial xor crap either
[03:21] <[mbm]> :/
[03:22] <groz_> had a very interesting talk about encryption/drm etc when i was out flying the other day
[03:22] <groz_> the person chartering the plane, a gal with quite a selection of albums out
[03:22] <groz_> a couple of them are platinum
[03:26] <CIA-17> nico * r4296 /packages/ (17 files in 17 dirs): change SECTION and CATEGORY to put packages in the right place, change PKG_BUILDDEPs to use package-name instead of source-name (seems to work best with WR sdk).
[05:03] <CIA-17> nico * r4297 /packages/net/ez-ipupdate/ (. Makefile): port ez-ipupdate to buildroot-ng
[05:05] <CIA-17> nico * r4298 /packages/net/ez-ipupdate/ (Config.in ipkg/): oops, forgot to remove unneeded files for -ng from my last commit
[05:19] <CIA-17> nico * r4299 /packages/libs/ (8 files in 8 dirs): remove ref to host include and lib dirs from pkgconfig files
[05:22] <CIA-17> nico * r4300 /packages/net/nsca/ (. Makefile): add send-nsca to buildroot-ng
[10:08] <florian__> ola
[10:10] <groz_> good morning
[10:15] <groz> not complaining, it's getting late, i had a very rude early start this morning
[10:15] <groz> damn phone
[10:16] <groz> so just fiddling with a little 'fun stuff' before i call it quits
[10:16] <florian__> lol
[10:17] <groz> well, i guess, that is if you call setting up network tunnels for a test network 'fun'
[10:17] <groz> I'm scripting the bring up of a few umls with openwrt, then set up some tunnels between them over pretend networks
[10:18] <groz> building it up so that it can all be driven by a script set, bring them up from fresh build, install the packages, install configs, then start bringing up and testing tunnels
[10:20] <florian__> has anyone ever backported a netfilter module from 2.6 to 2.4 ?
[10:20] <groz> i have no clue
[10:20] <groz> what do you want to backport ?
[10:23] <Kaloz> http://ars.userfriendly.org/cartoons/?id=20060726
[10:23] <Kaloz> :D
[10:24] <florian__> SIP connection tracking
[10:25] <groz> I thought i've seen that in 2.4 at some point
[10:27] <florian__> I merged it into whiterussian, but its licence was unclear
[10:27] <Kaloz> how could it be unclear?
[10:27] <Kaloz> it's gpl
[10:27] <florian__> I think you know better than me because you told me I had to remove it, don't you remember ?
[10:28] <florian__> the 2.6 sip connection tracking is GPL, the 2.4 one was taken apart
[10:28] <Kaloz> remove in that way :p add gpl header
[10:28] <Kaloz> iirc this happened with some of those in the netfilter svn as well
[10:28] Action: [mbm] mumbles .. is today "bad advice day" on the forum?
[10:28] <Kaloz> all of those are derivative works
[10:29] <florian__> I don't understand a word of what are you writing :p
[10:29] <Kaloz> florian__: gpl, search for derivative work
[10:30] <florian__> Kaloz: ok so you mean that some netfilter modules were integrated as GPL simply because the source was available or ?
[10:30] <Kaloz> The GPL implies that a derivative work is one that is linked, statically or dynamically, with the original work.
[10:31] <florian__> right, so ?
[10:31] <Kaloz> this is the reason why legally the broadcom wl.o is violating the gpl, as it doesn't have the "gpl'ed glue" like the ati, nvidia drivers and madwifi does
[10:31] <Kaloz> but that's the grey zone
[10:31] <Kaloz> with netfilet/iptables
[10:32] <Kaloz> additional modules are stricktly go into gpl
[10:32] <Kaloz> no matter what crap does a company write into the headers
[10:32] <murb> Kaloz: if you take a fundamtalist approach then the madwifi stuff is equally as illegal as the wl.o
[10:33] <florian__> Kaloz: so what's the point with the SIP connection tracking module I merged few months ago ? you mean that adding the GPL header back would be sufficient for it to be integrated back again ?
[10:34] <murb> florian__: which sip conntracking module that there have been atleast 2
[10:34] <Kaloz> florian__: remove the propietary header if exist, and add a gpl header that states where did it came from and why it's gpl
[10:34] <murb> ?
[10:34] <murb> florian__: you'd probably be on pretty safe ground if you took the patch posted to netfilter-devel
[10:34] <florian__> murb: was it a 2.4 or 2.6 module ?
[10:35] <murb> florian__: 2.6
[10:35] <murb> there was an early implementation from some italian guys iirc.
[10:35] <murb> for 2.4
[10:35] <murb> s/early/earlier/
[10:35] <florian__> ok
[10:36] <murb> the 2.6 on has been merged into 2.6.18 iirc.
[10:36] <florian__> that's why I say I would rather backport the 2.6 module, rather than using a 2.4 one that was posted before
[10:36] <florian__> the 2.6 is mature and stable
[10:36] <murb> florian__: problem is netfilter has changed a *lot* since then.
[10:36] <florian__> well, playing with existing connection tracking module seems to be understandable
[10:37] <florian__> and there are few macros, and types that changed regarding module development
[10:39] <florian__> {Nico}: did you test the au1000 port ?
[10:40] <florian__> well I don't know whether it's worth actually
[10:48] <[mbm]> florian__: if code was released as GPL and later modified, it's still GPL code and any headers added that claim otherwise can be removed
[10:49] <florian__> ok, maybe I should check then
[10:50] <florian__> https://dev.openwrt.org/browser/trunk/openwrt/target/linux/generic-2.4/patches/613-netfilter_nat_sip.patch?rev=3291
[10:50] <groz> actually mbm, the _original author_ can take something, and for example, go gpl->bsd license with a later revision
[10:50] <groz> and if you wanna stay technically pure, you go back to the last gpl release
[10:50] <groz> and fork from there
[10:51] <[mbm]> groz: yeah, the original other maintained copyright and can do whatever
[10:51] <groz> but, if joe smith takes the code
[10:51] <groz> they cannot do the same
[10:51] <[mbm]> right, in this case it sounds like broadcom has just blindly added useless headers
[10:52] <groz> oh, didn't you know, that's an automated process
[10:52] <groz> take code, strip comments, add new copyrights, feed it thru a beautifier to change th e'style'
[10:52] <groz> and call it 'my original work'
[10:52] <[mbm]> I didn't expect them to open every file by hand and type out the new header
[10:54] <[mbm]> Kaloz: we need better spam filters on openwrt-devel@
[10:56] <Kaloz> [mbm]: greylisting + razor + rbl + spamassasin is there...
[10:56] <[mbm]> yeah, but I'm still getting blatent spam :/
[10:56] <groz> kaloz, are you feeding the filter with a big batch of spam every day to train it ?
[10:57] <Kaloz> groz: no, but that's pretty useless if you have razor
[10:58] <groz> i set up a couple of common names, jane and frank
[10:58] <Kaloz> but we can give it a try later :)
[10:58] <groz> anything arriving for them, goes strait into the filters as 'training mateiral'
[10:58] <groz> it works pretty slick
[10:58] <groz> they get LOTS of spam
[11:08] <groz> mbm, whats your thoughts on this, building images for regression testing things
[11:08] <groz> should it start with a 'stock bare' image, and scripts intall packages
[11:08] <groz> or should we build images with a list pre-installed for the regression ?
[11:09] <groz> i'm hedging toward working with a stock bare starter image
[11:09] <groz> and then script the installation of packages as needed for testing
[11:09] <[mbm]> bare, then you catch dependancy issues
[11:16] <groz> I need a proper regression system for a couple of projects, so looking at some ideas here, and looking to roll this all into one setup
[11:16] <groz> thinking of re-working some of the existing scripts, so they actually populate database tables with results
[11:17] <groz> for a lot of things testing various packages and so on
[11:17] <groz> that coud be scripted all on one machine, thru umls, and then post results into a database
[11:17] <groz> which we could build some report pages into the trac system from
[11:18] <groz> so if we went to an automated scripting test on things, we end up with one page in the wiki to look at, which formats the results of the last test run
[11:18] <groz> you think that's headed down the right track, or, overkill ?
[11:18] <[mbm]> I'd like to get it so an svn checkin automatically triggers snapshots and testing
[11:18] <groz> well some of the regressions will take many hours to run
[11:18] <groz> example, exercise the dhcp server is a test that has to run until leases expire
[11:19] <groz> but that would only really need to be done if dnsmaq is touched
[11:20] <[mbm]> I'd be happy with automatically attempt to build, report compile errors and do some preliminary tests each checkin, perform detailed test every day or two
[11:21] <groz> and automated testing in umls is pretty strait forward, but, it's not so strait forward to do it on specific target hardware
[11:21] <murb> how close are any of the qemu-system-$arch to target hardware?
[11:22] <groz> hehe, for my own stuff, i'll be happy with full regression once a week
[11:22] <murb> there is now a qemu mtd emulated device as well.
[11:22] <groz> i've never played with qemu murb
[11:22] <groz> its on the list of things i'd like to play with, haven't gotten to it yet
[11:22] <[mbm]> ditto
[11:23] <groz> i got uml and colinux going here from necessity
[11:23] <[mbm]> although I really doubt qemu emulates all the interfaces and platforms nor do I think it will any time soon
[11:23] <groz> no, i think the best we could hope for short term from qemu
[11:24] <groz> is to run against a pretned mips processor
[11:24] <[mbm]> really my main concern is does openwrt even build
[11:24] <groz> not sure what that would even buy us
[11:24] <groz> yah, building is good
[11:24] <groz> no need to test beyond that, if it doesn't build
[11:24] <[mbm]> since we frequently have build errors caused by typos and omissions
[11:25] Action: [mbm] always makes sure to test before commiting .. not sure about the other devs
[11:25] <Kaloz> [mbm]: i'll take a look at hardening spam filtering when i'm back from the uk
[11:26] <[mbm]> Kaloz: yeah, no rush
[11:26] Action: Kaloz does as well, even on minimal changes :p
[11:26] <groz> well, the way I normally do it, but, haven't been at this one long enough to say here with certainty
[11:26] <Kaloz> if i commit a typo, those are because i commit the wrong patch :P but most of the time i take care
[11:26] <groz> but i normally do make clean make before I commit
[11:27] <[mbm]> I clear out any files involved but rarely will I go to the clean/distclean levels
[11:27] <groz> i use it as an excuse to go make coffee or food
[11:28] <groz> and now that computers getting so damn fast
[11:28] <groz> if you dont do a full clean
[11:28] <groz> you never get a decent break
[11:28] <[mbm]> heh, yeah, that's about the only time I do a distclean is when I want to take a coffee break at the office
[11:28] Action: groz pines for the good old days where full project build took 6 hours
[11:28] <groz> the first REAL big project i did for a customer
[11:29] <Kaloz> groz: select "compile all packages"
[11:29] <groz> after 9 months of work, the build took almost 8 hours
[11:29] <Kaloz> and you are set for half a day at least
[11:29] <groz> no kaloz
[11:29] <groz> i bought a new puter a few weeks ago
[11:29] <[mbm]> heh, well if you wipe the dl directory it usually will take 6+ hours .. the default set of mirrors are slow
[11:29] <groz> LOL mbm
[11:29] <murb> groz: and now you have no time for coffee?
[11:29] <groz> things build to fast now murb, ya
[11:30] <groz> make clean is no longer a full hour break
[11:30] <groz> altho now that we have a good swack of packages into buildroot-ng
[11:30] <groz> that's probly getting there
[11:30] <[mbm]> of course buildroot-ng supports a parallel build so if needed you could do a full distclean rebuild very fast
[11:30] <murb> someone needs to package openoffice that'll fix up your lunch breaks
[11:31] <groz> umm, these are embedded headless devices
[11:31] <[mbm]> murb: we need XFree86 first ;)
[11:31] <[mbm]> (and that aone is a nice coffee break)
[11:31] <murb> [mbm]: good point!
[11:31] <groz> i dont think openoffice will fit in 16 megs of ram
[11:31] Action: murb hands groz a soldering iron.
[11:32] <groz> well, we have till april
[11:32] <groz> we do need an Xfree package for april 1
[11:32] <[mbm]> groz: well, the first step is mounting the wrt on a moving platform that slowly sweeps scanlines .. next is to pulse the dmz light in perfect sync for 8bit green ..
[11:33] <groz> nahhh
[11:33] <groz> i got a thing beside my desk here
[11:33] <groz> an led display
[11:33] <murb> [mbm]: usb vga adaptors would be cheating
[11:33] <groz> i can feed it thru a serial port
[11:33] <[mbm]> wait! let's do tempest
[11:33] <[mbm]> there's a damned radio on the device
[11:34] <[mbm]> ;)
[11:34] Action: murb wonders if kdrive supports any such devices.
[11:59] <florian__> I think we can close the qt3 and konqueror tickets, don't you think so ?
[12:00] <[mbm]> think so
[12:47] <nbd> .
[12:49] Action: [mbm] having fun in #openwrt
[12:51] <nbd> :)
[13:27] <florian__> interesting :)
[13:28] <florian__> nbd: I think you made a mistake while commiting this : https://dev.openwrt.org/changeset/4285
[13:28] <florian__> should not the nameserver address be 127.0.0.1 ?
[13:30] <nbd> no error
[13:30] <nbd> this is intentional
[13:30] <florian__> ah, ok
[13:34] Action: crazy_imp is confused -> mr_pump# ping 0.0.0.1
[13:35] <crazy_imp> connect: Invalid argument
[13:35] <[mbm]> crazy_imp: that's the point
[13:37] <nbd> hmm... maybe i need to change it after all
[13:38] <nbd> for some reason it doesn't work for me anymore
[13:38] <nbd> the first test might have accidentally succeeded
[13:40] <florian__> nbd: what did you meant by 0.0.0.1 in fact ?
[13:41] Action: florian__ is just curious ;)
[13:41] <[mbm]> just a dummy address that nobody will ever use and will never route anywhere
[13:42] <florian__> ok
[13:42] <[mbm]> but will trigger enough network activity to cause ppp to dial
[13:42] <crazy_imp> if i traceroute 1.1.1.1 the routers of my isp are playing ping-pong with it :D
[13:42] <CIA-17> nbd * r4301 /branches/whiterussian/openwrt/package/ (ppp/files/ifup.pppoe pptp/files/ifup.pptp): change to a slightly less invalid ip address
[13:42] <florian__> ok, why not ping a multicast address instead ? maybe it won't trigger ppp to dial ?
[13:44] <florian__> is it ok to use the debian patches for ipsec in whiterussian, in order to use ipsec-tools with 2.4 kernel ?
[13:44] <nbd> i guess so
[13:44] <florian__> ok
[13:54] <nbd> lol. we finally got linked by heise
[13:55] <nbd> (about some planned fon stuff)
[13:55] <nbd> http://www.heise.de/newsticker/meldung/75948 (german)
[13:57] <[mbm]> about damned time :)
[14:01] <murb> hmm one of my neighbours has a FON open AP thing.
[14:08] <florian__> how far is your master mode patch integrated in the official branch ?
[14:10] <NDL> I hadn't followed up for some time already the situation there, but so far it looks like the patch is integrated to the wireless-dev tree with some additional bugfixes by Michael.
[14:10] <florian__> ok, it's just great
[14:10] <florian__> what about the hostapd side ?
[14:13] <NDL> florian__: at the time the patch was discussed on the mailing list (several weeks ago), it was mentioned that the hostapd was also adjusted, so it can be used together with the patch. However, as things are changing quite rapidly there, I'm not 100% sure it's still the case ;-)
[14:13] <NDL> Unless there were re-designs in dscape or hostapd, I would assume they should work in AP mode OK, though.
[14:14] <florian__> great
[14:14] <NDL> (I had no time to work on it further, so that's why all my info is a bit outdated already, sorry ...)
[14:15] <florian__> you don't have to be sorry, you made a great work !
[14:15] <NDL> Thanks! :-P
[14:16] <NDL> There's still the work left - I mean PS support - but that's currently beyond my abilities (mainly - in terms of time).
[14:19] <florian__> PS ?
[14:19] <florian__> power saving ?
[14:27] <florian__> db90h: is it technically possible to have a firmware image that embedds the vx killer, or do you always need 2 steps to flash to openwrt ?
[14:28] <NDL> florian__: yes, power saving (I mean, AP support for PS-enabled STAs).
[14:33] <db90h> florian__: the short answer is two steps are required
[14:40] <florian__> NDL: ok, this seems to be documented in the specs in fact
[14:41] <florian__> db90h: ok thanks
[14:41] <db90h> note that in my recommended procedure there is a third step (done first), the vxworks_prep firmware.. that's totally optional and just to make the process safer and easier for end users
[14:42] <db90h> but, yea, I wish there were a way to do it in one step.. but with a CFE compatible firmware image it's just not.. the best you can do is make it appear to look like one step through some abstraction (like a nice little utility that does all the steps for the user)
[14:43] <NDL> florian__: there are some things that are still missing in specs related to PS support in AP mode (but they can be figured out from the hardware behavior ... probably)
[14:54] <florian__> NDL: ok
[14:55] <florian__> I would rather have client power saving capabilities because I found them really useful tool, especially on my ibook ;)
[14:55] <florian__> and also WDS would be a great thing if implemented in bcm43xx
[15:02] <NDL> florian__: from the discussions on the bcm-specs IRC, I understood that they worked on the client PS issue for quite some time, but haven't succeeded so far :-(
[15:02] <florian__> what is their channel ?
[15:02] <florian__> #bcm43xx ?
[15:02] <florian__> #bcm-dev
[15:06] <NDL> florian__: it's #bcm-dev on freenode for development (not really much activity, though ;-) and #bcm-specs on sipsolutions.net for specs discussion (more active)
[15:14] <db90h> how is the RAM allocated to /tmp managed? if more application memory is needed, is its size reduced?
[15:16] <db90h> or maybe its allocated on-demand?
[17:36] <{Nico}> nbd: ping
[17:36] <nbd> pong
[17:37] <{Nico}> i have a problem when building -ng packages with the WR sdk
[17:37] <nbd> what problem?
[17:37] <nbd> and what package?
[17:37] <{Nico}> Makefile that depends on CONFIG_PACKAGE_* to change configure options
[17:38] <nbd> hmm...
[17:38] <{Nico}> freeradius for example
[17:38] <nbd> don't know how to solve that right now
[17:38] <{Nico}> can i add something like
[17:38] <{Nico}> ifeq(SDK,1)
[17:38] <nbd> yeah
[17:39] <{Nico}> define CONFIG_PACKAGE_foo=m
[17:39] <{Nico}> endif
[17:39] <nbd> huh?
[17:39] <nbd> that looks weird
[17:40] <{Nico}> nbd: that would fake the package Makefile as if all optionnal options were selected
[17:40] <{Nico}> s/optionnal options/config options/
[17:41] <nbd> either use CONFIG_PACKAGE_foo=m or define
[17:41] <nbd> but define in this case doesn't make sense
[17:42] <{Nico}> nbd: err, soryr
[17:42] <{Nico}> ifeq(SDK,1)
[17:42] <{Nico}> CONFIG_PACKAGE_foo=m
[17:42] <{Nico}> endif
[17:42] <nbd> yep
[17:43] <nbd> well, almost
[17:43] <nbd> $(SDK)
[17:44] <{Nico}> yeah, i planned to put that somewhere in the BuildPackage template in ./include/package.mk
[17:45] <nbd> why? doesn't it already build all packages?
[17:45] <nbd> i thought this problem only affected extra menuconfig options
[17:46] <{Nico}> those options are not defined in Build/Configure or Build/Compile
[17:47] <{Nico}> s/in/during/
[17:48] <{Nico}> nbd: wait, maybe my makefile is wrong
[20:50] <florian__> re
[23:34] <florian__> I have some news for the wrt54gx2 / ROME bootloader
[23:34] <florian__> it looks like it only accepts gzip'ed kernel
[23:34] <florian__> anyway, I did not yet succeeded in making it accept a gziped kernel
[23:34] <florian__> I think it embedds a gzip decompressor
[23:35] <florian__> following the makefile creation in the wpnt834 dont't seem to make valid images :(
[00:00] --- Thu Jul 27 2006