**** BEGIN LOGGING AT Mon Aug 31 19:23:55 2020 Aug 31 19:36:45 aparcar[m] Thankyou. Could you drop me a link when it's dun pleas? Aug 31 19:39:49 Of course Aug 31 19:50:06 aparcar[m]: Was just wondering again how we could implement PKG_RELEASE on multi-version packages like gcc or binutils Aug 31 19:51:09 if we want to have it there at all, I'd put it inside the ifeq ($(PKG_VERSION),xxx) clauses together with the PKG_HASH, so each version has it's own PKG_RELEASE Aug 31 19:51:50 but I'm not sure how helpful it would be at the end, and whether we really need it there Aug 31 20:00:16 is there an actual problem to solve? Aug 31 20:18:59 zorun: yes if we manage oot patches there must be a PKG_RELEASE Aug 31 20:25:27 adrianschmutzler: I guess that's the way to go except we want infinitely raising PKG_RELEASES, like used for base-files Aug 31 21:26:04 Hauke: yes. See https://github.com/openwrt/openwrt/pull/3004/commits/cf7c1e9d22c63c35e4eb03369031b4a845c1dc81 Aug 31 21:27:01 musl placed __LITTLE_ENDIAN in alltypes.h , which is leaking into mac80211 Aug 31 21:27:09 I have no idea how Aug 31 21:27:20 *placed in 1.2.0 Aug 31 21:45:45 mangix: thanks Aug 31 21:49:48 Hauke: https://github.com/openwrt/openwrt/pull/3373 Aug 31 21:50:43 mangix: your 10 year finding is a bit ridiculous ;) Aug 31 21:56:10 aparcar[m]: so.. with your browser, are not getting warnings about the self signed certs? Aug 31 21:56:23 I watned to retry this all today but was busy with other things Aug 31 21:58:01 I do, but it's a warning after all. Aug 31 21:58:19 All ubnt devices I used have self signed certs Aug 31 21:59:21 I feel that's dramatically understating it. Aug 31 21:59:59 SSL/TLS and X.509 are fundamentally broken by design for this kinda use... Aug 31 22:00:50 would be nice if there was a way to add a local cert or CA using QR-code or something like that Aug 31 22:02:02 but as long as we don't have that (and a way for the device to display the QR-code, or vendors to print them on a sticker and store the key in tpm) Aug 31 22:02:12 we this whole thing is doomed Aug 31 22:08:12 fast forward what would be a clean solution? APPs that handle the cert transparent to the user? A LED blink code which is scanable by webcams? Aug 31 22:12:32 mangix: the kernel includes stdarg.h from musl intentional: https://git.kernel.org/linus/30a7729771731971839cc969d2a321e6ea7a144b Aug 31 22:12:50 a DNS based acme challenge could be nice (referring to the mailing list), but obviously not quite possible out of the box (the http based challenges imho simply don't work for a router) Aug 31 22:12:53 and that includes include/bits/alltypes.h Aug 31 22:14:42 pkgadd: yea also most people have no public ip(v6) Aug 31 22:15:31 trsuted CAs will not sign certificates for (local?) IP addresses Aug 31 22:15:42 Hauke: correct Aug 31 22:17:03 aparcar[m]: that's going to change ;) I don't have a public IPv4 address anymore... (ftth/ Deutsche Glasfaser, cgNAT for IPv4, semi-static /56 IPv6 prefix) Aug 31 22:17:44 I don't have a IPv6 at all (Hawaiian Telcom Fiber) Aug 31 22:18:06 he.net has always served me well Aug 31 22:18:16 it's a kludge, but it helps Aug 31 22:18:46 it will probably take additional 20 years to get 98% of the host to IPv6 Aug 31 22:18:55 Yea I guess that's the way to go Aug 31 22:21:15 OEM firmwares are still problematic with IPv6 support, but at least it's usually there by now. root servers usually have it (if the admin enables it), shared hosters often don't - CDNs/ cloud setups often don't use it. wired ISPs are increasingly offering native IPv6, mobile ISPs are a mixed bag (T-Mobile has been optionally offering it for quite a while) Aug 31 22:22:10 IoT devices are a sad story, increasingly 'cloud' based, IPv4-only Aug 31 22:22:47 but, according to nlbwmon, 43.91% of the total traffic is IPv6 Aug 31 22:25:19 Hauke: I don't understand that commit. Something to backport? Aug 31 22:27:40 mangix: no it says that the kernel intentionally includes stdarg.h and this includes bits/alltypes.h Aug 31 22:28:00 so this is not leaking into the build but it is intentionally Aug 31 22:28:32 and the kernel has a lot of "#ifdef __LITTLE_ENDIAN" Aug 31 22:30:20 strange that only mac80211 breaks Aug 31 22:31:31 pkgadd: I remember reading somewhere about consumer routers claiming to "support" IPv6, but not supporting IPv6 firewalling. Aug 31 22:31:36 Hauke: are you okay switching uhttpd to EC by default? Aug 31 22:32:34 rsalvaterra: you mentioned some dhcpc6 and BusyBox stuff some while ago, do you remember? It was some legacy which could be cleaned up Aug 31 22:33:29 rsalvaterra: that issue appears to be generally fixed by now, the problems are usually smaller ones (which are still serious, but not in that regard). e.g. repeaters that have a serious packet drop for IPv6 or routers that have issues with session tracking (e.g. ssh sessions timing out) Aug 31 22:34:05 mangix: yes, __LITTLE_ENDIAN is only defined when included in backports, strange Aug 31 22:34:14 aparcar[m]: yes I am ok with using ECC Aug 31 22:34:18 aparcar[m]: I don't quite recall… I remember asking why we are using an external DHCPv6 client instead of the BusyBox one, yes… Aug 31 22:34:45 rsalvaterra: out of curiosity, are you still using your tl-wdr3600? I'm currently seeing mine freezing up while doing PPPoE (same with a tl-wdr4300, no issues with plain ethernet) Aug 31 22:35:39 it usually works on light usage (as in just the SIP pbx/ phones using it), but as soon as I start a speedtest or larger download, it freezes up within 10s-20s Aug 31 22:35:44 pkgadd: It's out of comission, at the time. I'm waiting for a USB 4G modem to use it as my holidays router. :P Aug 31 22:36:06 rsalvaterra: I'd recommend to test it properly, before relying on it ;) Aug 31 22:36:25 I never used PPP, though… but we'll soon find out. :) Aug 31 22:37:08 I've now switched to a BT home hub 5a for handling my VDSL/ PPPoE connection, which is 100% stable (ipq4019/ map-ac2200 works as well, just not enough ethernet ports to be useful) Aug 31 22:37:19 rsalvaterra: Want to test rustup for me? :D Aug 31 22:38:29 pkgadd: Ah, the Home Hubs are really great, too bad we can't find them here in Portugal (shipping from the UK is expensive)… but they're the only fully open xDSL devices, I believe. Aug 31 22:38:50 Grommish: Oh, dear… what are you up to this time? :D Aug 31 22:38:57 I've got it working ;p Aug 31 22:39:02 at least, on my platform Aug 31 22:39:09 there still seems to be weird issue with ar9344 (I've also seen hostapd segfaulting after DFS related channel switches, which kind of suggests a more common issue) Aug 31 22:40:00 rsalvaterra: I paid more for shipping to germany (~11 EUR) than for the device itself (~9 EUR), but it's a great device Aug 31 22:40:33 rsalvaterra: only thing I have to test right now is Suricata 5.0.3, but it'll test if rust installs correctly Aug 31 22:41:00 rsalvaterra: there are a few more VSDL2 capable lantiq VRX2xx devices, but most come with bad WLAN chipsets (lantiq WAV300 or RaLink/ rt2xxx) Aug 31 22:41:06 There are also many reports about this wan reconnect issue on ath79, but I'm not sure whether also on ar9344 ... Aug 31 22:41:50 pkgadd: Well… my regulatory.db is "special", maybe that's why I'm not hitting DFS issues…? Aug 31 22:42:09 I don't know if anyone wants to actually install Suricata (but me) but.. It'll give a good test Aug 31 22:42:40 adrianschmutzler: I'm still in the earlier debugging phases, but I don't see reconnecting, the router (both tl-wdr3600 and tl-wdr4300) completely freezes up while doing PPPoE under load, it no longer responds to pings, it doesn't recover without powering off Aug 31 22:43:52 I can test an old tl-wr1043nd v1/ ar9103 as well (haven't tried it yet), sadly I don't have any other ath79 devices than ar9103 or ar9344 Aug 31 22:44:34 Package libbsd is missing dependencies for the following libraries: Aug 31 22:44:35 libc.so.0 Aug 31 22:44:35 libgcc_s.so.1 Aug 31 22:44:37 hahahahaha Aug 31 22:44:40 pkgadd: Wow, that never happened to me, and I had the WDR3600 "managing" my 200/20 Mb/s connection while torrenting like crazy, for a couple of days (but this was on 19.07, not master) about three months ago. Aug 31 22:44:56 somebody broke the buildbots Aug 31 22:46:03 is curl still broken? Aug 31 22:46:09 I've not updated today Aug 31 22:46:17 how is it broken? Aug 31 22:46:39 rsalvaterra: I don't really know when it started, I've just recently (8 weeks ago) started using it 24/7 again - and the VDSL line is usually only used for my SIP landline phone numbers, until I can get them ported over to my fibre ISP - so it's usually barely seen any high usage Aug 31 22:46:57 mangix: https://pastebin.com/QiY0UWkv Aug 31 22:47:00 just noticed it when an excavator killed the ISPs POP for 1.5 days Aug 31 22:47:15 forcing me to switch over my network to VDSL again Aug 31 22:47:19 I last updated yesterday though, so it might be fixed by now Aug 31 22:47:24 Grommish: feeds install -a -f Aug 31 22:47:54 Yeah,i've removed the feeds directory completely, did a curl and libcurl uninstall and reinstall, etc etc.. i probably just need to update again Aug 31 22:48:06 you need -f Aug 31 22:48:10 Ok Aug 31 22:48:52 pkgadd: then it looks like a different problem, the wan reconnect issue is always reported to be visible in the logs Aug 31 22:50:00 I picked a random device to see if I can build it Aug 31 22:50:27 Indeed, it's quite weird… Also note that I'm using pepe2k's u-boot mod, overclocked to 680 MHz, and it's stable. Aug 31 22:58:09 mangix: fix the problem for mac80211: https://pastebin.com/EQNWx1zj mt76, ath10k-ct, mwlwifi and rtl8812au-ct still overwrite NOSTDINC_FLAGS and should still fail **** BEGIN LOGGING AT Mon Aug 31 23:04:12 2020 Aug 31 23:06:04 aparcar[m]: what about 10 years? Aug 31 23:06:28 mangix: that install -f -a worked.. thanks ;) Aug 31 23:06:38 I should probably do that every once in a while Aug 31 23:25:57 adrianschmutzler: this https://sl-h.de/tmp/tl-wdr4300-PPPoE-crash.log is the behaviour I'm seeing (with a current official master snapshot, just configured to use PPPoE and remote syslog) on the tl-wdr4300 Aug 31 23:35:38 pkgadd: Is the "OpenWrt netifd: Interface 'wan' is setting up now" caused by the problem or separate? Aug 31 23:36:18 because that's the symptom of the issue I mentioned beforehand Aug 31 23:36:45 e.g. https://bugs.openwrt.org/index.php?do=details&task_id=3226 Aug 31 23:36:54 adrianschmutzler: I think it's part of the problem, but the router is dead at this point, requiring powering off/ on Aug 31 23:37:35 unfortunately, I have almost no specific knowledge about PPPoE, since I've always had cable and never DSL Aug 31 23:37:51 that happened after ~900 MB traffic over PPPoE, it's 100% stable with plain ethernet/ dhcp Aug 31 23:38:37 you can try whether reverting those two commit helps for you as well Aug 31 23:38:52 e.g. mentioned at the bottom of the referenced bug report Aug 31 23:39:19 thanks, I'll have a try Aug 31 23:39:42 for the people reporting the wan reconnect issue, this seems to always helped Aug 31 23:40:26 so it should tell you whether you have a different flavor of the same problem (if reverting resolves it) or a different one (if reverting doesn't help) Aug 31 23:40:57 I've pinged 981213 about this some time ago already, but he said he couldn't reproduce it on his hardware Aug 31 23:42:56 I've just started a build and should know more in ~30 minutes Aug 31 23:43:22 So tell me tomorrow, I will hopefully sleep by then :-) Aug 31 23:43:37 thanks and gn8 Sep 01 00:26:05 Hauke: amazing. that fixes it **** ENDING LOGGING AT Tue Sep 01 02:59:57 2020