**** BEGIN LOGGING AT Tue Jun 12 03:00:04 2018 Jun 12 03:50:08 didn't know OVH is using a wrt version from Oct 2017 https://github.com/ovh/overthebox-lede Jun 12 05:24:28 Hi, I have version v.w.x.y in the 18.06 branch. In the master I have v.w.x.z. What is the 'way' to get z into the 18.06 branch? It contains bugfixes. Jun 12 05:36:58 wvdakker: i think you want cherry-pick Jun 12 05:49:06 m4t: And how do I pick a cherry? Jun 12 05:49:20 with git Jun 12 05:49:29 wvdakker: git cherry-pick --help Jun 12 05:50:11 Thanks Jun 12 05:50:21 strange. I'm trying to build an ext4 image + rootfs for my target, even tough I've that configured, I only get the rootfs Jun 12 05:50:49 user error most likely, but this worked two days ago Jun 12 05:51:31 azarus: maybe make -j1 V=s and see what it's doing during the image generation Jun 12 05:51:52 m4t: it does call ext4 generation somewhere Jun 12 05:51:58 but I will paste it, a sec Jun 12 05:52:12 i've never used ext4 images Jun 12 05:54:19 I kinda want them, my target supports booting from sd cards Jun 12 05:55:23 does it resize2fs on first boot? Jun 12 05:55:47 m4t: i just set the image fs size to the size of my sdcard Jun 12 05:55:51 oic Jun 12 05:56:05 or I could just resize2fs before I even boot the first time Jun 12 05:56:32 m4t: http://ix.io/1d4D Jun 12 05:56:46 it does call make_ext4fs Jun 12 05:58:58 yes i see that Jun 12 05:59:05 and yet, my bin// is: http://ix.io/1d4F Jun 12 05:59:10 only get the rootfs Jun 12 05:59:48 does /home/azarus/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl_eabi/linux-mvebu_cortexa9/root.ext4 exist? Jun 12 05:59:53 m4t: yes Jun 12 05:59:54 idk if it's supposed to copy it to bin Jun 12 06:00:06 oh so what is the issue? it's not being mv/cp'd? Jun 12 06:00:07 it should, I should end up with a .img Jun 12 06:00:16 with partitions and everything Jun 12 06:00:21 gotcha Jun 12 06:00:31 currently, I only get the root filesystem Jun 12 06:05:47 Ah, I know what's wrong! Jun 12 06:05:55 * azarus facepalms Jun 12 06:06:13 I tried to build a target that wasn't even in the source, wrong git checkout... Jun 12 06:07:29 dsorry for wasting your time :/ Jun 12 06:07:34 s/d/// Jun 12 06:07:52 i just tried to build an ath79 ext4 Jun 12 06:08:12 it almost looks like fwtool is eating the iamge Jun 12 06:10:35 nvm Jun 12 06:10:57 it's Build/check-size in include/image-commands.mk that rm's it Jun 12 06:17:04 yup, works again :) Jun 12 06:18:49 cool Jun 12 06:19:25 i started running f2fs on my sdcard last release...seems to work well Jun 12 06:19:35 extroot Jun 12 06:19:46 er overlay, whatever it's called nowadays Jun 12 06:19:55 ah, don't need extroot on my target ;) Jun 12 06:19:58 targets* Jun 12 06:20:10 my router has so much flash I don't know what to do with it Jun 12 06:20:26 and my helios4 target boots off SD, so I have no need for it Jun 12 08:12:42 mkresin: ping - Tommy calling Fritz https://github.com/ldir-EDB0/openwrt/commit/479d99e12d5fe1c24fbcbe80aa70ad21a0f04e6c Jun 12 08:13:46 mkresin: the really curious thing.... no debug printed... so code correct or incorrect isn't even being called. Jun 12 14:35:17 Hi all Jun 12 14:40:12 Bit confused about the different parts of package installation. So I know you'll run SUBMAKE install by default in the package build directory. Is the destination of this the staging directory? You then have the Package/x/install part which is included in the packages OpenWRT Makefile. I believe this is installed to ipkg-platform inside the packages build directory. This is used to build the ipkg file. But where Jun 12 14:40:13 is the final root directory that becomes rootfs? And how is that directory populated? Is it populated by running ipkg on the build machine with all the packages? If what I said is correct so far my overall question is what is the point of running SUBMAKE install when you're just going to use ipkg to install it to rootfs directory? Jun 12 14:40:16 thanks in advanced Jun 12 14:43:54 because you also need the packages for packages that are =m for instance. Jun 12 14:44:05 and so that the rules for buliding the package are consistent Jun 12 14:50:09 thanks for help karlp. I guess my question is that there are two directories build ipkg-install and ipkg-. The former is generated by running SUBMAKE install DESTDIR=package/ipkg-install and the latter is built by using the INSTALL_DATA, INSTALL_BIN, INSTALL_DIR macros ( part of the process of building the package ). My two questionsa are: Jun 12 14:50:33 1) What is the point in the fomer (the make install generated version) if you're just going to use ipkg to make rootfs and the packages themselves Jun 12 14:51:10 ipkg-install is a scratch directory for upstream "make insteall" Jun 12 14:51:12 2) Why use the DATA,BIN,DIR macros if make install will install the the files anyway? Shouldn't you just use ipkg-install as the directory for the package? Jun 12 14:51:15 one of them is the DESTDIR for upstream makeinstall Jun 12 14:51:34 ipkg- is where the relevant bits get staged for .ipk packaging Jun 12 14:51:40 Neighbor1111111: you don't walways want to just use upstream's "make install" target Jun 12 14:51:57 usually you *never* want it Jun 12 14:52:09 it ships stuff like man pages, shared files, examples Jun 12 14:52:16 sorry guys what do you mean by 'upstream' make install? Jun 12 14:52:17 well, you often want it to be able to pick things from, but yeah, it's never usually good enough by itself :) Jun 12 14:52:25 the "make insatll" from the package source itself Jun 12 14:52:45 Ahh I see, yeah that is my point. You don't want that stuff so what is the point in it? Is it every used? Jun 12 14:52:55 because lots of upstream packages do things that matter in that stage Jun 12 14:53:01 Why not just skip the upstream make install? Jun 12 14:53:21 ahh ok I see - can you give an example (sorry not trying to be annoying) Jun 12 14:53:22 and/or make it vastly simpler to pick out DESTDIR/usr/blah rather than trying to find things inside the source tree's autohell Jun 12 14:53:28 Neighbor1111111: because it sometimes does preparation steps you cannot/do not want to do manually Jun 12 14:53:40 soversion symlinks, arranging directory layouts Jun 12 14:54:25 you can then copy entire subdirectory trees from the make isntall scratch dir instead of assembling build artifacts from all over the place (like src/.libs/libblah.so) Jun 12 14:54:26 oh I see, in the package openwrt makefile can you access the ipkg-install when installing the ipkg-? Jun 12 14:55:12 granted for simple packages it might seem overkill Jun 12 14:55:12 sweet - that is exactly what I thought. So I guess in your package Makefile you'd do INSTALL_DATA $(PKG_BUILD_DIR)/ipkg-install/... $(1)/... Jun 12 14:55:26 is that correct? Jun 12 14:55:46 but large ones with lots of loadable plugin objects or huge amounts of resource files you really do not want to maintain a 2nd install procedure Jun 12 14:56:06 gotcha jow that makes sense. So the OpenWRT install lets you be more discerning about what you install Jun 12 14:57:22 So just one last question. If you are indeed happy with the upstream install. Can you just do cp -r $(PKG_BUILD_DIR)/ipkg-install/* $(1)? Jun 12 14:57:23 yes this is almost correct Jun 12 14:57:45 instead of $(PKG_BUILD_DIR)/ipkg-install/ you should refer to $(PKG_INSTALL_DIR) Jun 12 14:57:47 jow: could you do me the favour and close https://patchwork.ozlabs.org/patch/926983/ as rejected Jun 12 14:58:14 Thanks for your time and help jow and karlp Jun 12 14:58:17 oh nooo my patch dies!!!!! :-) :-) Jun 12 15:08:34 mkresin: done Jun 12 15:09:51 Neighbor1111111: you're welcome Jun 12 15:10:05 jow: thx Jun 12 15:12:14 Neighbor1111111: you can also use $(CP) to make sure things get copied nicely. Jun 12 15:13:14 :) Jun 12 15:14:51 karlp: just updated lighttpd PR, first time to submit a PR, is it OK or i need re-do a new PR? Jun 12 15:15:06 just push -f to the same branch, it will autoupdate the PR Jun 12 15:16:06 updated it via github UI directly Jun 12 15:16:39 can that fix the missing Sob? Jun 12 15:16:47 ausjke: that will just update the description text (which is initially copied from the commit Jun 12 15:17:11 changes made to that textbox will not affect the commits in the pr Jun 12 15:17:34 ok let me check how to do PR right Jun 12 15:24:35 karlp: for the lighttpd patch file PR, just checked it's fixed in packages/master-branch, maybe just merge from master? Jun 12 15:25:12 the issue is on 18.06, unless 18.06 is frozen on packages/, a merge will suffice Jun 12 15:27:20 that's something for you and the package maintainer. Jun 12 15:27:29 I was just skimming over and doing basic sanity reviews Jun 12 15:28:11 how does the patch even apply? Jun 12 15:29:51 did not find maintainer's info in the Makefile, never did PR, checking around Jun 12 15:30:12 look harder in the makefile? Jun 12 15:30:18 it's even in caps. MAINTAINER Jun 12 15:30:32 and like the template says, look at the git history for the package files... Jun 12 15:33:01 git log -p Makefile found Maintainer email back in 2014 commit, it's removed from Makefile sometime since then, anyways Jun 12 15:33:39 or shit, it's still there Jun 12 15:33:55 will contact Jun 12 15:34:18 for now please deny that PR Jun 12 15:34:27 jsut close it yourself. Jun 12 15:34:44 what makefile are you looking at? Jun 12 15:34:54 just actually open up net/lightttpd/Makefile Jun 12 15:34:56 net/lighttpd/Makefile Jun 12 15:35:02 now search for MAINTAINER Jun 12 15:35:08 somehow i did not spot MAINTAINER :( Jun 12 15:35:17 look harder :) Jun 12 15:35:17 old guy here Jun 12 15:35:51 so, MAINTAINER and the git log of, say, "git log net/lightttpd" shows you the people you may wish to reach out to. Jun 12 15:35:58 (there's no easy way to work out their github nicks though) Jun 12 15:37:37 i did 'git log -p Makefile > /tmp/xyz' then 'vi /tmp/xyz' to search with 'set -ic', did not realized it's all uppercase Jun 12 15:37:52 actually, anyway to search git-log while ignoring cases Jun 12 15:38:47 only know using vi to search case-insensitive via CLI for git-log Jun 12 15:39:35 look like 'git log -i' will do it, cool Jun 12 15:40:40 sometimes it's faster to just look at the current file, rather than reading the entire log of changes for when it might have been introduced a long time ago. Jun 12 15:41:37 yes i'm stupid enough to not see MAINTAINER in current Makefile Jun 12 15:41:48 anywa, 'git grep -i maint' will find it quick Jun 12 16:09:19 hello - anybody here who has some experience with ramips devices? Jun 12 16:09:42 i got a bootflag for tftp recovery here, which seems to be set after loading openwrt, but not with the stock firmware Jun 12 16:10:01 and i cant find the location in flash that could correspond to the flag Jun 12 16:10:22 is there any other sram or smth on the mt76x8 devices that could hold the flag? Jun 12 16:45:07 how do you stop / start a network interface from shell ? Jun 12 16:48:50 ifconfig up/down? or use ip link Jun 12 16:49:57 ldir: I just use /etc/init.d/network restart (and rely on ssh "just working" to stay connected) Jun 12 16:51:39 turns out I can use ifup / ifdown logical_if_name Jun 12 17:10:52 ifup / ifdown; /etc/init.d/network start/stop/restart; wifi/wifi down; Jun 12 17:14:25 why is uclibc++(70k) and stdcpp(370k) that small? Jun 12 17:16:01 ok, found my error: reserved partitions should not be overwritten ... Jun 12 17:18:31 ausjke: what do you mean "why are they small" ? Jun 12 17:18:32 morning Jun 12 17:18:38 don't do it ultito Jun 12 17:18:52 ? Jun 12 17:19:05 this is where you normally ask "is 18.06 available yet?!" Jun 12 17:20:10 karlp: need do some cpp on openwrt and worried about the c++ lib size, turns out they're not big Jun 12 17:21:04 karlp: is 18.06 available yet? /me awaits a stoning as per the life of brian stoning scene :-) Jun 12 17:21:27 i'll fetch my fake beard, brb Jun 12 17:21:37 packet of gravel? Jun 12 17:21:47 ausjke: even on my laptop, libstdc++ is smalelr than libc. Jun 12 17:21:48 got my own! Jun 12 17:21:50 snapshot runs good enough, let's just release today by calling it a 'rolling' release Jun 12 17:21:58 :) Jun 12 17:23:37 * ausjke is trying to get ipset+dnsmasq working with 60000 dns names Jun 12 17:33:26 karlp: me? No, that had to be someone else... :-D Jun 12 18:06:11 Hi can some pleas give me the link to the 18.6 snapshots again pleas Jun 12 18:06:22 I did have it but now cant find it Jun 12 18:17:52 Tapper: https://downloads.lede-project.org/releases/18.06-SNAPSHOT/targets/ Jun 12 18:20:08 hi guys, I'm having troubles to set parameters (ex: diversity) to an ath5k and an ath9k cards, it seems to ignore the /etc/config/wireless settings. How can I be sure that the parameters are committed? Jun 12 18:21:58 in particular, I need to test my new antennas setting diversity rxantenna, txantenna, txpower, and so on. But I can't get a clue without fixing those (and other) parameters. Jun 12 18:29:22 Hi. I have a TP-Link Archer C7 running Lede 17.01.4 (r3560-79f57e422d). I've seen issues lately where DNS just plain stops working. Perhaps it comes back after some time, but I don't really think it does. Previously I've rebooted it, but this time I stopped and started dnsmasq, and it recovered. Jun 12 18:29:34 DNS straight to the upstream servers worked fine. Jun 12 18:29:38 Have anyone seen this? Jun 12 18:31:09 pkgadd thanks dude Jun 12 18:53:14 Can one use a newer upstream uboot when building an image? Jun 12 18:55:19 https://downloads.openwrt.org/snapshots/targets/ar71xx/generic/openwrt-ar71xx-generic-archer-c7-v2-il-squashfs-sysupgrade.bin <- is this the correct image for TP-Link Archer C7v2? Jun 12 18:56:00 Hmm, perhaps not il Jun 12 18:56:04 What's -il-? Jun 12 19:02:57 a special hardware variant (well, different hwid enforced by the bootloader) for Israel Jun 12 19:03:06 pkgadd: Ah, thanks Jun 12 19:03:20 So dropping the -il- and I should be on the correct version? Jun 12 19:04:17 they have very strict regulatory domain requirements, which require vendors to make extra sure to obey the required settings and not to allow flashing firmwares for another country Jun 12 19:04:57 the non-il version is what you want, unless you've bought your device in Israel Jun 12 19:05:44 Yeah, non-il sounds good Jun 12 19:06:06 I wonder what bites dnsmasq, though Jun 12 19:09:37 hi, why do pakcages that contain a service sometimes disable it on package removal? Normal removal of a package already seem to do that. Jun 12 19:10:41 example: https://github.com/openwrt/packages/blob/master/net/vpnbypass/Makefile#L85 Jun 12 19:24:48 ldir: How go the ATM woes? Jun 12 19:25:36 not well at all. Jun 12 19:26:51 Monkeh: mkresin and I are hitting more questions than answer.... I'm broken for the day. Jun 12 19:27:26 Legacy crap built on legacy crap then wrapped in legacy crap and given a nice coat of paint? Jun 12 19:27:34 With a side of spaghetti? Jun 12 19:28:37 Something like that. We're both trying to work out why something doesn't work but starting from a point of not know how it's supposed to work in the first place. Jun 12 19:29:28 Sounds like a pretty normal day then :) Jun 12 19:29:47 which has to be read as: The code path that is for sure broken isn't called at all Jun 12 19:30:24 the pppoe stuff works because it uses a bridged ATM setup.... so it presents as an ethernet interface. Jun 12 19:31:05 pppoa uses the ATM layer directly....somehow. Jun 12 19:31:45 Probably just not saying the right prayer to the machine spirit Jun 12 19:32:07 what platform are you playing with? Jun 12 19:32:10 I would gladly sacrifice a goat if it would cure it! Jun 12 19:32:33 lantiq Jun 12 19:32:39 ah Jun 12 19:32:45 ive had success Jun 12 19:32:52 not tried 4.14 though Jun 12 19:33:02 but it requires very specific config Jun 12 19:33:20 had to do lots of trial and error Jun 12 19:33:56 ldir: Perhaps I should find my uplifting primre and see if there's a suitable prayer Jun 12 19:34:12 (presumably nobody has any idea what I'm on about now) Jun 12 19:34:19 I have a working system on K4.9 - it breaks on 4.14 Jun 12 19:34:51 ah Jun 12 19:34:53 lantiq was switched to K4.14 and previously working PPPoA setups stopped. Jun 12 19:35:11 i never got chance to test pppoa on 4.14 Jun 12 19:35:18 only pppoe on ptm Jun 12 19:35:34 pppoe is fine on atm and ptm Jun 12 19:35:44 I am thankfully free of ATM Jun 12 19:36:13 i do have some adsl sites left though Jun 12 19:36:25 i womt upgrade just yet :D Jun 12 19:36:28 wont Jun 12 19:36:33 Doing the hack that some ISPs support running PPPoEoA makes it work again.... but it is less efficient (more overhead) and some ISP really only want PPP in ATM AAL5 Jun 12 19:36:42 i bumped my own tree to 4.14 ages ago Jun 12 19:36:54 yeah thats no good Jun 12 19:37:58 are you stepping through commits? Jun 12 19:38:16 what you mean in openwrt tree? Jun 12 19:38:25 or is it just an impossible task given difference between .9 abd .14? Jun 12 19:38:33 kernel commits to atm Jun 12 19:39:18 we've looked at that...and there's a change that would break stuff. But here's where it gets fun... Jun 12 19:39:22 could be the dsl driver i guess Jun 12 19:40:13 so we fixed up the driver to speak the new kernel method. And put some debug in. To find our code whether fixed or not.. doesn't get called. Jun 12 19:40:19 gotta go - food Jun 12 19:40:33 doh, fun Jun 12 19:40:54 i dont have a dslam in any facility, i should probably get one Jun 12 19:44:51 I'd really like to pick one up for home.. Jun 12 19:45:06 For educational purposes, obviously Jun 12 19:54:45 Anyone doing an OpenWRT/LEDE install for wi-fi on a boat? Jun 12 19:57:20 is that a question for a vendor doing that or do you need help with something specific? Jun 12 19:57:44 you mean like "going river-rafting and using putty on your mobile to remotely do something"? Jun 12 19:58:18 I was *assuming* jason_meredith meant a slightly bigger boat but who knows? :) Jun 12 19:58:47 its a liveaboard :) 40' sailboat Jun 12 19:59:01 congrats Jun 12 19:59:10 good news is: OpenWRT won't know Jun 12 19:59:49 so it works the same as on land. Have fun with radio compliance within coastal areas / harbours though Jun 12 20:00:45 Im wanting to use a picostation out on the rail to grab a local wifi connect the picostation to a MR3020 to use as a local AP/router Jun 12 20:01:40 so have a local BoatWifi hotspot that is getting its internet from an externally mounted picostation Jun 12 20:01:49 make sense? Jun 12 20:03:05 just not sure which mode I should be using on the 3020 running OpenWRT Jun 12 20:03:32 if I want it to act as an Access Point and Router, which do I select? Jun 12 20:09:41 I have no idea what the difference between these in your mind is. But technically the MR-3020 would be an access point that's making the network on its eth0 interface available to the wifi clients, too Jun 12 20:11:03 So is the mr-3020 wlan0 interface giving out dhcp addresses? Jun 12 20:12:02 if you tell it to do so Jun 12 20:12:47 guess thats where Im confused, I only see DHCP settings with LAN Jun 12 20:16:22 if you use Luci there is a "DHCP Server" section at the bottom of :/cgi-bin/luci/admin/network/network/ Jun 12 20:21:14 I see that here: http://192.168.1.1/cgi-bin/luci/admin/network/network/lan lan is showing br-lan, so its bridging the wifi interface to the lan interface, correct? Jun 12 20:22:21 would I need a crossover cable between the PicoStation and the 3020? or will a regular patch cable work Jun 12 20:23:15 I do not really know about your power envelope, but the tl-mr3020 would be a bit too limited for my own taste Jun 12 20:23:41 I need low voltage, its a sailboat Jun 12 20:24:07 we run off batteries a lot of the time Jun 12 20:31:12 * ldir is back Jun 12 20:43:47 hello guys! Jun 12 20:44:23 one question: its possible install LEDE in DSL2750B rev C1? Jun 12 20:48:46 i try firmware A4001N but not WAN Jun 12 20:49:13 I create new Interface, but not success Jun 12 20:51:47 if by "WAN" you mean the xDSL support, that's never going to work Jun 12 20:55:53 but its not possible LAN to WAN? Jun 12 20:56:31 eg. LAN4 to ===> WAN port? Jun 12 20:57:22 that would be a matter of configuring the switch and vlan setup accordingly Jun 12 20:59:05 @pkadd yeah... I trying, but not sucess Jun 12 20:59:22 @pkgadd Jun 12 20:59:25 neoraider: Any thoughts on the squashs blocksize discussion here? https://patchwork.ozlabs.org/patch/922805/ Jun 12 21:00:04 oi Jun 12 21:00:17 Crap, everyone hide Jun 12 21:00:18 :) Jun 12 21:01:56 run for the hills, baby. Jun 12 21:03:06 boot perf impact is, well, bad Jun 12 21:03:33 It is quite ugly, yes Jun 12 21:03:48 Hence my inclination to split the difference Jun 12 21:04:22 things ain't fun if they ain't breaking. Jun 12 21:04:28 But I'd like to hear from people who may have memory constraints on such devices Jun 12 21:05:48 i have a 32 MB device but it's just an AP now Jun 12 21:05:56 no LuCI or whatever, just AP Jun 12 21:06:03 same here Jun 12 21:06:22 8 MB flash devices tend to have more issues with 32 MB RAM than 4 MB flash ones, just because the availability of more flash coaxes one into installing more Jun 12 21:06:53 Right, and this change is explicitly to squeeze another erase block out on 4M devices Jun 12 21:07:07 It's getting that tight. Jun 12 21:07:49 :) Jun 12 21:07:51 (I have some wpad/wpa_supplicant changes to squeeze some more out for my application forthcoming, though) Jun 12 21:08:11 even with the tiny subtarget, luci is out i suppose on 4/32 for 18.06 Jun 12 21:08:21 it was already pretty much choking 4/32 with 17.01 Jun 12 21:08:29 No, I believe it'll squeeze Jun 12 21:08:47 At least, with a few little tweaks like that patch Jun 12 21:08:48 on the flash, but it will kill the ram Jun 12 21:09:01 Borromini: as long as ar71xx stays on kernel 4.9 it should be tight, but possible - kernel 4.14 (at least for ar71xx) kills luci on 4 MB flash Jun 12 21:09:24 Borromini: Those RAM tests were all with luci installed - haven't checked consumption after using it a bit, though Jun 12 21:09:27 so the move to dts won't compensate enough? Jun 12 21:09:40 to account for swapping 4.9 for 4.14 Jun 12 21:09:47 It helps, but 4.14 got fat Jun 12 21:09:48 i thought/hoped that would mean lighter kernels Jun 12 21:09:51 hehe Jun 12 21:10:07 Upstream needs to go on a diet Jun 12 21:10:08 Borromini: can't say for sure, none of my devices works with ath79 yet Jun 12 21:10:11 Monkeh: my 1043 v1 would be usable until you tried to use luci Jun 12 21:10:26 pkgadd: yeah, same here. had only the 1043 v1 to test but i bricked that :D :D Jun 12 21:10:37 Well I'm browsing around in luci right now on that device Jun 12 21:10:41 and? Jun 12 21:10:52 It's made all of about 100K difference in RAM usage? Jun 12 21:11:07 Borromini: ath79 is supposed to work on the tl-wr1043ndv1 already, but both of my tests didn't work (no network access, no failsafe) Jun 12 21:11:14 Monkeh: maybe it was between my ears then. Jun 12 21:11:17 There's still a whole 12M free on this Jun 12 21:11:26 pkgadd: so you got it back with serial? Jun 12 21:11:31 Mind you it's very barebones currently - no firewall, no NAT going on, etc Jun 12 21:11:42 I use this one purely as a bridge, so it's slimmed Jun 12 21:11:47 Borromini: tftp Jun 12 21:11:59 But I'm pretty sure a no-frills device would just about survive Jun 12 21:12:25 Not.. entirely convinced on opkg via luci, though Jun 12 21:12:28 i had made a proposal last year that could help squeeze out one more EB (among other benefits). It remained largely ignored tho :) Jun 12 21:12:45 Haven't tried, there's only 84K free.. Jun 12 21:12:46 Monkeh: oh. thought it was a default config Jun 12 21:13:00 anyhow, time to bail. bbl Jun 12 21:13:15 unfortunately my tl-wr941nd v2.4 (4 MB flash, marvell switch) doesn't have tftp push-button support in its u-boot implementation, while the tl-wr1043ndv1 does... Jun 12 21:13:15 Which is why I need to make some changes with wpad/wpa_supplicant - full wpad is rather large and I only need one thing from it (which is rather small) Jun 12 21:13:21 f00b4r0: What proposal? Jun 12 21:13:55 pkgadd: You're making me consider grabbing a WR1043ND to see what's going on Jun 12 21:13:59 Please don't do that. I'm broke. Jun 12 21:13:59 :) Jun 12 21:15:28 if only the case wouldn't be that stubborn ;) Jun 12 21:15:38 (admitted, there are worse) Jun 12 21:15:42 Monkeh: regarding sysupgrade format. Should be easy to find on the m-l, I sent maybe 4 emails there :) Really bailing now Jun 12 21:15:54 f00b4r0: I'll have a dig, thanks Jun 12 21:16:06 was at end of 2017 Jun 12 21:16:08 pkgadd: How does one ID a v1 anyway, case style? Jun 12 21:16:29 Monkeh: yep, the case design changed between v1 and v2 Jun 12 21:16:40 Black is the newer models I take it? Jun 12 21:16:55 Monkeh: https://lists.openwrt.org/pipermail/openwrt-devel/2017-November/009879.html Jun 12 21:16:59 i'm out now Jun 12 21:17:03 f00b4r0: Shoo Jun 12 21:17:07 ;) Jun 12 21:17:07 Monkeh: yes, the old one was white Jun 12 21:17:20 pkgadd: https://www.ebay.co.uk/itm/TP-Link-TL-WR1043ND-300-Mbps-Gigabit-Wireless-N-Router-bundle-extra-antennae/223014437966 Jun 12 21:17:21 *sigh* Jun 12 21:17:35 that is a v1, yes Jun 12 21:17:45 Not a horrible price.. perhaps I'll pick one up to see if there's actually something wrong Jun 12 21:17:51 but I really wouldn't recommend buying one these days Jun 12 21:17:57 People seem very proud of them Jun 12 21:18:20 they were among the first affordable routers with 1 GBit/s ethernet Jun 12 21:19:00 I really like mine, even if I don't really use it anymore (expect for doing weird test setups with strongswan) Jun 12 21:19:36 ...I just regret having bought a tl-wr941nd v2.4 half a year before its release Jun 12 21:19:48 pkgadd: Well, there seems to be a limited number of people working on ath79, so I'm always on the lookout for something affordable which needs a port. Jun 12 21:20:15 v1 is a yellowed white by now Jun 12 21:20:21 looks ugly and old. Jun 12 21:20:32 v2 and higher are blue UFOs, they look more non-descript :P Jun 12 21:20:35 Borromini: mine looks pretty much as new Jun 12 21:20:44 pkgadd: your v1? Jun 12 21:20:48 yes Jun 12 21:20:53 did you bleach the plastic lol Jun 12 21:21:00 nope Jun 12 21:21:06 Those things are meant to be blue? Jun 12 21:21:07 you probably don't live in a city :) Jun 12 21:21:15 Monkeh: v2 is a dark blue Jun 12 21:21:22 Very dark, apparently.. Jun 12 21:22:37 Borromini: depends on your definition of 'city', technically a population of 20'000 does qualify for that label Jun 12 21:23:44 https://openwrt.org/toh/tp-link/tl-wr1043nd that visual guide is correct, only v1 was white Jun 12 21:30:40 speaking of white things, anyone interested in an EasyBox 904 xdsl? I got two here I will not be using anymore. Jun 12 21:31:12 maybe someone who's actually interested in developing its port? Jun 12 21:32:45 pkgadd: hardly offers metropolitan problems :) Jun 12 21:33:19 it would be such a nice device, if Arcadyan wouldn't have "customized" the hardware a bit too much (non-standard ECC badblock markers, RaLink rt3883 iNIC), everything else wouldn't be too difficult (o.k., display support is a unique feature/ problem) Jun 12 21:33:45 they would be rather nice, but honestly I don't have the energy to invest more time in it. Jun 12 21:34:11 I went for a FritzBox 3370 now, and that bugger just works lovely out of the box. Jun 12 21:35:43 and it's got mainline OpenWrt support. Jun 12 21:38:07 not every version of it i was told recently in this channel Jun 12 21:38:13 thats why i hesitate to buy Jun 12 21:38:27 apparantly revision 1 is unsupported, yes. Jun 12 21:38:48 just get a newer version, you can tell from the serial number (which lots of ebay sellers helpfully post the image of) Jun 12 21:39:24 how to make 'ping -I nic1-ipaddr nic2-ipaddr' work? ip_forward is 1, even enabled proxy-arp Jun 12 21:39:40 default route is sane to me too Jun 12 21:42:34 mwarning: no particular reason, the disable does seem pointless, possibly just paranoid attempts to "clean up" on removal Jun 12 21:44:03 Hi I have a 1043ND some one can have for postage from UK Jun 12 21:44:07 v2.1 Jun 12 21:44:33 drmr: how do i tell from the serial? is it obvious? Jun 12 21:44:42 even disable rp_filter, nothing works by google Jun 12 21:44:48 rotanid: somewhat. they code in the production time. Jun 12 21:45:13 Monkeh ? Jun 12 21:45:24 rotanid: first letter A=2010, B=2011, C=2012, D=2013, the next three digits are production week and day-of week. Jun 12 21:46:13 rotanid: apparantly it has been produced between 2011 and 2013, mine is a D305... so 30th week of 2013, a revision 5. Jun 12 21:46:53 I'm willing to wager that if you get one starting in "D" you wouldn't get a revision 1. Jun 12 21:47:25 drmr: thanks! Jun 12 21:48:01 (no promises though) Jun 12 21:48:55 also you still need to open them to check the flash chip maker. Jun 12 21:49:41 Monkeh are you in the UK? Jun 12 21:49:43 ftp'ing in OpenWrt is slightly tricky, mainly because most of Windows clients su-u-uu-u-u-uuu-uck. Jun 12 21:50:35 karlp: ok, thanks. Is there a way to prevent to enable a service by default? Or is it the expected default? Jun 12 21:50:46 (after trying half a dozen clients, Total Commander worked for me in the end, which funnily is my go-to everyday ftp client anyway) Jun 12 21:51:02 I at least consider installing a service and having it be enabled to be expected, and iirc, that's how things are now. Jun 12 21:51:17 in the old days, things that were built in were enabled, and things that were added later weren't, which was confusing. Jun 12 21:51:32 if you really want to customize things like that, build your own images. Jun 12 21:51:54 or evaluate whether you really need to be bothered by such things. Jun 12 21:52:07 _some_ services ship in "enabled, but with a config flag deactivating them" Jun 12 21:52:15 similar to how wifi is disabled out of the box Jun 12 21:52:16 drmr: netkit-ftp-ssl works just as well Jun 12 21:52:36 ...at least on the newer AVM devices, but I don't expect there to be a significant different Jun 12 21:52:40 difference* Jun 12 21:52:48 pkgadd: good to know. Jun 12 21:53:02 For the low mem devices OpenWRT should do a build to make them in to dumb APs with all the stuff removed that is need for a proper router. Jun 12 21:53:46 EG dnsmasq and firewall Jun 12 21:53:47 karlp: yes, it is confusing. the specific problem I have is that starting my vpn package generates a key that is stored in the uci config and the uci commit drops all help texts from the config file. Which is unfortunate Jun 12 21:54:14 and all the protocols needed for wan connections Jun 12 21:54:37 pkgadd: then again, regular *nix ftp should work just fine. it's just Windows clients where apparantly everyone needed to create their own set of commands, for reasons that are completely beyond me. Jun 12 21:55:21 well, I never even considered testing a windows ftp client ;) Jun 12 21:55:31 and Windows' own FTP command line client cannot do passive transfers. don't ask. Jun 12 21:55:44 mwarning: that doesn't sound like anythign to do with, "installing services enables them" Jun 12 21:56:00 it sounds like, "the init script for this _particular_ package is weird and possibly wrong" Jun 12 21:56:39 are you really storing things as comments in uci files? Jun 12 21:57:04 that's totally unreliable :) Jun 12 21:57:33 things like luci apps that save models again will write things that are semantically the same as before, but not necessarily identical. Jun 12 21:58:41 karlp: that's interesting, just today I found a PHP script that, I kid you not, stores its own configuration in a JSON-encoded comment. it reads itself, cuts off the commend markers and tries to json-decode the line. Jun 12 21:59:14 whoever came up with that should be nailed to a random city wall. butt naked, face against the wall. Jun 12 21:59:18 drmr: as in, "option comment 'long string of json'" ? Jun 12 21:59:31 or as in "# hoho magic " ? Jun 12 21:59:50 the former is valid, but odd, the latter is stupid :) Jun 12 22:00:17 drmr:also, I'm talking very specifically about uci config files Jun 12 22:00:23 karlp: second line: // { 'configuration' : '...' } Jun 12 22:00:30 hi. i configured openconnect(by juniper connection), by log - the connection is ok, ping ok. but in case - i defined firewall zone with it (in goal to redirect traffic) - i have reloading firewall and >> kernel.info segfault .. error 4. is it a bug ? :) Jun 12 22:00:32 karlp: I realize that. Jun 12 22:00:49 old fahsioned update_rc.d scripts used special comment formats too didn't they? Jun 12 22:01:09 mt7621a is 32bit mipsel, any mipsel 64bit used in openwrt? Jun 12 22:01:22 karlp: it's just so odd that the topic came up right now. Jun 12 22:01:48 ausjke: name a mipsle64bit soc? Jun 12 22:02:09 karlp: not sure actually, recall broadcom etc had it Jun 12 22:02:52 cavium has a few too Jun 12 22:03:15 ARMv8/ 64 bit is slowly appearing on the highend router market, e.g. ipq807x and annapurna labs Jun 12 22:04:29 arm still has a long way to go for network i feel, performance wise Jun 12 22:04:49 its IO was optimized for media processing, not net IO or not even pcie IO Jun 12 22:05:07 i mean the bus architecture Jun 12 22:05:09 i had an issue i thought was due to a SIM slot not working, but it was actually the SIM itself Jun 12 22:05:13 of, the 'fabric' Jun 12 22:05:22 SIM had gotten fried, because I tried to hot-swap it Jun 12 22:05:23 s/of/or/ Jun 12 22:05:26 at least mvebu is very fast, so should be mt7622/ mt7622 Jun 12 22:05:44 the slot doesn't support hot swapping Jun 12 22:06:09 ausjke: you're making shit up again aren't you :) Jun 12 22:06:55 * ausjke thinks nobody does good networking other than powerpc, though it's dying Jun 12 22:07:48 you're living in the past, at least 10 years Jun 12 22:07:54 when you have a dts line like, "<&pinctrl 4 GPIO_ACTIVE_LOW>" how do you specify which gpiochip that's attached to? Jun 12 22:08:09 I have "pin 4" on two chips, one is a led, one is a button? Jun 12 22:09:48 tell me one arm can do pciex8, or 40+Gbps network, yes ppc does it a few years ago already Jun 12 22:10:16 but then asic and intel NIC ruined it Jun 12 22:10:25 "ruined" it Jun 12 22:10:54 * Borromini hugs his realtek nics Jun 12 22:13:24 I'd really love to switch to 10 GBit/s, but all that stuff is still way too expensive for home uses Jun 12 22:14:32 (and no, don't start about 2.5 GBit/s or 5 GBit/s, that stuff isn't any cheaper) Jun 12 22:16:13 does any want this 1043nd before i go to bed? Jun 12 22:16:38 I will be going to the post office in the morning. Jun 12 22:16:50 reading gnubee, not cheap as mt7621a board is not expensive, but very interesting, any place to see some NAS benchmark data for that Jun 12 22:18:19 For any RAID5 beyond a hobby toy you pretty much have to go with x86... Jun 12 22:19:32 * karlp disocvers that the pinctrl gpio numbers are just consecutive. so gpio 4 on bank 2 is just 36... Jun 12 22:21:05 ah, damned, that should have clicked with me (it's the same on ipq8065 with the LED GPIOs handled by ath10k) Jun 12 22:25:00 np :) Jun 12 22:25:19 now on to the hrader things, like, why doesn't the wifi actually _work_ even thogh it probes and responds well to iw list, and _claims_ to be online? Jun 12 22:25:52 I can't test the SFP port, as I don't have that fibre anymore, and I can't test nor care about the dsl ports, so this board is ~finished Jun 12 22:25:59 10g is cheap Jun 12 22:26:07 cheap is relative Jun 12 22:26:11 x520 is 100 gnp a port Jun 12 22:26:27 and can do wirespeed wity dpdk Jun 12 22:26:30 with Jun 12 22:26:42 gbp even Jun 12 22:27:03 100 per port os *really* cheap for a nic Jun 12 22:27:13 again, that's relative... Jun 12 22:27:17 switches only just started hotting 100 a port Jun 12 22:27:23 hitting Jun 12 22:27:44 for the cost/market its very cheap Jun 12 22:28:01 personal circumstances are irrrelevant really Jun 12 22:28:20 no. they're the only thing that is relevant when it comes to whether something is "expensive" :) Jun 12 22:28:49 they arent consumer products though Jun 12 22:29:22 jwh: *I* would be fine with ~40 EUR a card and <100 for a 8-port switch, which is considerably more than most other would be willing to pay Jun 12 22:30:16 there's a reason why I explicitly mentioned home uses Jun 12 22:31:38 switches are pretty cheap tbf Jun 12 22:31:55 nics are more expensive because they are more intricate Jun 12 22:32:12 the chips have to do much much more Jun 12 22:33:00 cant just switch ethernet frames, have to do offload, virtualisation crap etc Jun 12 22:33:27 all of them also do variable packet matching now Jun 12 22:33:43 for nic packet filters Jun 12 22:34:00 none of that is easy or cheap Jun 12 22:35:13 did you just say they weren't cheap? ;) Jun 12 22:41:55 I wonder if lantiq chipsets are prone to overheating. Jun 12 22:42:32 my easybox 904xdsl is easily between 65-70 °C Jun 12 22:42:50 without taxing it hard Jun 12 22:43:24 I just had my 3370 lock up on the pppoe connection. whatever I did, dsl restart, network restart, it wouldn't reconnect pppoe anymore. Jun 12 22:43:59 a reboot helped, but only for a few minutes, then a lockup again. looked into the cupboard, and that thing was *hot*. Jun 12 22:44:14 freed it a bit, made sure it got better airflow, now it seems to run fine again. Jun 12 22:45:12 made that observation on a completely different lantiq machine (tp-link 9980) too, that one I have actually run a watchdog that reboots the machine if wan stays down for 10 minutes. Jun 12 22:50:13 although I'm not sure about the connection between pppd and a hot cpu. Jun 12 23:10:06 drmr: They seem to run quite warm but I haven't had one get to the level of overheating so far Jun 12 23:56:18 Monkeh: something is a bit off there. it most certainly is a Lantiq thing though. maybe the dsl thingy runs full or whatever, I wouldn't know. Jun 12 23:59:22 the AVM Fritz!Box line has a bad case designs when it comes to heat dissipation and natural convection, it's known that you're better of mounting them vertically rather than setting them down Jun 13 00:01:05 I guess I'm just a bit puzzled that it affects the ppp handshake, but seemingly nothing else. Jun 13 00:02:01 I can still ssh in, dsl stays synced and even resyncs when restarted. it's like something breaks and doesn't un-break until rebooted. Jun 13 00:03:02 I'm-a need a sammich. Jun 13 00:06:02 but the vertical mounting is a good idea. Jun 13 00:06:58 karlp: npt as cheap as switch ports, no :p Jun 13 00:07:45 but i still think that for what they are, they are pretty inexpensive Jun 13 00:08:26 but nobody needs 10g at home, and ge nics are practically nothing now Jun 13 00:22:56 last I heard NAS devices are pretty popular, but 1 GBit/s throughput can't keep up with the transfer speeds of even a single HDD (not even to start about SSDs) anymore. and yes, there are people who deal with larger files (up to lower 3 figure GB range) at home regularly. yes, even with predominantly local disk space I'd really like network transfers to be faster, on a reasonable home budget Jun 13 00:29:40 Bring on the low cost NBASE-T. Please? Jun 13 00:31:40 doesn't feel like it - hardly cheaper than full 10 GBit/s Jun 13 00:31:52 neither more readily available Jun 13 00:32:04 That's why I said bring it on Jun 13 00:32:26 I was hoping to see a lot more 2.5G around Jun 13 00:32:44 im trying out some shell scripting but failed of course. can someone help me fixing this https://pastebin.com/DuF236Pm ? ty ^^ Jun 13 00:33:11 i think is a problem of quoting on the "-" in the string? Jun 13 00:50:16 omg nvm fixed it Jun 13 00:54:25 here the complete one maybe it is useful for someone https://pastebin.com/cRSWLmmx Jun 13 02:06:12 updated :> **** ENDING LOGGING AT Wed Jun 13 03:00:01 2018