**** BEGIN LOGGING AT Sun Oct 25 02:59:57 2020 Oct 25 05:18:50 @aparcar are you proposing an alpine based openwrt? <-- it's a sign Oct 25 05:29:04 mangix: a sign for what? Oct 25 05:45:33 just build it on top of debian or ubuntu ;p Oct 25 05:45:47 why stop at alpine Oct 25 05:53:52 I'm missing something here, I propose a different package manager and people see an enemy takeover. OpenWrt is more than its packages manager 🙄 Oct 25 06:25:49 hahaha Oct 25 08:30:13 been thinking - https://downloads.openwrt.org/releases/19.07.4/targets/sunxi/cortexa53/ that boards have ext4 and squashfs image for sd card... but there is sysupgrade image. is that intended? Oct 25 08:30:38 there is no sysupgrade image* Oct 25 08:45:53 * mangix has no idea Oct 25 09:02:44 how does the packages feed work these days Oct 25 09:02:52 I have 3 new trivial python libs Oct 25 09:03:02 do I have to send a PR or can I just push the, ? Oct 25 09:39:56 blogic: one is fine Oct 25 09:40:04 but three commits please Oct 25 09:40:41 also please fix the SDKs... Oct 25 09:50:42 damex: mangix for your personal delight https://gitlab.alpinelinux.org/alpine/infra/alpine-mksite/-/issues/2 Oct 25 09:53:19 fascinating how bad upstream cavium/octeon sources are Oct 25 09:54:25 that boards could have let's say 4 qsgmii ports + 1x loop port plus something else. it does not even check that ports are valid. it just initialize them. so for 4x qsgmii + 1x loop system will have 20 network interfaces. Oct 25 10:35:04 anyone else getting "tmp/.config-package.in:63474:warning: leading whitespace ignored" which is to do with "config LUCI_LANG_nl" ? Oct 25 10:58:37 blogic: it is preferred to create a PR. But given that the buildbots are totally broken courtesy of a broken SDK, feel free to puah directly. Oct 25 11:01:19 aparcar[m]: I will say merge requests on aports is pretty painless Oct 25 11:03:09 blogic: actually jefferyto and bkpepe will complain. Create a PR. Oct 25 11:12:47 ldir: yes me too looks like a bug in the luci feed Oct 25 11:14:07 does someone have a recent example how people fix drivers in a tree when they ignore properties like status = "disabled"; ? Oct 25 11:14:32 linux tree ofcourse Oct 25 11:15:13 Hauke: I tried to fix but I've been out of things so long, failed Oct 25 11:19:17 damex: of_device_is_available() checks it Oct 25 11:20:29 PaulFertser: thanks, gonna read about it now Oct 25 11:25:34 damex: target/linux/mediatek/files-5.4/drivers/net/phy/mtk/mt753x/mt753x_mdio.c can be considered an example I guess Oct 25 11:26:02 damex: and the wiki has section on how to use quilt to prepare a kernel patch if you need to fix a driver that's already upstream. Oct 25 11:26:04 Does Dutch (the language) in Dutch mean Nederlands? Oct 25 11:28:02 PaulFertser: drives is upstream but it is in a such bad shape so i am not so sure anyone would accept a fix Oct 25 11:28:34 s/drives/driver/ Oct 25 11:28:35 damex: if it's not in staging then you should be able to send a patch following the regular Linux contribution guidelines (with scripts/check_maintainers.pl etc) Oct 25 11:28:45 ldir: I think I found the fix Oct 25 11:28:48 PaulFertser: what if it is staging? Oct 25 11:29:14 damex: chances that your patch will be merged without review are even higher then :) Oct 25 11:29:26 damex: because people do not give a damn about staging Oct 25 11:30:33 Hauke: great! Oct 25 11:30:41 https://github.com/torvalds/linux/blob/master/drivers/staging/octeon/ethernet-sgmii.c#L28 that comment Oct 25 11:30:58 so turns out we don't have autonegotiation on SGMII interfaces Oct 25 11:31:39 PaulFertser: nice, thanks ;p Oct 25 11:40:23 damex: yes, if your code doesn't look obviously wrong and doesn't break the build GregKH will just merge it afaict. Oct 25 11:40:39 ldir: https://github.com/openwrt/luci/pull/4540 Oct 25 11:42:37 Ha, now that's what I tried but I obviously didn't do the right magic somewhere as I still got the error. :-) Oct 25 11:42:39 Even without autoneg logic it should be sane to check the node "status" and not enable what's disabled. I wonder if the DTBs that are already present on hardware have status=okay for those nodes because if they do not you might risk breaking it. Oct 25 11:43:21 ldir: I had to delete my tmp directory Oct 25 11:43:35 ahhhhh! Oct 25 11:47:26 PaulFertser: status absent == status okay as far as i see Oct 25 11:47:50 so if it is not specified on dtb that is already present - it should be fine? Oct 25 11:49:24 https://github.com/torvalds/linux/blob/master/arch/mips/cavium-octeon/octeon-platform.c#L657 i think i can adjust that to apply the same logic if subnode is disabled? Oct 25 11:57:27 probably i can do the something similar in driver itself in staging tree Oct 25 12:00:40 i wonder why such logic is in arch/mips/cavium-octeon instead of octeon_ethernet or something like that though Oct 25 12:06:42 damex: you're right, absent == okay Oct 25 13:30:45 hmmmm - package/feeds/packages/ntpd/Makefile: user ntp id 123 is already taken by package/utils/busybox/Makefile - that's new Oct 25 13:43:43 what do i need to do for files from target/linux//files to be copied to a kernel directory? Oct 25 13:54:53 Just saw Daniel Golle commits dropping some capabilities Oct 25 13:55:06 first thanks ! Oct 25 13:55:58 Would it make sense to move capabilities / no_new_privs / user switching in procd ? Oct 25 13:58:02 with ambient capabilities and ubus acl this allow to run many things as non root, and this should not be optional Oct 25 14:07:13 Hmm… not so good… Oct 25 14:07:15 package/feeds/packages/ntpd/Makefile: user ntp id 123 is already taken by package/utils/busybox/Makefile Oct 25 14:07:31 for.... ntp :) Oct 25 14:08:05 So, just ignore it? :P Oct 25 14:09:38 maybe it works if both are named the same Oct 25 14:09:46 make oldconfig complains loudly. Oct 25 14:10:11 ntp vs ntpd Oct 25 14:10:27 https://paste.debian.net/1168625/ Oct 25 14:11:01 I'm using ntpd from busybox. Oct 25 14:14:50 i mean the user, if if you change ntpd to ntp in busybox Oct 25 14:17:52 I know how to fix it, it's just that this broke master again. :/ Oct 25 14:20:56 indead :( Oct 25 14:28:30 not really the same breakage level Oct 25 14:31:31 I don't understand how that can get committed to master without having been encountered already. Oct 25 14:34:32 don't install ntpd from packages? wouldn't be seen until you do right? Oct 25 14:38:36 karlp: yes, hadn't thought of that. Question is, how to fix it? 2 packages that use the same user ID. Oct 25 14:39:54 isn't that the intention? you're not going ot want to ntpds anyway? Oct 25 14:40:04 or, again, switch to names instead of uids? :) Oct 25 14:41:32 I'm not trying to build them both Oct 25 14:43:38 the dependencies are blowing up just because they exist. Oct 25 14:44:42 + USERID:=ntpd=123:ntpd=123 is trying to use names Oct 25 14:45:45 I don't think things should blow up at 'make menuconfig' stage simply because ./scripts/feeds/install -a has been run. Oct 25 14:57:20 could do as gentoo does and use a meta-package for user/group creation and both busybox-ntp and ntpd could depend on this meta package Oct 25 14:57:34 but ... meh Oct 25 14:59:54 stintel: gentoo does not create any users in meta packages. or atleast should not Oct 25 15:02:34 if busybox has now claimed ownership of user & group 'ntpd' is the simplest thing to drop the claiming of the same uid/gid in packages/ntpd since isn't it very unlikely that busybox won't be installed? Oct 25 15:04:05 how come files from /files directory is copied to kernel directory only once when you do a fresh kernel build? this is so bad Oct 25 16:04:41 damex: have you used gentoo recently? Oct 25 16:04:56 stintel: did they change something? Oct 25 16:05:09 how recently ? :) Oct 25 16:08:44 damex: acct-user/$username acct-group/$groupname was introduced more than a year ago Oct 25 16:10:01 https://www.gentoo.org/glep/glep-0081.html omg Oct 25 16:10:58 stintel: okay, thanks. i haven't used gentoo for a while and that change is pretty recent ;p Oct 25 16:11:41 you and I have a different definition of recent :) Oct 25 16:14:39 yeah, i still think that tmux is pretty new piece of software ;p Oct 25 16:15:12 compared to screen Oct 25 16:23:36 my ERL is running an image built from master 46d ago. this is almost ancient to me :P Oct 25 18:07:36 Back from my nap and master is fixed! Thanks, dangole! :) Oct 25 18:12:47 rsalvaterra: got surprised by having broken it again and realized it just after my nap Oct 25 18:13:04 rsalvaterra: hope not two many of the phase2 builds will break because of that Oct 25 18:13:09 *too many Oct 25 18:15:23 Heh… it's easy to forget about the buildbot when we do our own buids. :P Oct 25 18:17:52 updated openwrt/upstream, https://sdwalker.github.io/uscan/index.html Oct 25 18:28:00 i have a device that is hard to recover but need to test pll settings. is there a way to apply pll settings for ag71xx on fly? Oct 25 18:47:29 I don't see any fixes of the SDK? blogic ? Oct 25 18:52:31 mangix: what is NIH stuff? Oct 25 19:08:33 so https://biot.com/switches/models that hardware is my best bet for openwrt switch? Oct 25 19:09:59 tried to play with edgeswitch 8 150w and switch ports are not available in uboot. no way to test a new build and it also have Oct 25 19:10:36 https://biot.com/switches/gs1900-10hp that one looks tasty ;p Oct 25 19:30:47 yes, the zyxel devices have been tempting me for quite a while as well (the gs1900-24e, to be more exact), contrary to D-Link (for which you have to check the hardware revision in detail) they seems to stick to one SOC Oct 25 19:37:08 aparcar[m]: NIH is "not invented here" Oct 25 19:38:36 karlp: thank you Oct 25 19:39:10 I can't speak to what mangix was referring to, but you can read about NIH behaviour and draw your own conclusions :) Oct 25 19:51:01 hello Oct 25 19:51:45 i am trying to add a new set of graphs to the statistics module, however, my module with the graphs definitions does not seem to be loaded Oct 25 19:52:05 do I need to do anything else other than dropping a file in /usr/lib/lua/luci/statistics/rrdtool/definitions/ ? Oct 25 20:55:39 I'm never sure when mangix is trolling and when actual opinions are stated Oct 25 21:06:52 Yikes, my Archer C6 image grew almost 200 kiB in the last couple of days, without changing my configuration. What gives…? O_o Oct 25 21:08:05 I don't see what could have caused this… recent changes were pretty minor… Oct 25 21:24:38 Ah, I see. My bad. I noticed I'm now building tor with -fPIC, which I wasn't. Carry on. :P Oct 25 21:32:14 is there comprehensive openwrt/dsa documentation anywhere? Oct 25 21:36:30 i need untagged packets arriving on lan ports to be delivered to the cpu on vlan 300 and vlan 300 tagged packets to be untagged leaving the LAN ports, while other vlans (100 and 200) leave the lan ports tagged. Oct 25 22:09:20 aparcar[m]: not invented here. stuff like procd, libubox, etc... totally custom stuff. Oct 25 22:10:37 Yea Nobody (Not me) plans to merge alpine with openwrt. How is that related to the package manager? Oct 25 22:31:39 it isn't Oct 25 22:31:53 my point was that apk was as far as it will go Oct 25 22:32:10 since that user was asking about it Oct 25 22:32:41 anyway, what's the recommended 802.11ax hardware nowadays? Oct 26 01:22:20 mangix: there are basically only two options, ipq807x with qcn5024/ qcn5054 or mt7915e with either mt7621a or mt7622a underneath. the former exists, but is expensive - the later hasn't really reached the market yet (there are some underpowered APs with mt7621a+mt7915e announced, but I don't think you can get them yet) Oct 26 01:24:49 ipq807x has basic target support, and there are some efforts around adding support for the Xiaomi Mi AIoT Router AX3600 - but that isn't really available yet. hnyman seems to have scored the Netgear rax120, due to its price the ax3600 seems to have a little more backing behind it Oct 26 01:25:55 it's in sight, but not really there *yet*. ipq60xx and ipq50xx will be cheaper, but that's still very volatile (patches are still in the process of getting processed mainline, both basic SOC support and for ath11k) Oct 26 01:27:35 due to the well-known SOCs underneath, Mediatek m7915e might be quicker to get OpenWrt support - once some devices shipping it arrive on the market (although I'd skip anything with mt7621a in favour of waiting for mt7622a+mt7915e). but ipq807x pretty much has a head start so far Oct 26 01:31:53 Xiaomi Redmi Router AX6 would be a cheaper variant of the AX3600, only two radios instead of three (the third radio is a 1x1 ath10k 2.4+5 GHaz WLAN card, originally meant for Xiaomi's AIoT ecosystem) - but lower tx power (half of the ax3600 values) Oct 26 01:37:10 pkgadd: mt7621 + mt7915 is available since months in form of the D-Link DIR-X1860 Oct 26 01:38:34 blocktrron: ah, o.k., thanks Oct 26 01:40:17 It's niche i admit Oct 26 01:41:08 And mt7915 support isn't there in master yet, but i think this is the path of least resistance atm Oct 26 01:42:48 Also i assume the MT7915D+MT7621 will be the most prominent choice for sub-100 2x2 DBC designs. Oct 26 01:43:29 I've seen a lot of activity around it on linux-wireless@vger, so I just assume it won't be too far away. and obviously using a well-known SOC underneath will make supporting it easier ('just' the new wireless driver to take care of), but at least mt7621a imho feels too slow for an 802.11ax capable router Oct 26 01:46:01 Well, certainly a upgrade considering 2x2 802.11ac gear with 100 Mbit/s ports Oct 26 01:47:52 the ax3600 with ipq8071a and 4*1.4 GHz cortex a53 is kind of tempting me, considering its price to performance ratio **** ENDING LOGGING AT Mon Oct 26 02:59:57 2020