**** BEGIN LOGGING AT Thu Dec 10 02:59:56 2020
Dec 10 08:34:20 <owrt-snap-builds> build #642 of apm821xx/nand is complete: Failure [failed targetupload]  Build details are at http://buildbot.openwrt.org/master/images/builders/apm821xx%2Fnand/builds/642  blamelist: Brian Norris <computersforpeace@gmail.com>, Hans Dedecker <dedeckeh@gmail.com>, John Audia <graysky@archlinux.us>
Dec 10 09:02:03 <aparcar[m]> mangix: https://github.com/openwrt/packages/pull/13512
Dec 10 09:02:07 <aparcar[m]> Comments?
Dec 10 09:29:12 <mangix> I don't have any TBH
Dec 10 09:30:53 <mangix> probably just merge
Dec 10 09:31:03 <aparcar[m]> yea that's what I thought
Dec 10 09:36:02 <rsalvaterra> Heh… speaking of which, I'm a bit miffed with the bloat of most Do{H,T} resolvers… and the bugginess of others (stubby/getdns). ¬_¬
Dec 10 09:42:27 <mangix> DNS resolvers?
Dec 10 09:43:24 <rsalvaterra> Well, stub resolvers.
Dec 10 09:43:37 <mangix> ah of course
Dec 10 09:43:42 <mangix> I'm stuck with knot
Dec 10 09:44:37 <rsalvaterra> Yeah, but you're recursing, right? That's great, but… I like the convenience of AdGuard DNS. :P
Dec 10 09:46:28 <rsalvaterra> I used stubby/getdns, which worked fine for a while, but at some point I had problems accessing certain sites (YouTube, from the top of my head).
Dec 10 09:47:03 <mangix> I currently have some issue with Knot and WireGuard where it's not passing DNS
Dec 10 09:48:13 <rsalvaterra> That's strange… I don't remember having issues with DNS over WireGuard.
Dec 10 09:48:43 <mangix> yeah I have no idea
Dec 10 09:48:54 <mangix> dnsmasq works fine
Dec 10 09:49:35 <rsalvaterra> But you're recursing through the tunnel? Or is knot on the other side of the tunnel?
Dec 10 09:50:01 <jow> rsalvaterra: it's mystery to me why all those Do{H,T} resolvers are such an instable bloat fest
Dec 10 09:50:21 <jow> I don't even dare to think how the internet would look like today if core software would've been written like that back in the day
Dec 10 09:50:34 <rsalvaterra> jow: I know, right?! :P
Dec 10 09:50:57 <rsalvaterra> Man… Go? Really? Why?! *facepalm*
Dec 10 09:51:08 <jow> either the protocol is too complex to be implemented in a secure and lean manner, or the people implementing those packages are not competent
Dec 10 09:51:12 <jow> I doubt the latter
Dec 10 09:51:38 <rsalvaterra> Well… getdns messages are… JSON objects.
Dec 10 09:52:07 <jow> plug together libcurl, libjson-c, libcares
Dec 10 09:52:13 <jow> shouldn't be that hard
Dec 10 09:52:29 <jow> throw in libevent if manual poll/epoll is too hard/unportable/ugly
Dec 10 09:52:30 <mangix> DNS is indeed complex
Dec 10 09:53:13 <rsalvaterra> jow: If you already have libevent for other stuff, using it isn't such a bad idea, for performance reasons too.
Dec 10 09:53:49 <rsalvaterra> mangix: DNS is complex, but a stub resolver shouldn't be. The complexity is on the server side. :)
Dec 10 09:57:35 <mangix> which contradicts your earlier bloat statement :)
Dec 10 09:58:37 <rsalvaterra> mangix: I'd say it corroborates. The stubs should be simple, but they're bloated beyond belief. :P
Dec 10 09:59:37 <mangix> maybe we need odnsd to replace dnsmasq
Dec 10 10:00:36 <rsalvaterra> OMG, another one?
Dec 10 10:00:56 <mangix> https://xkcd.com/927/
Dec 10 10:01:21 <owrt-snap-builds> build #671 of apm821xx/sata is complete: Failure [failed targetprune]  Build details are at http://buildbot.openwrt.org/master/images/builders/apm821xx%2Fsata/builds/671  blamelist: Brian Norris <computersforpeace@gmail.com>, Hans Dedecker <dedeckeh@gmail.com>, John Audia <graysky@archlinux.us>
Dec 10 10:01:38 <Grommish> That sounds like a challenge :D
Dec 10 10:01:45 <rsalvaterra> It's a miracle this industry came this far… :P
Dec 10 10:02:10 <rsalvaterra> If people knew how software is written…
Dec 10 10:04:39 <rsalvaterra> Also, is there anything odhcpd does which dnsmasq doesn't (or vice-versa)?
Dec 10 10:04:40 <mangix> I just pushed 7 totally untested commits. I expect buildbot failures soon
Dec 10 10:04:55 <mangix> rsalvaterra: no stub resolver
Dec 10 10:05:04 <rsalvaterra> mangix: #yolopush :P
Dec 10 10:05:29 <Grommish> mangix: :D
Dec 10 10:05:58 <rsalvaterra> So… why use odhcpd, then? ;)
Dec 10 10:06:12 <rsalvaterra> Since dnsmasq does everything and it's there by default…
Dec 10 10:06:13 <mangix> https://www.quickmeme.com/meme/2gs6
Dec 10 10:06:26 <mangix> whoops
Dec 10 10:06:31 <mangix> http://www.quickmeme.com/meme/2gs6
Dec 10 10:06:38 <rsalvaterra> Firefox can’t establish a connection to the server at www.quickmeme.com.
Dec 10 10:07:21 <rsalvaterra> mangix: Like I said, if people knew how software is written… :P
Dec 10 10:07:25 <Grommish> Just to complicate the issue.. you know,, you could always look into Trust-DNS
Dec 10 10:07:37 <Grommish> :D
Dec 10 10:07:49 <mangix> rsalvaterra: odhcpd is used because... NIH
Dec 10 10:08:13 <mangix> It also has actual ubus integration
Dec 10 10:08:24 <jow> mangix: was about to say s/NIH/ubus integration/
Dec 10 10:08:29 <rsalvaterra> mangix: I see… just like odhcp6c… XD
Dec 10 10:08:56 <jow> we evaluated the feasibility of using dnsmasq multiple times, but it simply does not integrate well since it insists on managing the entire delegation chain itself, without any instrumentation capabilities
Dec 10 10:09:27 <jow> at the time odhcp6c was written, the only choices for DHCPv6-PD were dibbler and ISC dhcpd
Dec 10 10:09:29 <rsalvaterra> (Not really, I don't think the BusyBox DHCPv6 client existed when odhcp6c was written.)
Dec 10 10:09:32 <jow> both not exactly lightweight
Dec 10 10:09:56 <jow> udhcpc was feature incomplete
Dec 10 10:11:04 <rsalvaterra> Right, I went through the history, odhcp6c was written a *long* time ago. :)
Dec 10 10:11:25 <jow> the single most missing feature that prevents replacing dnsmasq + odhcpd with just dnsmasq-dhcpv6 is the ability to add and remove IPv6 delegation prefixes at runtime without config rewrite + service restart
Dec 10 10:12:12 <jow> the ubus integration would be neglectible if such a capability exists since we could do that in whatever service takes care of driving dnsmasq (likely netifd then)
Dec 10 10:13:21 <jow> when looking into the feasibility of adding such functionality, the dnsmasq code was deemed to be too messy to work with
Dec 10 10:13:41 <jow> (didn't extensivley check it myself but from a cursory look it *was* complicated at least)
Dec 10 10:13:50 <rsalvaterra> I don't have enough experience with IPv6 to understand the importance of being able to change delegation prefixes without restarting the service, so I'll just take your word for it. :P
Dec 10 10:13:52 <mangix> as an aside to that, odhcpd can replace dnsmasq. there's a UCI config option for it
Dec 10 10:14:05 * Borromini is using odhcdp + unbound for dhcp/dns
Dec 10 10:14:17 <Borromini> s/odhcdp/odhcpd/
Dec 10 10:15:26 <jow> mangix: is (1)/usr/bin/pcre-config
Dec 10 10:15:29 <jow>  a script?
Dec 10 10:15:34 <mangix> yes
Dec 10 10:15:53 <jow> actually stupid question, you likely wouldn't be able to sed around in it if it were an ELF executable
Dec 10 10:16:12 <mangix> I think it's for packages that predate pkgconfig
Dec 10 10:16:21 <jow> yeah, a horrible concept
Dec 10 10:16:31 <mangix> unfortunately, they're still around
Dec 10 10:16:51 <jow> imagine the 200 or so libs on your system would all ship a $libname-config executable
Dec 10 10:16:58 <jow> conveniently put in your standard $PATH
Dec 10 10:17:22 <jow> and omitting the "lib" prefix
Dec 10 10:18:41 <jow> welcome ssl-config and crypto-config printing random gcc flags
Dec 10 10:18:53 <mangix> I recently went through all the ones in the packages feed. Fixed all except maybe 1 or 2.
Dec 10 10:19:15 <mangix> I was shocked to see how many of them there are
Dec 10 10:19:30 <owrt-snap-builds> build #655 of mediatek/mt7622 is complete: Failure [failed targetupload]  Build details are at http://buildbot.openwrt.org/master/images/builders/mediatek%2Fmt7622/builds/655  blamelist: Brian Norris <computersforpeace@gmail.com>, Hans Dedecker <dedeckeh@gmail.com>, John Audia <graysky@archlinux.us>
Dec 10 10:19:48 <jow> also who came up with "-config" as suffix. Something like "-buildflags" would've made more sense
Dec 10 10:20:27 <jow> but then it's from a time where "a.out" was deemed to be a sensible default output file name
Dec 10 10:22:08 <jow> anyhow, back on topic. mangix I've been wondering for a while if these absolute pathname sed fixes can be somehow generalized into a PKG_FIXUP macro
Dec 10 10:23:50 <jow> could be implemented using something like  $(FIND) $(PKG_INSTALL_DIR) -path '*/usr/bin/*-config' -or -path '*/pkgconfig/*.pc' -print0 | xargs -0 -r $(SED) 's/.../.../g'
Dec 10 10:25:20 <mangix> I have not investigated.
Dec 10 11:11:01 <russell--> well, that was embarrassingly easy, ubnt-bullet-ac works, patch inbound
Dec 10 12:37:15 <owrt-snap-builds> build #665 of ramips/rt288x is complete: Failure [failed targetupload]  Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Frt288x/builds/665  blamelist: Brian Norris <computersforpeace@gmail.com>, Hans Dedecker <dedeckeh@gmail.com>, Karel Ko?? <karel.koci@nic.cz>, Paul Spooren <mail@aparcar.org>
Dec 10 13:17:54 <owrt-1907-builds> build #235 of lantiq/ase is complete: Exception [exception sharedwd df ccachestat]  Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/lantiq%2Fase/builds/235  blamelist: Davide Fioravanti <pantanastyle@gmail.com>
Dec 10 13:55:54 <owrt-snap-builds> build #589 of mpc85xx/p2020 is complete: Failure [failed kmodupload]  Build details are at http://buildbot.openwrt.org/master/images/builders/mpc85xx%2Fp2020/builds/589  blamelist: Brian Norris <computersforpeace@gmail.com>, Hans Dedecker <dedeckeh@gmail.com>, Karel Ko?? <karel.koci@nic.cz>, Paul Spooren <mail@aparcar.org>
Dec 10 14:38:02 <owrt-1907-builds> build #226 of at91/sam9x is complete: Failure [failed targetupload]  Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/at91%2Fsam9x/builds/226  blamelist: Davide Fioravanti <pantanastyle@gmail.com>
Dec 10 15:17:09 <rr123> full dnsmasq(dns+dhcp) worked well here, why need odhcpd? never used that actually
Dec 10 15:18:11 <rsalvaterra> rr123: Indeed. I don't find the limitation jow mentioned that cripling…
Dec 10 15:19:16 <rr123> dnsmasq uses hash to store /etc/hosts, for ads blocking when you have 100K lines it becomes an issue, memory and speed, but still OK, when you have 20K lines of address=/domain.com/1.2.3.4 in dnsmasq.conf it becomes a serious problem, the restart time of dnsmasq becomes a true bottleneck
Dec 10 15:19:50 <rr123> i put 100k lines of address there just to compare with 100k lines in /etc/hosts, the former will take 30 minutes to restart and eats up all the cpu
Dec 10 15:25:33 <rr123> jow: do you think the new luci will be ready for next major release, I read the new code Rafel added for uhttpd/sse it looks very interesting, in time for my websocket study, will try sse instead for now
Dec 10 15:25:52 <owrt-1907-builds> build #227 of x86/64 is complete: Failure [failed kmodupload]  Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/x86%2F64/builds/227  blamelist: Davide Fioravanti <pantanastyle@gmail.com>
Dec 10 16:18:44 <owrt-1907-builds> build #229 of tegra/generic is complete: Failure [failed targetupload]  Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/tegra%2Fgeneric/builds/229  blamelist: Davide Fioravanti <pantanastyle@gmail.com>
Dec 10 17:52:14 <aparcar[m]> why are these failing?
Dec 10 17:55:58 <adrianschmutzler> rsync error: timeout in data send/receive (code 30) at io.c(204)
Dec 10 17:56:09 <adrianschmutzler> upload@mirror-01.infra.openwrt.org
Dec 10 18:02:19 <rr123> disk full?
Dec 10 20:24:40 <owrt-snap-builds> build #643 of apm821xx/nand is complete: Success [build successful]  Build details are at http://buildbot.openwrt.org/master/images/builders/apm821xx%2Fnand/builds/643
Dec 10 21:28:51 <philipp64> karlp and jow: questions about procd… I’m trying to do DNS integration of dhcp into bind for dynamic updates.  (1) what’s the risk associated in moving named from 50 to 18 (dhcpd runs as 19) and (2) how do I make sure named finishes starting before dhcpd gets started?
Dec 10 21:31:32 <owrt-snap-builds> build #672 of apm821xx/sata is complete: Success [build successful]  Build details are at http://buildbot.openwrt.org/master/images/builders/apm821xx%2Fsata/builds/672
Dec 10 21:33:32 <karlp> I don't hink you can in general.
Dec 10 21:33:54 <karlp> this is where people start writing ubus events and how you end up with upstart/systemd.
Dec 10 21:34:20 <karlp> I think there's a few other scripts that do a ubus subscribe or some other wait
Dec 10 21:34:31 <karlp> I've seen it in patches at least, can't think of one off the top ofmy head.
Dec 10 21:35:56 <philipp64> or a sleep… sigh.
Dec 10 21:44:34 <philipp64> other than that, can named start as 18?
Dec 10 21:45:17 <karlp> no opinion there sorry :)
Dec 10 21:45:42 <philipp64> also… firewall starts at 19 (same as dhcpd)… and network starts at 20…. why not have dhcpd start *after* network?
Dec 10 21:45:53 <philipp64> jow: can you chime in?
Dec 10 21:46:29 <Piraty> yocto's opkg maintainer steps down
Dec 10 21:46:44 <philipp64> how do we come up with these numbers, anyway?  seems arbitrary...
Dec 10 21:48:11 <karlp> seems fairly arbitrary yes.
Dec 10 21:52:49 <aparcar[m]> blocktrron: can we close https://github.com/openwrt/openwrt/pull/3654?
Dec 10 21:54:36 <ynezz> aparcar[m]: s/close/merge/
Dec 10 21:55:14 <aparcar[m]> ynezz: it's merged already and was reopened for further discussion
Dec 10 21:55:35 <ynezz> ah, ok, wasn't aware about that
Dec 10 21:55:49 <philipp64> okay, different question… seeing this:
Dec 10 21:55:51 <philipp64> Dec 10 14:52:38 OpenWrt kernel: [269800.621617] dhcpd[13275]: segfault at 3 ip 00007fb852fc7a4b sp 00007fff135ad020 error 4 in libc.so[7fb852fb9000+49000]
Dec 10 21:56:01 <philipp64> how to go about converting that into a stack trace?
Dec 10 21:59:13 <rr123> some init scripts are time sensitive and it's hard to define their dependencies
Dec 10 22:00:59 <rr123> so you may see a script starting early(e.g. dnsmasq) before its dependency is ready(e.g. network)
Dec 10 22:07:34 <owrt-snap-builds> build #656 of mediatek/mt7622 is complete: Success [build successful]  Build details are at http://buildbot.openwrt.org/master/images/builders/mediatek%2Fmt7622/builds/656
Dec 10 22:14:55 <nick[m]1> someone here, who had already issues when in some c code a realloc is done, that the programms fails with out of memory?
Dec 10 22:16:07 <dangole> ynezz: testing your heroic ustream-ssl and uclient efforts, works great, actually verifies certificates, doesn't seem to leak memory!
Dec 10 22:16:52 <dangole> yness: tested with wolfssl (client and server), mbedtls is about to go next
Dec 10 22:22:09 <rr123> dangole: curl is heavy, busybox wget can do http and https, what does uclient bring?
Dec 10 22:24:35 <karlp> presumably when it was done, it allowed ssl with mbed-tls or something that busybox wget didn't support
Dec 10 22:25:17 <karlp> uclient is however "thin" on top of ustream, which is used in other places, that just to provide a wget command line equiv
Dec 10 22:25:26 <rr123> busybox wget ssl has two options now: interla tls, and openssl
Dec 10 22:25:45 <rr123> s/interla/internal/
Dec 10 22:25:57 <karlp> ok, so we don't want to have internal tls, as that's _yet another_ one.
Dec 10 22:26:05 <karlp> and people that don't have space for openssl....
Dec 10 22:28:29 <rr123> i decide to marry openssl until all packages can run with non-openssl, in my case, freeradius mandates openssl
Dec 10 22:29:07 <karlp> that's fine for some of us, but not all people can/will use openssl.
Dec 10 22:29:25 <rr123> agree
Dec 10 22:31:33 <blocktrron> aparcar[m]: as me made the decision go ahead
Dec 10 22:34:43 <jow> philipp64: firewall starts before network to establish base policies before setting up wan connectivity
Dec 10 22:34:58 <jow> it is reloaded on subsequent ifup's anyway
Dec 10 22:35:52 <philipp64> okay.
Dec 10 22:36:33 <philipp64> Integrating rndc hooks into dhcpd is causing rdnc to coredump.  How do I troubleshoot this?
Dec 10 22:36:45 <jow> aparcar[m]: presumably the buildbot rsyncs are failing because the OSUOSL lab uplink in oregon is saturated atm
Dec 10 22:36:58 <jow> philipp64: gdb(server)
Dec 10 22:37:42 <jow> philipp64: or scp the coredump to your buildroot, run ./scripts/remote-gdb /path/to/dump.core build_dir/target-*/rdnc-*/rdnc (or whatever the executable path in build_dir is)
Dec 10 22:38:27 <jow> the latter is somewhat easer and should provide full symbol resolution without mucking too much with the target system
Dec 10 22:38:48 <philipp64> uh… where would the coredump be?
Dec 10 22:39:17 <philipp64> find / -xdev -name “core*”
Dec 10 22:39:20 <philipp64> turns up nothing.
Dec 10 22:39:46 <jow> n openwrt, coredumps are usually placed in /tmp
Dec 10 22:40:03 <jow> and named $processname.$pid.$time.core or similar
Dec 10 22:42:07 <philipp64> do I need to enable any config to permit coredumps?
Dec 10 22:42:36 <jow> ulimit -c unlimited, then make the process crash in the same shell
Dec 10 22:43:29 <jow> if started through procd, try adding  procd_set_param limits core=unlimited  to the init script
Dec 10 22:43:56 <jow> somehwere next to procd_set_param command ...
Dec 10 22:45:29 <philipp64> Dec 10 15:44:57 OpenWrt procd: Not starting instance dhcpd::limits, command not set
Dec 10 22:46:22 <philipp64> oh, inside the procd_open_instance ...
Dec 10 22:50:58 <dangole> ynezz: `wget -O- --no-check-certificate https://localhost` still fails somehow, says Connection error: Connection failed
Dec 10 22:51:17 <philipp64> got it… http://paste.ubuntu.com/p/jDrH9tsKyx/
Dec 10 22:51:23 <dangole> without the --no-check-certificate flag it correctly reports Connection error: Invalid SSL certificate
Dec 10 22:54:56 <philipp64> catd is -1????
Dec 10 23:01:45 <mangix> jow: ping
Dec 10 23:36:14 <philipp64> jow: building isc-dhcp with DISABLE_NLS:=1 … any reason we’d need NLS?
Dec 11 00:54:48 <pkgadd> rsalvaterra: there are still ISPs with dynamic IPv4 addresses and dynamic IPv6 prefixes, both combined with daily disconnects to discourage server uses, for those odhcpd's runtime reconfiguration/ prefix update is rather essential - dnsmasq is lacking in this regard
Dec 11 00:56:18 <pkgadd> rsalvaterra: dnsmasq is all fine if you're dealing with static IPv6 prefixes, but beware if it changes
Dec 11 00:58:49 <rr123> pkgadd: you mean when ISP's dhcp prefix changed, dnsmasq won't be able to detect/react so ipv6 from the house won't work at all afterwards
Dec 11 00:59:20 <rr123> for my use case I disabled all ipv6 on openwrt and nothing seem broken
Dec 11 01:00:34 <Dagger> we (as in the networking community) put in a huge amount of effort to make sure v6 could be deployed without disruption, and to have that summarized as "I disabled it and nothing seemed to break" is really unfortunate :(
Dec 11 01:01:48 <rr123> that's not my intention, I wanted to play with ipv6 when I have more confidence, thus the above question, sorry for the tone
Dec 11 01:02:17 <rr123> one reason is the ad block, with ipv6 it's harder to do i feel
Dec 11 01:02:25 <pkgadd> rr123: more or less that, yes
Dec 11 01:03:26 <rr123> in that case I can have a pingcheck for ipv6, if it failed i restart dnsmasq, maybe that's a dnsmasq workaround, ugly? yes
Dec 11 01:03:50 <Dagger> ad blocking should be no harder on v6 than on v4, although identifying client machines (for exclusion rules) might be harder due to privacy addresses
Dec 11 01:04:28 <pkgadd> that's not just ugly, it's borderline DoS'ing the poor peer you're pinging
**** ENDING LOGGING AT Fri Dec 11 02:59:56 2020