**** BEGIN LOGGING AT Fri Oct 05 03:00:02 2018 Oct 05 07:01:06 nbd: here's a table of results for connecting the Linkit 7688 to various nearby access points in my testbed: https://pastebin.com/FJ08gDyN Oct 05 11:53:31 stintel: I only just noticd that your agentx enablign only enabled domain sockets, not tcp :) Oct 05 11:53:35 two years too late. Oct 05 12:37:07 karlp: :D Oct 05 12:52:15 just fiddling with the ui for it. an empty section enables agentx with /var/agentx/master, you provide a default of /var/run/agentx.sock just gotta make it clear how to turn it on and off if needed. Oct 05 12:52:28 and make sure my own stuff migrates from tcp:localhost:705 to the "right" socket. Oct 05 12:54:17 snmp :( Oct 05 12:55:25 hm, theres something going on with my dev environment Oct 05 12:55:38 oooh, I wonder if its newer gcc Oct 05 12:56:09 nope... both 8.2.0 Oct 05 12:56:18 e2fsck.h:598:13: warning: type of 'fatal_error' does not match original declaration [-Wlto-type-mismatch] Oct 05 12:56:21 extern void fatal_error(e2fsck_t ctx, const char * fmt_string); Oct 05 12:56:35 in e2fsprogs, target stage Oct 05 12:56:58 it won't build the kernel either though, unknown type __s8 or something Oct 05 12:58:02 but.. builders are 8.2.0 and they manage it just fine Oct 05 13:02:13 what's the sad face for on snmp? Oct 05 13:02:50 dunno, guess I just find it a pain in the ass these days Oct 05 13:02:53 karlp: oh, empty enables it too? Oct 05 13:03:07 but /var/agentx/master doesn't exist? Oct 05 13:03:12 its also pretty annoying to represent some info in snmp, given the limited data types Oct 05 13:03:13 it creates it itself Oct 05 13:03:29 I see, oh well, I don't really remember everything from back then Oct 05 13:03:47 but we can improve stuff if needed Oct 05 13:04:12 eyah, no biggie, was just trying to work out why one update in my stuff was now saying, "can't reach agentx master" :) Oct 05 13:04:37 although I have to say I dislike /var/agentx/master, sockets should go in /run or /var/run Oct 05 13:04:47 I think I can tweak the luci app documentation to say "delete this to turn it off", and if you want to add it, you can find it easily enough Oct 05 13:05:05 yeah, I'm ok with /var/run, the var/agentx/master is just net-snmp default setting. Oct 05 13:05:34 means I need to add the -x /var/run/agentx.sock path explicitly to subagents, btu that's fine, being explicit about things is less magical anyway Oct 05 13:05:48 whaaaat the hell Oct 05 13:06:16 alright Oct 05 13:06:17 jwh: well, it's like modbus. it "just works" even if it's not really tuned for anything. Oct 05 13:06:19 all of this is glibc related Oct 05 13:06:28 does nobody test with glibc? Oct 05 13:06:33 not really, no. Oct 05 13:06:51 I mean libc really shouldn't cause any of this Oct 05 13:06:59 but I guess if stuffs not being built properly Oct 05 13:07:44 http://ix.io/1ooy Oct 05 13:09:08 aaand Oct 05 13:09:30 http://ix.io/1ooz Oct 05 13:10:01 I mean, it could totally be a broken build environment that doesn't present itself with musl targets Oct 05 13:10:37 or maybe I need to go back to what I was doing before, building glibc targets on musl and vice versa, as that seemed to avoid weirdness Oct 05 13:11:15 all this hassle because musl won't implement fake ucontexts :((( Oct 05 13:12:20 I still feel like you must have serious legacy cruft to be using ucontexts anyway Oct 05 13:12:42 posix 2008 dropped it said, "this is dumb, don't use this" Oct 05 13:12:43 they were used fairly recently because they're fast Oct 05 13:12:52 s/used/added/ Oct 05 13:13:15 what's wrong with pthreads? Oct 05 13:13:38 this is designed to handle 1000s of ipoe/ppp sessions in userspace, so Oct 05 13:13:55 he must have had a valid reason for thinking ucontexts are faster Oct 05 13:14:43 I don't know enough about ucontexts or pthreads to ifdef it to make it musl compatible (at the expense of performance) Oct 05 13:14:52 and I'm not sure how else I can work around it Oct 05 13:15:15 or, they just had been using ucontexts for years and write it like that without even thinking? Oct 05 13:15:21 I *did* have a horrible bunch of patches that just reverted the switch to ucontexts but it's a nightmare to maintain and it's kinda old Oct 05 13:15:33 what application is this? Oct 05 13:15:33 I think they only appeared last year Oct 05 13:15:36 accel-ppp Oct 05 13:15:57 and you've asked thema bout this right? Oct 05 13:16:18 nope, won't even fix musl header stuff, so Oct 05 13:16:30 sounds like you should jsut toss it in the bin then. Oct 05 13:16:45 are you really planning on supporting thousands of ppp sessions on your little owrt router? Oct 05 13:16:53 linux is absolutely terrible as LNS, this is the only project even *close* to being useful :( Oct 05 13:17:01 karlp: none of my owrt routers are little :P Oct 05 13:17:12 except maybe the NTEs Oct 05 13:17:50 what's LNS to you? Oct 05 13:18:19 traditional sense, l2tp (ppp) termination, but modern lns' also terminate IPoE etc, metro ethernet Oct 05 13:18:56 https://github.com/xebd/accel-ppp/tree/master/accel-pppd/triton Oct 05 13:19:21 thats where the ucontext stuff is, I can revert the commits but would kinda like to at least figure out how to make it compatible with musl and send pr Oct 05 13:21:00 I was hoping someone had made a thin library that can pretend to do ucontexts but nope :( Oct 05 13:21:25 thjat would be such an incredibly niche library Oct 05 13:21:35 yeah Oct 05 13:21:44 but it would mean many things work on musl that don't currently Oct 05 13:21:57 they're already using pthreads, I don't get it Oct 05 13:22:47 hmmm Oct 05 13:24:23 running out of time to get this sorted, so not sure which option is best... revert specific commits and hope it doesn't die in the middle of the night, or try and figure out my glibc build issues Oct 05 13:25:59 run a glibc container..... ;) Oct 05 13:26:07 ontainers fix everythign right? Oct 05 13:26:11 haha! Oct 05 13:26:12 good one Oct 05 13:26:29 I mean I would settle for that... Oct 05 13:26:39 they cause more problems than they fix really Oct 05 13:26:46 fixing glibc is valid and should be done though, Oct 05 13:26:55 and that's at least meant to eb supported in openwrt, Oct 05 13:27:12 alright, well I can't figure out if its me or not Oct 05 13:29:02 nuked the lot, gonna see if it happens again Oct 05 13:31:50 curse me and my bleeding edge arch container, right? Oct 05 14:21:08 nope, broke at the same point Oct 05 14:22:21 This bug is annoying the fuck out of me: https://bugs.openwrt.org/index.php?do=details&task_id=1449 Oct 05 14:22:41 Anyone available to shortly mentor me on how to go about patching it to resolution? Oct 05 14:33:31 I'd suspect autoneg problems on the cable, Oct 05 14:33:36 not sure about "#patching it" Oct 05 14:34:52 karlp: explain? this looks like a regression, didn't have this before 18.06.X Oct 05 14:35:17 then it should be easy to bisect Oct 05 14:37:55 doesn't bisect mean that on each bisection step i have to reflash and wait for the random behavior to appear? Oct 05 14:38:21 that sounds... unefficient. unless i'm missing out on how to test the bisection? Oct 05 14:42:46 nbd: does this commit have anything to do with the behavior i'm seeing? https://github.com/openwrt/openwrt/commit/b605a84a74170d2a07f53512cd57cd564db77249 Oct 05 16:03:53 you were making out like it happened all the time, so easy to reproduce Oct 05 16:17:33 karlp: i like that, all those words used together in the same sentence, making out and so easy to reproduce Oct 05 16:22:58 hey everyone Oct 05 16:23:20 was there a change in how the list_cb callback in the init.d script works between LEDE 17 and OpenWRT 18? Oct 05 16:24:15 I'm getting stuff like uamdomain_ITEM1="www.facebook.com" and uamdomain_LENGTH="1" which is totally unexpected on LEDE Oct 05 16:24:49 I added a debug output and the debug is printing out correctly but then I find that _ITEM1 and _LENGTH="1" which I don't understand where it is coming from Oct 05 16:39:34 I found this in LEDE 17: https://github.com/openwrt/openwrt/blob/lede-17.01/package/base-files/files/lib/functions.sh#L73-L85 Oct 05 16:40:00 but it's there in openwrt 18 too Oct 05 16:44:05 I don't really understand what that piece of code does, I also don't understand why I'm seeing this issue on LEDE 17 but not on OpenWRT 18 Oct 05 19:15:08 russell--: Try using .11w or WDS to make things even more unreliable Oct 05 19:23:54 Monkeh: i'd be satisfied with ap and station mode Oct 05 19:24:26 weirdly, the mt7628 i have works Oct 05 19:41:57 russell--: Eh, throw some WDS at it and try again. ;) Oct 05 20:14:36 Monkeh: i abandoned WDS as soon as I possibly could, and haven't looked at it again in the last 10 years Oct 05 20:15:33 batman-adv is more reliable for us Oct 05 23:01:50 http://www.win-star.com/eshop/goods.php?id=233 Can anyone develop openwrt for this device, MT7621A+MT7615N+MT7603E? I can pay to develop. Please contact me via skype: giangtran1211 Oct 05 23:05:00 there are no drivers for mt7615, the rest should be reasonably Oct 05 23:05:22 Kiru_: where can we get them? Oct 05 23:05:39 I ordered 500 pcs for them Oct 05 23:05:48 Only for OEM Oct 05 23:06:35 I’m not sure if they sell online somewhere Oct 05 23:06:53 they got a lot of models Oct 05 23:06:58 why not ask someone _before_ ordering 500 ? Oct 05 23:07:43 Well they don’t offer sdk and I need some more features Oct 05 23:07:57 and I assume wifi to work is one of those features Oct 05 23:09:44 I see. Is there any hope? Oct 05 23:11:44 unless mediatek changes their policies, openwrt does not plan to support the 7615 Oct 05 23:11:45 the hope for mt7615 support would be non-zero, but I wouldn't expect that to happen anytime soon Oct 05 23:12:12 7603 and 7621 are supported Oct 05 23:12:30 s/7621/7612/ Oct 05 23:12:49 the soc Oct 05 23:13:49 and the 7603 support has some issues Oct 05 23:17:09 This router has very long range and very good wall penetration. I got 50mbps speedtest with 5ghz , 1 brick wall and 50m range Oct 05 23:17:47 Is there any good alternative long range router like that and also support openwrt? Any recommendation? Oct 05 23:17:53 that won't help you without mt7615 driver support, as the resulting range on the 5 GHz band would be 0 Oct 05 23:19:01 I’m looking for alternative now because my clients also request long range and stable router. 50 meter is good start Oct 05 23:19:19 Device connecting is android tablet and iphone **** ENDING LOGGING AT Sat Oct 06 03:00:01 2018