**** BEGIN LOGGING AT Wed Nov 28 03:00:01 2018 Nov 28 06:52:32 Hi, I'm calling sysupgrade from another script and I have the issue that sysupgrade does return with value 10 (which is supposed to be UBUS_STATUS_CONNECTION_FAILED). Am I calling it the wrong way or how can I deal with this return value? Nov 28 08:52:36 hm, is there a sensible way of extracting the dts from a kernel? running or just vmlinux? Nov 28 08:52:49 well, might be a uImage actually Nov 28 08:57:14 jwh: there is /proc/device-tree or whatever Nov 28 08:57:25 oh hm Nov 28 08:57:45 is that just a kernel representation or the compiled dts? Nov 28 08:58:26 it's splayed out in a fs tree Nov 28 08:58:31 ohhh Nov 28 08:58:37 * jwh looks Nov 28 08:58:41 so, not a plainjane .dts file Nov 28 08:59:34 might be enough, assuming its in the vendor kernel of course Nov 28 09:00:29 find /proc/device-tree Nov 28 09:04:56 hm, useful Nov 28 09:05:01 terminal corruption in minicom Nov 28 09:05:11 is cu included in anything in the openwrt packages? Nov 28 09:06:08 picocom is. Nov 28 09:06:14 oh that will do Nov 28 09:06:27 simpler/easier/smaller/morereliable.... :) Nov 28 09:06:31 heh Nov 28 09:06:51 ok maybe its not minicom then Nov 28 09:07:08 well, probably not, it's a pig, but it should _work_ Nov 28 09:07:15 maybe I need a new db9 <-> rj45 thinger Nov 28 09:07:18 it looks kinda shady Nov 28 09:07:31 you need less db9s in your life, that's what you need. Nov 28 09:07:34 haha Nov 28 09:07:43 well I don't think my usb one works either Nov 28 09:07:49 but I still need the adapter for that anyway Nov 28 09:07:53 maybe I got a spare Nov 28 09:09:35 bleh, will have to wait until I fly home at the weekend Nov 28 09:10:41 https://i.imgur.com/SUEIBNJ.png :D Nov 28 09:10:45 blogic: Well, that created as many questions as answers, but at least they're providing :) Nov 28 09:12:45 ohhh hang on Nov 28 09:12:52 might be my bad Nov 28 09:13:21 Hi people Nov 28 09:14:20 How do I build for a second target with out over writing the .config I have saved for the first target? Nov 28 09:14:44 hm, no inittab reloading still? Nov 28 09:14:45 reboot it is Nov 28 09:16:39 yeah, it's a "missing" feature, but it's not really somethingyou need real often :) Nov 28 09:17:03 yeah Nov 28 09:22:03 Monkeh: What did you guys got? Could you forward it to me? :) Nov 28 09:25:33 gch981213[w]: A simple listing of the values valid for those registers, but I'm unclear if it's assigning a function to a pad, or a pad to a function. Also no sense to be made of the drive strength. Nov 28 09:25:49 gch981213[w]: Either way it's incomplete for our purposes but hopefully we can get the rest.. Nov 28 09:29:24 Someone working in MTK told me what their internal datasheet says about those bits but that's even less than what the code comment said and that guy providing me these info isn't familiar with mt7628 :( Nov 28 09:31:28 There's going to be a few dozen pages of context needed Nov 28 09:36:16 If these registers work the way I think, even the descriptions of all of them won't suffice without mapping the pad names to physical pads on the parts.. Nov 28 09:39:42 Indeed. Nov 28 09:40:45 And then we only have a massive code cleanup to do to use the information :P Nov 28 09:54:46 https://store.ubnt.com/collections/routing-switching/products/edgerouter-12 Nov 28 09:54:50 holy shit thats cheap Nov 28 09:56:57 hrm. I just backported the rpcd updates for exec times to 18.06, and now ubus is running, rpcd-mod-file is installed, but the file object isn't in "ubus list" Nov 28 09:57:15 strace shows it opening /usr/lib/rpcd/file.so Nov 28 09:57:36 * karlp tries a cleaner clean build. Nov 28 10:00:32 ohh, https://mikrotik.com/product/crs305_1g_4s_in Nov 28 10:01:02 looks like mvebu, is it in openwrt yet? Nov 28 10:02:25 jwh, it can be ported Nov 28 10:02:50 well obviously, that wasn't the question though ;) Nov 28 10:04:08 hm Nov 28 10:04:17 lots of decent devices about now, I'm kinda impressed Nov 28 10:04:34 who do I throw money/beer/women at? Nov 28 10:05:03 blogic: take a chill pill dude :) Nov 28 10:11:36 jwh: that er12 poe is a bit crap poe, Nov 28 10:12:00 wouldn't trust them to supply poe Nov 28 10:12:07 even if it is passthrough Nov 28 10:12:28 but 2 real eth ports plus sfp, then switched ports Nov 28 10:12:52 that mikrotik is the crazy one though, 4 * SFP+ for 149 bucks RRP Nov 28 10:13:13 mvebu so DSA probably works Nov 28 10:16:47 hmmm, 802.3ad "offloading" on that edgerouter Nov 28 10:17:03 because processing LACP is hard Nov 28 10:20:01 mkresin: what do you think of v3 of the ath79 airrouter support? https://patchwork.ozlabs.org/patch/1003523/ and https://patchwork.ozlabs.org/patch/1003524/ Nov 28 10:31:57 russell--: focused on ramips at the moment. will get back to ath79 as soon as I'm finished with ramips Nov 28 10:44:44 'lo Nov 28 10:47:28 eh, nope, still no ubus file. what's exploded in rpcd? Nov 28 10:48:41 karlp: rpcd and mod-file are the same version? Nov 28 10:49:01 yeah, and rpcsys is there, just not file Nov 28 10:49:11 interesting Nov 28 10:49:20 I've just backported your two timing things, Nov 28 10:49:33 (ran into the "invalid timeout" thing yesterday) Nov 28 10:49:45 yeah that was a stupid brainfart Nov 28 10:50:11 no biggie Nov 28 10:50:48 I believe https://git.openwrt.org/?p=project/rpcd.git;a=commitdiff;h=2cc4b998ec78fc8fb9e12a5e67e71c34b2511719 could've broken file Nov 28 10:51:15 I guess the exec_timeout symbol is unresolved when dlopen()'ing file.so Nov 28 10:51:27 will take a look Nov 28 10:51:31 shouldn't that fail somehow then? Nov 28 10:51:52 I'm not very pro at reading strace output, but it seemed to just open it and continue Nov 28 10:51:58 yeah, dlopen fails, so its silently skipped Nov 28 11:02:51 https://mikrotik.com/product/CRS326-24G-2SplusRM Nov 28 11:02:52 ohhhh Nov 28 11:03:14 that would be really nice to get openwrt on... Nov 28 11:03:21 first "real" switch Nov 28 11:05:36 what's wrong with just using the vendor sw in this case? Nov 28 11:05:45 its mikrotik Nov 28 11:06:01 they really don't understand switching, or at least how to present it for configuration Nov 28 11:06:17 also why not openwrt? Nov 28 11:06:32 that is crazy cheap too Nov 28 11:09:40 https://mikrotik.com/product/crs328_4c_20s_4s_rm Nov 28 11:09:40 man Nov 28 11:11:37 I mean it still suffers from the typical mikrotik design of there being a weak cpu and crappy link between switch and cpu, but otherwise... Nov 28 11:14:23 hi, i'd like to submit a change to uhttpd, but as it is not on github i'm not sure of the process Nov 28 11:15:16 timk: send a patch to the openwrt-devel list please (with git-send-email) Nov 28 11:15:30 prefix it with [PATCH uhttpd] Nov 28 11:15:42 jow: thanks Nov 28 11:20:19 karlp: pushed some fixes (and runtime tested them this time) Nov 28 11:21:47 ok, thanks, I'd been doing my "does this idea of using rpcd even work" just using a hard patch on the timeout value Nov 28 11:26:58 karlp: and does it work? Nov 28 11:27:42 rebuilding, sorry Nov 28 11:28:09 I mean if using rpcd works in principle for your use case. I think for file to become a viable default plugin it needs an additional level of acl control Nov 28 11:28:24 like it is being done for uci where we can define per file access Nov 28 11:28:43 for me, yeah, I'm going with the "only root is allowed anyway" so adding file access is ... no worse than it alreayd was. Nov 28 11:29:17 the exec function is problematic, its essentially a wildcard allow anything function Nov 28 11:30:04 it is, yah, but writing individual luci controllers or ubus/rpcd plugins was just tedious. Nov 28 11:30:20 I can imagine, yeah Nov 28 11:30:54 I'll probably start converting some luci json endpoints to direct ubus access soonish Nov 28 11:31:16 it helps with performance Nov 28 11:31:55 but to achive some code size reductions, a lot of code has to be converted to overcome the initial code size increase due to the addtional code paths Nov 28 11:41:41 jow: the biggest thing really was just adding the acl that allowed other ubus stuff to work. Nov 28 11:42:25 even if I do single things instead of exec, having access to "the same" ubus via the json-rpc as root has on the console is much simpler. Nov 28 11:58:01 Hi jow,karlp. The toolchain appears to be working. For curiosity sake I'm trying to determine where the include/site variables are exported. And I'm come across a weird Makefile syntax. Nov 28 11:58:34 This is inside a define block: "$(1) : export CONFIG_SITE:=$$(CONFIG_SITE". What is this doing? It looks like the export is part of the prerequisites Nov 28 11:58:49 that doesn't make sense. Trying to google this syntax is hard. Nov 28 11:59:16 personalyl I'd recommend not doing "for curiosity" projects with the makefiles unless you've got something broken. Nov 28 11:59:49 but that looks like setting a shell env var to the makefile var, but only when it's used, not what it was set to when it was defined. Nov 28 12:01:20 Neighbor11111114: yes, this essentially exports CONFIG_SITE to the shell for the target noted in $(1) Nov 28 12:01:39 oh I see that is shell syntax not Makefile syntax. Doh, the define confused me Nov 28 12:01:40 thanks guys Nov 28 12:01:56 Neighbor11111114: this is part of a larger macro which is evaluated through $(call ...) and $(eval ...) hence the $$ escaping Nov 28 12:02:12 in Make you need to use two dollar signs to escape one dollar sign Nov 28 12:02:30 and four to escape two? ;) Nov 28 12:02:37 yes Nov 28 12:02:38 that can go pretty deep... Nov 28 12:02:39 Yup that makes sense to me. I just assumed it was Makefile syntax. Nov 28 12:02:52 iirc the scan.mk has 6 $ Nov 28 12:03:11 that's why I hate Makefile Nov 28 12:03:25 you have to know how "deep" your code is used and use a proper amount of $ Nov 28 12:03:34 which you pretty much never remember Nov 28 12:03:41 and have to check over and over Nov 28 12:03:51 *and use Nov 28 12:04:01 the love of $ is the root of all evil Nov 28 12:04:23 lol russell-- Nov 28 12:04:30 :D Nov 28 12:05:06 I used to dislike Makfiles a lot. Not so much now. There are quite a few problems that direct acyclic graph solves nicely and you can use Make in cool ways to take care of it for you Nov 28 12:05:27 autotools is the one I don't like. But it's because I don't think I truly understand it yet lol Nov 28 12:06:59 * russell-- wishes make had an option to export graphviz dot files Nov 28 12:07:41 jow: that last rpcd patch works thanks. Nov 28 12:18:55 jow/karlp. Sorry for pestering. I still don't understand. I thought ':' is builtin for true. And if you look at man page of true it states anything that follows it is ignored. Why or how does the shell realise that export x=y is an export statement and not options that should be ignored by true Nov 28 12:19:45 Neighbor11111114: the colon in this context is make syntax Nov 28 12:19:56 $(1) : shell code Nov 28 12:20:40 Isn't that where the prerequisites go? Shouldn't it be Nov 28 12:20:44 $(1) : Nov 28 12:20:48 \t shell code Nov 28 12:23:47 goal:export is special make syntax Nov 28 12:24:20 it will cause GNU make to export prior to invoking the shell commands of "goal" Nov 28 12:25:58 (this is why you don't go down the rabbit hole unles you need to...) Nov 28 12:26:00 ok awesome, now it makes sense. Thanks! Nov 28 12:26:10 take this example: https://pastebin.com/Af8C5zQY Nov 28 12:26:15 karlp: the opposite! Glad I know. I'll be doing that in future Nov 28 12:27:04 if you run "make -f /tmp/test.mk" you'll see it first echo "ENV of 'foo': ... BAR=1 ... FOO=1 ...", then "ENV of 'bar': ... BAR=1 ..." Nov 28 12:27:22 protip: writing makefiles using all these sorts of features is a surefire way to have maintennance problems, and put people off working on them. Nov 28 12:28:13 karlp: add some self generating code and three levels of eval to the mix :) Nov 28 12:29:09 heh Nov 28 12:29:14 very cool, thanks Nov 28 12:44:46 fatal: Cannot create /var/lib/iptraf-ng: No such file or directory Nov 28 12:44:46 hm Nov 28 13:07:03 ldir: stintel: kernel bumps pushed to staging, but I will not push them to master probably (as there are many targets in the process of updating currently) Nov 28 13:10:45 xback: unfortunately I messed up my build host and I'm currently travelling Nov 28 14:09:32 jow: can this be merged? https://github.com/openwrt/packages/pull/6621 Nov 28 14:11:52 is there anything in the update we need? Nov 28 14:11:57 or security fixes? Nov 28 14:12:06 iirc we had some fallout the last time this got bumped Nov 28 14:12:23 since I don't have time to runtime test this atm I'd rather onyl merge it if here's an urgent need Nov 28 15:19:20 build #176 of malta/be is complete: Success [build successful] Build details are at http://release-builds.openwrt.org/18.06/images/builders/malta%2Fbe/builds/176 Nov 28 16:42:39 am I going crazy or is there no dts files for the octeon target? Nov 28 16:42:46 are they else where? Nov 28 16:48:53 jwh: arch/mips/boot/dts/cavium-octeon/ in the kernel ? Nov 28 16:50:52 jow: AFAIK no Nov 28 17:02:32 will need to add PKG_CPE_ID Nov 28 17:10:39 rmilecki: :P nobody likes regressions Nov 28 17:13:47 jwh: search for hex d00dfeed in Your image to get where the dtb starts, then use dtc to decompile it Nov 28 18:09:47 My c7 is now on ath79 all is good so far Nov 28 18:30:01 karlp: good that you got the ramdisk booting up Nov 28 21:01:01 Hauke: yah, bit of fiddle, next step to do it completely via FEL instead of fel then tftp, but now I can _do_ stuff :) Nov 28 23:15:37 nbd: FYI, had an mt76 device which lost wifi connection to its neighbors after 10 days of uptime (running openwrt 18.06 code) but it recovered a few hours later. no log entries whatsoever Nov 28 23:16:08 MT7628N actually Nov 28 23:16:44 will try owrt master now Nov 29 00:19:56 Can somebody explain the constant rebases in the repository to me? Why not just rebase your local code ontop of origin/master and then commit that? Nov 29 00:20:37 there are multiple staging trees Nov 29 00:21:04 there's no rebases in the actual shared tree Nov 29 00:23:10 Not sure if I understand that. My remote is https://git.openwrt.org/openwrt/openwrt.git, and I'm on origin/master, and whenever I pull a change, it appears to me as somebody had rebased something. How would I check out the no-rebase shared tree? Nov 29 00:24:03 i just get fast-forwards when i pull that Nov 29 00:25:47 On my git log, I see 5ffacceb7b (3 hours ago), then e880a30549 (4 weeks ago), then 795211093f (5 hours ago), among the recent commits, which looks rebased to me Nov 29 00:26:58 merges are not the same as rebases. Nov 29 00:27:07 did it pull fastforward? Nov 29 00:27:25 merge will leave the chron history "weird" but not cause any merge problems. Nov 29 00:29:16 hmm, is there an easy way to query the used spi-nor and/or nand flash chips (manufacturer/ model) from sysfs (I'm not really interested in /proc/mtd, as that's pretty messed up with ubi)? I'm usually looking at dmesg, but the buffer flows over before I can get a chance to see it. Nov 29 00:30:32 @karlp I know that, but I don't see any merge commits in my git log. I've checked the last few months. The only thing I could think of is that [pull] rebase = preserve in my .gitconfig doesn't do what I think it does Nov 29 00:33:02 monx: try git log --format=fuller and check the commit date, not the author daye Nov 29 00:33:30 It's not that I'm having merge problems, I was just trying to find out which recent change made my images not fit on my 4/32 anymore. Must have been something in the last couple days Nov 29 00:34:00 for that, git bisect is your friend Nov 29 00:36:15 Well, if I check out HEAD~10, will the feeds update to what the newest feeds were back then or will I still get the newest? Because whether or not git bisect works kind of depends on that **** ENDING LOGGING AT Thu Nov 29 02:59:59 2018