**** BEGIN LOGGING AT Tue Dec 10 02:59:58 2019 Dec 10 07:22:27 build #201 of omap/generic is complete: Failure [failed tools] Build details are at http://buildbot.openwrt.org/master/images/builders/omap%2Fgeneric/builds/201 blamelist: Daniel Golle Dec 10 07:33:09 hm, tools/Makefile:157: recipe for target 'tools/flex/compile' failed Dec 10 07:33:17 configure: error: cannot run /usr/bin/env bash build-aux/config.sub Dec 10 08:26:08 ynezz, https://gitlab.com/openwrt/docker/pipelines/101832084 failed Dec 10 08:26:53 I know Dec 10 08:27:09 someone has to fix it Dec 10 08:27:10 :) Dec 10 08:27:25 IIRC aparcar[m] is aware about it Dec 10 08:28:09 probably some "if 18.06.y branch; do something_differently_foo; fi" is needed Dec 10 08:28:27 master/19.07 dont have this issues Dec 10 08:34:45 is planned rootfs also for 18.06? Dec 10 08:38:30 IIRC it would probably need backporting of some patches, so probably nope Dec 10 08:39:01 for init? Dec 10 08:39:11 I don't know what is missing Dec 10 08:39:48 if 18.06 rootfs doesn't work, then probably something has to be missing, right? Dec 10 08:42:19 and missing support for Docker could be hardly considered as a fix, thus unlikely, that anyone would backport those patches into 18.06 and risk some regressions Dec 10 08:42:54 missing IB/SDK for 18.06.5 is issue and should be fixed Dec 10 08:44:10 s/SDK/SDK Docker images/ Dec 10 08:44:20 greearb_: The patches posted by Nicolas are for BA issues upon reconnection of the remote side Dec 10 12:08:48 hrm. luci status and firewall pages are giving me XHR timeouts on chrome, but work perfectly well in firefox. (on today's luci master) Dec 10 12:28:30 > karlp: try disabling http keepalives in uhttpd, see if it helps Dec 10 13:02:04 Google probably "optimized" HTTP agai Dec 10 13:18:23 is that http_keepalive set to 0? Dec 10 13:19:08 yep, that works. Dec 10 13:20:19 still very noticeably slower in chrome than ff. (loading the status->firewall page) Dec 10 13:20:29 but at least it now _works_ so that's great thanks. Dec 10 13:35:47 karlp: could you cpature me some PCAPs by any chance? Dec 10 13:36:07 I''ve seen random chatter about this issue in the forum from time to time but I was never able to reproduce it Dec 10 13:36:50 ideally once with keepalive enabled and showing some xhr timeouts and once with keepalive disabled Dec 10 13:37:24 maybe uhttpd does not properly fend off quic and http 2 and all that newfangled google crap Dec 10 13:37:54 sure, like wireshark pcaps? Dec 10 13:41:23 you want me to turn off tls on uhttpd first? Dec 10 13:41:55 karlp: hm yeah Dec 10 13:42:07 wouldn't surprise me if it works totally fine after turning off tls Dec 10 13:42:25 yes tcpdump/wireshark pcap Dec 10 13:44:45 I'm doing some testing to make sure, but I think that new "auth_cache 1" default may be causing my wireless clients to disassociate every 12 hours. Dec 10 13:49:03 well, I can't figure out how to actually turn off tls. I've turned off the listener, turned off redirect_https and commented out the certs, and now I just can't log in :) Dec 10 13:54:38 karlp: thats firefox, it refuses the login cookie if it knows that the site was https previously Dec 10 13:54:57 karlp: try a private/incognito tab or clear the site preferences (not sure waht the chrome equivalent is) Dec 10 14:00:45 ahh, both chromium/ff were doing it. ok. Dec 10 14:03:36 ok. so yes, chromium is "slow" on tls, but same (fast) as firefox on non-tls. Dec 10 14:05:19 which of these combinations would yuou like pcaps from then? keepalive on, xhr timeouts on chromium on non-tls? and keepalive on, no timeouts on firefox? Dec 10 14:07:22 can you reproduce the timeouts without tls on chromium? Dec 10 14:17:52 yes. I can. bit slow testing sorry, other things happening in the house. Dec 10 14:19:38 no. apparently I can't Dec 10 14:19:58 and apparently I can't tell between the two chromium (incog and regular) and two firefox windows well enough... Dec 10 14:22:33 ynezz: what do you think about moving the docker.git to gitlab and setup gitlab to mirror it to github. currently it is the other way around Dec 10 14:25:17 you're gonna love this trace! Dec 10 14:30:24 /o\ Dec 10 14:31:34 karlp: it is known that the performance with chrome and tls is absymal if the device happens to be slow Dec 10 14:31:52 karlp: chrome attempt to negotiate half a dozen tls sessions in parallel with each request Dec 10 14:32:14 we even reported it as bug back then but the response can be summarized as "fuck off, we're google" Dec 10 14:32:48 Hauke enabled tls session caching in mbedtls back then which improved the situation somewhat Dec 10 14:36:21 aparcar[m]: I'm not in favor of doing that unless the final decission is made (probably after meeting in Barcelona), as its impossible to anticipate the voting result... Dec 10 14:43:16 ynezz: ack Dec 10 14:45:17 ack Dec 10 15:06:55 ynezz: https://github.com/openwrt/docker/pull/42 Dec 10 15:07:45 once the usign key is added 18.06.x should build fine. People can still make use of the image builders and the rootfs can be used if desired, tho procd will cause some errors Dec 10 15:47:43 jow: please have a look at https://patchwork.ozlabs.org/patch/1194619/ and https://patchwork.ozlabs.org/patch/1193405/ Dec 10 15:47:59 the primer is trivial and the latter just a fixup of a previously approved commit Dec 10 15:50:36 i get this error trying to use the static glib2? I can only fix this by adding "TARGET_CFLAGS += -mno-interlink-compressed" for mips to the glib2 makefile? Dec 10 15:50:42 https://www.irccloud.com/pastebin/G0jJS1Lz/ Dec 10 15:51:08 i mean my adding "-minterlink-compressed" Dec 10 15:52:02 any idea why and if this a bug or i'm missing something? Dec 10 16:11:38 aparcar[m]: did you check that "ifdef CONFIG_INSTALL_LOCAL_KEY" is working as expected? Dec 10 16:12:27 wondering because the other conditional right before it used "ifeq ($(CONFIG_...),)" Dec 10 16:41:44 jow: okay let me check again I'll ping you then **** BEGIN LOGGING AT Tue Dec 10 17:30:37 2019 Dec 10 17:43:51 jow: http://palmtree.beeroclock.net/~karlp/bugs/luci-xhr-timeout-chromium-tls-firewall-status.pcapng.gz Dec 10 17:44:39 that's from a "quiet" start, opening a tab with the https://192.168.3.1/cgi-bin/luci/admin/status/iptables and it working, reloading, then going to status/overview and it failing and back again. Dec 10 17:45:09 it's not _always_ failing, but when it fails, it stays stuck for a qhile, and no other requests will respond either. Dec 10 17:45:43 I'm nto sure how much help it is, it's very spammy Dec 10 17:45:56 I can give you the uhttpd key file if it helps decode it at all? Dec 10 17:46:04 do you want a "same" one with keepalive off? Dec 10 17:48:50 http://palmtree.beeroclock.net/~karlp/bugs/luci-slow-chromium-tls-status-overview-no-keepalive.pcapng is with keepalive off, it works more reliably, but is definitely noticeably slower than firefox Dec 10 17:49:02 that's much simpler, just a single tab open that was slow, and stopped. Dec 10 20:09:24 "Intel released a statement regarding issues with certain CPUs when running in virtualized environments." Dec 10 20:10:01 maybe they should make that microcode opensource so someone can fix it finally Dec 10 20:17:08 They should... Though I bet heavily on not going to happend... for few reasons, code quality being one suspect:P Dec 10 20:22:12 Plundervolt, nice name Dec 10 20:54:49 indy: docker run --rm --tty --interactive openwrtorg/sdk:x86-64-18.06.5 bash -c "id && sudo id" Dec 10 20:55:14 aparcar[m]: thanks for fix! Dec 10 21:20:04 hi, i'm trying to build the package hostapd for openwrt v18.06.5, it seems to compile fine with "make package/hostapd/compile", but "make package/hostapd/install" fails with No rule to make target 'package/hostapd/install', what am i doing wrong? Dec 10 22:21:25 swalker: can you look at this? https://github.com/openwrt/packages/pull/10744#issuecomment-564273748 Dec 10 22:21:38 I'm not too familiar with how uscan looks at CVEs **** ENDING LOGGING AT Wed Dec 11 02:59:58 2019