**** BEGIN LOGGING AT Thu Nov 14 02:59:58 2019 Nov 14 07:38:09 aparcar[m]: ping hi when will be 18.06.5 container images on docker.io/ :) Nov 14 07:45:07 indy__: I can add it but it's missing some procd features used by docker Nov 14 07:49:13 aparcar[m]: i'm using podman not docker and works fine, at least sdk and imagebuilder Nov 14 07:51:07 hmm, it appears that sysinfo(2) is broken on mips Nov 14 07:51:13 wigyori: typo after Notes: Nov 14 07:51:22 meh Nov 14 07:51:50 :) Nov 14 07:52:34 thanks! Nov 14 08:04:23 indy__: see here https://gitlab.com/openwrt/docker/pipelines/95845213 Nov 14 08:04:25 this should upload 18.06.5 Nov 14 08:04:40 jow: please let me know if you got some ansible scripts for me Nov 14 08:29:40 indy__: oh well, I just realized that the 18.06 release keys aren't in the docker build container. Can I motivate you to do some tests on 19.07-rc1 instead? Nov 14 08:29:49 indy__: if it's of importance I modify the container Nov 14 08:33:18 aparcar[m]: you mean 19.07-SNAPSHOT? Nov 14 08:43:55 indy__: I can add rc1 if you want Nov 14 09:05:53 indy__: anyway here is the rc1 for now, should take a bit more time until it's on docker.io. if you really need 18.06 for whatever reason i'll set it up tomorrow https://gitlab.com/openwrt/docker/pipelines/95859330 Nov 14 10:01:52 thanks for all the work on luci.js - have to say to me the web ui seems much more responsive Nov 14 10:07:26 someone needs to fix KGB-2 Nov 14 10:09:04 kgb is always enthusiastic about things Nov 14 10:45:04 or upgrade it to FSB Nov 14 10:51:35 so, change in name only? :) Nov 14 10:55:38 ldir: great to hear Nov 14 10:55:53 was wondering if it makes any difference after all Nov 14 10:56:41 yes, now it has spinning wheels while data loads :-) Nov 14 10:56:57 I'm sure it helps with the responsive feeling Nov 14 11:24:08 jow: I've sent you a couple of language tweaks - the first one has always bothered me - but feel free to ignore Nov 14 11:49:51 Hi any one know how to fix a prob in luci. Nov 14 11:49:54 Error Nov 14 11:49:54 XHR request timed out Nov 14 11:50:10 I am on r11487 Nov 14 11:50:32 Should i just do a new build? Nov 14 11:51:30 Tapper: on which page does it happen? Nov 14 11:52:00 the Overview page Nov 14 11:52:15 and When I click on SQM it says bad gateway Nov 14 11:53:15 Let me update my build and I will let you know how it gos Nov 14 11:53:20 ok Nov 14 11:53:31 if it does not help you could give me ssh and I'll take a look Nov 14 11:54:06 OK Nov 14 12:08:36 jow seems to be OK now Nov 14 12:09:01 I don't know what was going on but I will keep a eye out for any bugs. Nov 14 12:11:20 thank you Nov 14 12:25:35 someone remind again why
and
is better than and ? Nov 14 12:32:45 Is there a best practice about how to apply uboot-envtools changes when adding device support? Separate commit for uboot-envtools or include it in the device support PR? Loobs like both has been done depending on who committed it ... Nov 14 12:37:36 karlp: it can wrap on narrows screens Nov 14 12:40:10 I just tried "doing the right thing" and using table-title and even data-titles and I just get unaligned garbage, so I'm just going to stick with tables. there's only so much modern I can handle :| Nov 14 12:41:57 I feel we must have done something wrong when tabula data can't be tables because of layout, Nov 14 12:42:19 not using tables for layout, fine, but unable to use tables for tables and expecting the browser to layitout for you? bah Nov 14 12:42:33 anywya, bit too OT Nov 14 12:48:05 The GCC 7.5 release brings more than 215 bug fixes https://www.phoronix.com/scan.php?page=news_item&px=GCC-7.5-Released Nov 14 12:48:26 O we switched to gcc 8 I forgot! Nov 14 12:56:29 karlp: you can keep using plain tables if that suits your use case Nov 14 13:21:02 jow: yeah, I mean, I can't get the browsers to show me tables on a mobile, so I'll have to do something eventually, but I can't even work out how to so far, so motivation isn't real high. Nov 14 13:21:20 I've managed to drop a few things I had with custom css and hacks though, so progress I guess. Nov 14 13:26:54 I wonder who invented the "iface6rd" uci options Nov 14 13:27:17 its completely unintuitive Nov 14 13:27:24 0 - disable 6rd Nov 14 13:27:30 any other value - enable 6rd Nov 14 13:27:54 but if it is 1 or empty, the resulting interface will be named ${parentname}_6 Nov 14 13:28:05 otherwise it will be called as specified in that option Nov 14 13:28:15 don't even know how to start descvribing that in the ui Nov 14 13:28:35 and why didn't it simply reuse option ipv6 which already is defined for pppoe Nov 14 13:28:38 sigh Nov 14 13:30:15 jow: should is switch to libpcre2 from libpcre or should i stick with pcre? Nov 14 13:31:07 does one of them ship by default with openwrt? Nov 14 13:31:18 gladiac1337: uhm, I don't know Nov 14 13:31:55 https://github.com/openwrt/packages/blob/master/libs/pcre/Makefile Nov 14 13:32:11 don't know if 8.43 is "pcre" or "pcre2" Nov 14 13:32:28 8.43 is pcre(1) Nov 14 13:33:26 also - pcre has a jit compiler which should work on arm and mips but it is disabled Nov 14 13:33:52 https://github.com/openwrt/packages/blob/master/libs/pcre2/Makefile Nov 14 13:33:56 seems that is there, too Nov 14 13:34:06 so use whatever floats your boat. Base images ship with neither Nov 14 13:35:25 alright, thx Nov 14 13:36:05 i will try and add jit-support - should boost regex performance significantly Nov 14 13:49:50 adrianschmutzler: i think in the cases where the uboot-envtools change was separate, they might simply not have thought of making the change at the same time. since both styles have been accepted, i assumed both styles to be acceptable Nov 14 14:23:50 is someone already working on adding 5.4 support? Otherwise I would commence now (Hauke?) Nov 14 15:09:21 are all QorIQ CPU's arm based or ppc? Nov 14 15:10:37 both Nov 14 15:12:27 don'ðt reallllly know why layerscape is also called qoriq, but whatevs Nov 14 15:24:05 jow: any tips on debugging ubus acls? I had an acl I added a while ago as per https://openwrt.org/docs/techref/ubus?s[]=acl#acls (superuser style, anyone logged in can do anything they could do via ssh) Nov 14 15:24:22 and now in 1907 I get access denied, doing a file exec Nov 14 15:26:03 luci_access.write.ubus.file still has exec in it though? in luci-base.json Nov 14 15:26:07 what's blocking me? Nov 14 15:26:50 or, how does read/write separation work? because luciÖaccess.read.ubus.file doesn't list exec, so I presume that's where I've gone wrong? Nov 14 15:26:57 is that just GET vs POST? or... Nov 14 15:27:31 I don't really want to add a rpcd plugin just to do "ntpdate -q localhost" Nov 14 16:04:23 jow: I can sort of go one better than that - interfaces - dhcp(v4) release '0' "Don't release address on shutdown" - dhcpv6 norelease '1' "Don't release address on shutdown" Nov 14 16:16:52 if someone were to make it consistent.... which way should it be? I prefer 'release = 0' = don't release Nov 14 16:17:35 adrianschmutzler: Re: https://patchwork.ozlabs.org/patch/1169555/ pls do whatever you think is appropriate Nov 14 16:18:27 ynezz: okay, then I will just change the SUPPORTED_DEVICES for ath79 Nov 14 16:18:40 ok, thanks! Nov 14 16:18:46 1. this is the same I already did for the tp-link cases Nov 14 16:19:09 2. ar71xx is (unfortunately) designed this way, so that one board name is for several devices Nov 14 16:19:30 3. much less change to break anything in ar71xx Nov 14 16:20:06 I've just marked the patch as rejected Nov 14 16:21:22 k. The only disadvantage is that technically one could now flash with the wrong firmware for the same board (i.e. flash wdr4300 with image for wdr3600) Nov 14 16:22:18 However, technically this has been always possible with ar71xx ... Nov 14 16:22:55 karlp: you need to both allow "ubus": { "file": [ "exec" ] } and "file": { "/usr/bin/ntpdate": [ "exec" ] } Nov 14 16:23:07 adrianschmutzler: I just dont want to spend more time on this :) Nov 14 16:23:16 karlp: the former allows access to the exec method, the latter allows the exec method to invoke /usr/bin/ntpdate Nov 14 16:23:46 ynezz: And that's exactly why I think exploiting SUPPORTED_DEVICES is the better thing to do Nov 14 16:23:48 karlp: alternatively a wildcard (*) instead of /usr/bin/ntpdate if you want to allow exec on all paths Nov 14 16:26:30 ldir: I prefer release 0 as well as double inversion feels unelegant. But then there's another popular opinion that you never should introduce negative options (meaning options that need to be set to a negative value to have an effect) Nov 14 16:28:42 disabled=0 Nov 14 16:29:45 I *think* the default for dhcpv4 case is release 1 Nov 14 16:32:50 jow: but my superuser acl file is still in place, is it being overridden by the luci-base.json file as being more specific? Nov 14 16:33:10 karlp: no Nov 14 16:33:22 all ACL definitions are merged Nov 14 16:33:37 you can see the effective outcome using "ubus call session list" once logged in Nov 14 16:33:42 none of the acl stuff has materially _changed_ has it? you're just actually using more of them? Nov 14 16:33:54 well there was one change in the file plugin Nov 14 16:33:58 ynezz: FYI: https://git.openwrt.org/?p=openwrt/staging/adrian.git;a=commitdiff;h=3f12d2b1ab7f88cadfe6b089bf88b5ee98c82ae1 Nov 14 16:34:21 what determines what access group I end up in? Nov 14 16:34:24 it requires an additional file: { ... } scope now and it recognizes two new permissions "list" and "exec" Nov 14 16:34:29 I suppose you do not want to be mentioned as co-developer/suggestor for such a simple patch? Nov 14 16:34:39 karlp: the login definition of /etc/config/rpcd Nov 14 16:34:54 karlp: the default one binds all groups (*) Nov 14 16:35:25 ok, so the "superuser" acl that is on the wiki, if you actually want to use "file" like root on the shell, now need sthe extra file{} scope? Nov 14 16:36:06 yes, to be "superuser" it needs an additional: "file": { "*": [ "list", "read", "write", "exec" ] } Nov 14 16:36:23 is all the acl stuff in luci intended so that you can have levels of users in luci, unlike what we have now? Nov 14 16:36:29 yes Nov 14 16:37:21 I am preparing to switch controller index() functions to static declarations in JSON format, these can then specific access-group (acl), uci and file dependencies Nov 14 16:37:54 you can then tell "the route /admin/foo/bar" requires the "luci-base" access group, should only render if /etc/config/blah exists and /sbin/blah is installed Nov 14 16:38:09 * karlp had just started becoming somewhat comfortable with how luci worked :) Nov 14 16:38:20 though I'm sure this new stuff is more approachable for everyone else these days. Nov 14 16:38:56 the additional file: scope is at the top level, read, write, file? Nov 14 16:39:02 I want to get red of the dynamic route building code, thats my main motivation Nov 14 16:39:49 second Nov 14 16:41:03 karlp: https://pastebin.com/ZngjUKJB Nov 14 16:41:06 this should suffice Nov 14 16:41:32 gotcha, that was where I was trying it now, one each under read/write, not at the otp level. Nov 14 16:41:36 * karlp restarts rpcd Nov 14 16:45:40 yes! Nov 14 16:45:43 * karlp updates the wiki Nov 14 16:46:38 here's a more fine grained example: https://github.com/openwrt/luci/blob/master/modules/luci-base/root/usr/share/rpcd/acl.d/luci-base.json Nov 14 16:47:13 yeah, the luci-base.json has heaps of stuff in it, but as long as I only have a single web user, I don't see any gain from trying to put all the acls on. (for me at least) Nov 14 16:47:36 thats true. I am personally just wary of file/exec Nov 14 16:48:20 since it basically allows any shell code while with the old Lua stuff you only could invoke things predefined in code (unless there was a code injection bug or similar) Nov 14 16:49:12 totally, I used to use a separate controller to do the ntpdate query, but it... surprise surprised, blocked the webserver a lot, Nov 14 16:49:13 but thats just a technicality I guess. If you have write access to /etc7rc.local or /etc/firewall.user and to /sbin/reboot, you can execute arbitrary code anyway Nov 14 16:49:27 and if you're logging into the web ui as root, you can alwaysjust ssh Nov 14 16:50:10 so, I've converted tabs to use standard tabs, converted some tables, got a ubus exec again. now just need to fix some dynamic tabs and we're nearly good to go on 1907 luci new hotness :) Nov 14 16:54:39 with a bit of luck, the next openwrt will come with multiuser luci Nov 14 16:54:46 the one after 19.07 that is Nov 14 16:55:49 bbl, off to some dumb pc gaming Nov 14 17:04:32 * Slimey goes to build npx layerscape targets Nov 14 17:12:47 those are arm Nov 14 17:25:33 DonkeyHotei i think the QorIQ cpu's are both some how.. its confusing Nov 14 18:19:29 Slimey: on what HW ? Nov 14 18:19:41 Slimey: i have been told its ultra fast on the networking part Nov 14 18:33:48 * Monkeh pokes blogic Nov 14 18:37:48 you're brave! :-) Nov 14 18:38:23 I have a long and well insulated stick Nov 14 18:38:47 wise man Nov 14 18:39:05 lol Nov 14 18:46:04 blogic: are you aware of any SPI bus related issues in MT7629 similar to those in MT7621 and MT76x8? Nov 14 18:51:53 xback: no I haven't looked into kernel 5.4 support yet Nov 14 18:54:01 xback: I do not know if we had a decison for the kernel for the first 20.X Nov 14 19:00:26 Hauke: AFAIK, we didn't Nov 14 19:03:02 pepe2k: nope Nov 14 19:03:15 pepe2k: i'll ask, i got patches from mtk today for 7629 Nov 14 19:03:29 Monkeh: i'll poke back in a couple of days Nov 14 19:03:43 blogic: ok, thanks, I think the IP is different Nov 14 19:06:19 DonkeyHotei yeah they are ppc, like i said confusing documentation Nov 14 19:06:49 ppc and arm are not in the same chip Nov 14 19:09:49 depending on the soc QorIQ has both in its product family Nov 14 19:10:27 wikichip got me straighten out Nov 14 19:13:11 wikichip... thanks! Nov 14 20:21:23 jow: git clone git://git.openwrt.org/project/libnl-tiny.git fails with fatal: remote error: access denied or repository not exported: /project/libnl-tiny.git Nov 14 20:21:32 jow: https://gitlab.com/ynezz/openwrt-ci/-/jobs/351258741#L2483 Nov 14 20:22:10 same here localy Nov 14 20:30:56 ynezz: any reason to not use https:// ? Nov 14 20:32:07 sure, I can use https:// just saying, that it was working before, but suddenly stopped working Nov 14 20:32:25 for example libubox works, urngd doesnt Nov 14 20:32:34 should both work now Nov 14 20:32:38 thanks! Nov 14 20:32:52 the git server uses a different directory tree composed of symlinks Nov 14 20:33:11 to allow for the legacy aliases to still work. I forgot to update that symlink tree with the latest repos Nov 14 20:33:34 ok, anyway https:// is prefered? Nov 14 20:33:50 I have no strong preference Nov 14 20:33:59 I've read somewhere that git:// is more efficient Nov 14 20:34:30 while https:// is most likely to work with all kinds of networks and firewalls Nov 14 20:37:23 yep Nov 14 20:45:11 mangix went around heavily replacing git:// with https in packages a while ago, so while some people might not care, there is some heavy leanings to https for git urls Nov 14 20:46:49 there it makes sense, but it has failed for me in CI, which should be environment "under control" Nov 14 22:29:34 that I did Nov 14 22:30:06 the last one in packages seems to be minidlna. Unfortunately that project seems abandoned. Nov 14 22:46:08 abandonded aka "works just fine" ? :) Nov 14 23:00:03 karlp: actually it does not: https://github.com/openwrt/packages/issues?utf8=%E2%9C%93&q=is%3Aissue+is%3Aopen+minidlna Nov 14 23:01:01 gerbera seems to be an active alternative. Unfortunately, it's huge. **** ENDING LOGGING AT Fri Nov 15 02:59:58 2019