**** BEGIN LOGGING AT Mon May 03 02:59:56 2021 May 03 03:44:52 i seem to recall being able to select CONFIG_ALL=m, make defconfig is telling me that's not valid May 03 03:48:00 CONFIG_ALL=y May 03 05:40:22 russell--: You can use CONFIG_ALL=y but you'llwant to tag it with IGNORE_ERRORS="m n" or it'll fail on ones that won't compile for the target May 03 05:40:34 on the make May 03 05:40:51 IE make -jx IGNORE_ERRORS="n m" May 03 06:38:45 yeah, IGNORE_ERRORS was known May 03 06:39:46 * pkg_hash_fetch_best_installation_candidate: Packages for ip-tiny found, but incompatible with the architectures configured May 03 07:01:08 a few packages fail on missing header file: fatal error: gnu/stubs-32.h: No such file or directory May 03 07:02:32 gnuplot, knot-resolver, luajit, mtools, snort, snort3 May 03 07:43:12 Damnit, I ran into the stub issue at one point, but I don't remember what fixed it May 03 07:45:57 Oh May 03 07:46:10 russell--: x86? Did you install the multi-libs? May 03 07:47:25 russell--: gcc-multilib May 03 07:48:11 It's required for 32-bit x86 under a 64-bit build environment.. I ran into the issue with luajit and some other things back when I started the suricata stuff May 03 07:48:58 as a host package I mean May 03 07:56:41 ath79 May 03 07:57:04 Yeah, but your build system is x86/amd64? May 03 07:57:10 yes May 03 07:57:17 You need gcc-multilib then May 03 07:57:23 or it'll die on the 32-bit code May 03 07:57:44 why is it looking at the host includes for package builds? May 03 07:57:48 At least, that was my experience May 03 07:57:52 Because its a requirement May 03 07:58:01 https://openwrt.org/docs/guide-developer/build-system/install-buildsystem May 03 07:58:05 Check LuaJIT May 03 07:58:26 under Package prerequisites May 03 07:58:56 Might be worth a look anyway.. It's how I cleared my issues if I remember right May 03 07:59:01 in the end, i don't actually care that they didn't build May 03 07:59:07 just mentioning it May 03 07:59:15 s'Ok :) May 03 07:59:19 they aren't on my critical path May 03 08:00:46 my build box is archlinux May 03 08:01:17 Shouldn't matter.. you're still buildig with gcc May 03 08:02:57 Anyway.. G'luck with it :) I'm out for the evening.. Have a great one! May 03 09:40:16 git grep 'pkg_hash_fetch_best_installation_candidate' returns nothing May 03 09:41:26 russell--: git grep in opkg repo? May 03 09:41:48 there are a few recent bug reports about this, it's related to a change in opkg from a few months ago May 03 09:42:19 i did the git grep build TOPDIR May 03 09:42:51 IIRC this error can be triggered by unrelated issues like missing dependencies (so it's a case of wrongly reported error) May 03 09:51:09 i did a make package/opkg/host/{clean,compile} and trying again .... May 03 09:53:18 fwiw, this exercise is really only about getting a broad sample of package configure.ac's to see how people deal with embedding version strings in built software May 03 09:53:47 most seem to just be static strings May 03 09:55:52 http://lists.openwrt.org/pipermail/openwrt-devel/2021-May/034936.html May 03 11:02:07 diffconfig.sh gives me lots of "syntax error" / "unknown option "imply"" on master May 03 11:02:34 this "imply" thing seems to have been introduced with the big wireguard change, anybody knows what it is? May 03 11:04:46 I get this when trying to build 21.02 branch with LuCI: May 03 11:04:51 * pkg_hash_fetch_best_installation_candidate: Packages for luci-base found, but incompatible with the architectures configured May 03 11:04:52 * satisfy_dependencies_for: Cannot satisfy the following dependencies for luci: May 03 11:04:54 * cgi-io May 03 12:35:52 rmilecki: I suspect some opkg regression May 03 12:36:07 rmilecki: there's been similar reports in the tracker May 03 12:38:40 rmilecki: https://git.openwrt.org/?p=project/opkg-lede.git;a=commit;h=5936c4f9660248284e8a9b040ea3153d3ea888de May 03 12:38:51 rmilecki: could you check if reverting this helps? May 03 14:08:44 zorun: ping? May 03 14:09:04 f00b4r0: pong May 03 14:09:45 zorun: you mentioned a "hard_config patch" on GH: which patch? May 03 14:10:35 the one you ACKed May 03 14:10:53 https://patchwork.ozlabs.org/project/openwrt/patch/20210503103947.3527098-1-baptiste@bitsofnetworks.org/ May 03 14:11:44 ah ok, I misread the message: I thought "this" meant the size mod. I see now you meant you're merging the device backport :) May 03 14:12:01 did you figure out what's going on with your device btw? May 03 14:12:21 no, but I'm starting to think there's something wrong with it... May 03 14:15:50 well, it's made by mikrotik ;P May 03 14:25:06 yeah :D it's good hardware for the price, but if there were good alternatives we would switch to something else May 03 14:37:18 I have a second one, it's not in a pretty shape (we opened it and soldered wires to get serial access during last battlemesh) but I will try this one May 03 14:37:24 bbl May 03 16:20:45 jow: sorry, I had to leave, i'm back, working on that May 03 16:21:03 jow: btw: how do you check if given device uses DSA or not (runtime, e.g. in LuCI)? May 03 16:35:18 jow: i reverted 55a43e1ab0ad ("opkg: update to git HEAD") but it didn't help in my 21.02 branch May 03 16:35:33 master works fine on the other hand, i've yet to compare my .config May 03 17:02:53 jow: ok, that was my fault after all, I didn't have packages repo updated (indexed) May 03 17:03:24 so feeds update & feeds install -a -p packages did the trick May 03 17:14:21 rmilecki: dsa... actually LuCI does not check for DSA but for the lack of swconfig May 03 17:14:39 jow: does it check for binary? May 03 17:14:44 without swconfig, DSA and non-DSA config are pretty much the same from the LuCI pov May 03 17:15:27 rmilecki: afair it checks for both the swconfig executable and for switch topology info in /etc/board.json May 03 17:15:36 jow: ah, ok May 03 17:15:38 thanks for info May 03 17:15:48 and in some places it also checks for the presence of a "config switch" section in /e/c/network May 03 17:16:09 thanks, I'm curious to try DSA on bcm53xx one day May 03 17:16:54 simply speaking, DSA is just bridging netdevs together May 03 17:17:31 and speaking tagged VLAN on the wire or doing vlan trunking would be setting up bridge vlan filters on top of a bridge May 03 17:18:25 whether the netdevs your bridge together happen to be DSA ports or distinct phys/nics does not matter May 03 17:18:47 in fact you have to dig pretty deep in sysfs to actually figure out if a port happens to be a DSA one May 03 17:18:59 you generally only need that if you want to display some kind of status information May 03 17:19:50 May 03 17:55:42 jow: thanks for info May 03 17:55:47 (explanation) May 03 18:01:33 rmilecki: could you try this patch for opkg: https://patchwork.ozlabs.org/project/openwrt/patch/20210502205912.23753-1-hauke@hauke-m.de/ May 03 18:02:24 I looked into a similar problem yesterday and this should be the solution: https://patchwork.ozlabs.org/project/openwrt/patch/20210502210228.23942-1-hauke@hauke-m.de/ May 03 18:05:59 jow: i was thinking about adding "switch" LuCI page that would display each switch port port and its status (carrier, speed) May 03 18:06:20 jow: so there is no easy way to list available switches and their ports? May 03 18:07:58 rmilecki: you can read this information with ethtool, and use the board.d config to know which switch ports are there May 03 18:08:54 info (carrier, speed) is already in ubus in network.device May 03 18:09:15 Hauke: so i guess i'm looking for board.json then May 03 18:12:01 rmilecki: there is an easy way actually May 03 18:12:27 rmilecki: recent kernels expose a DEVTYPE=xxx property in the sysfs uevent files May 03 18:13:58 rmilecki: # grep DEVTYPE= /sys/class/net/lan1/uevent May 03 18:13:59 DEVTYPE=dsa May 03 18:14:17 oh, nice! May 03 18:16:06 to group DSA ports by switch, I found two ways May 03 18:16:20 1) group by the result of readlink /sys/class/net/*/device May 03 18:16:44 2) group by the contents of /sys/class/net/*/phys_switch_id May 03 18:17:12 however due to the lack of an actual multi-DSA-switch device here I cannot confirm that it'd actually work multiple DSA switches May 03 18:17:17 but I suppose it would May 03 18:18:49 ah, actually I forgot about a 3rd way which is the most reliable May 03 18:19:00 cat /sys/class/net/*/iflink May 03 18:19:11 that will contain the ifindex of the physical parent interface May 03 18:19:26 e.g. on a zbt wg2626 it reports iflink=2 for lan1, lan2, lan3, lan4 and wan May 03 18:19:33 2 corresponds to "eth0" May 03 18:19:52 so one can deduce that lan1..lan4 + wan are DSA ports of switch eth0 May 03 18:20:09 all that info is available through "ubus call luci-rpc getNetworkDevices" in current master May 03 18:20:21 it is just not exposed as a kind of switch overview in LuCI yet May 03 18:20:55 should be fairly easy to do though, e.g. by providing a partial status applet in /www/luci-static/resources/view/status/include/ May 03 18:28:59 jow: thanks, i'll see if I can implement that using info you provided! May 03 18:40:26 rmilecki: in luci-rpc getNetworkDevices filter for "devtype": "dsa" and group by "parent": ... May 03 18:40:36 jow: just found it :) May 03 18:40:46 jow: getNetworkDevices is great May 03 18:41:16 thought i'm not big fan of naming things like that with "luci" prefix May 03 18:41:34 well it's LuCI's private luci.so rpcd plugin May 03 18:41:59 we could move it into a separate sysinfo.so kind of plugin, along with other generic functionality May 03 18:42:18 sounds good May 03 18:42:24 sysinfo like plugin May 03 18:57:57 this may need an update :) > grep TITLE feeds/luci/libs/rpcd-mod-luci/Makefile May 03 18:57:58 TITLE:=Rapid reverse DNS rpcd module **** ENDING LOGGING AT Tue May 04 02:59:57 2021