**** BEGIN LOGGING AT Tue Apr 14 02:59:57 2020 Apr 14 03:01:07 ok I must have uwsgi module for nginx on Apr 14 03:01:11 now it's working :) Apr 14 03:01:40 except overview page takes 6 seconds to load Apr 14 03:19:09 rr123: that shows a diff, not the individual commits Apr 14 03:19:34 greearb: does ath10k-ct support temperature reading with qca9880? Apr 14 03:20:25 from reading the upstream code, it seems dependent on firmware support. I'm on CT-HTT Apr 14 03:23:34 mangix: git log -p tag branch? Apr 14 03:24:27 doesn't look like it Apr 14 03:24:47 it's running git log on two branches independently Apr 14 03:25:34 anyway for a 7-hour fun run with 19.07 on archer A7-v5, 1. i removed mtd so sysupgrade never worked(by mistake) 2. while trying to make nginx lighter I must not remove uwsgi(?) otherwise it will not run 3. you can not enable nginx-luci and nginx-ssl-luci at the same time unlike luci+luci_ssl which are ok. 4. luci default ssl(uhttpd) is 3 times faster than nginx when loading pages Apr 14 03:25:47 ah you need .. between them Apr 14 03:27:57 merge conflict. of course. Apr 14 03:28:06 short answer, will stay with uhttpd, bye nginx Apr 14 03:30:35 rr123: :) Apr 14 03:48:02 mangix, yes it does Apr 14 04:18:46 nbd: alive? Apr 14 04:31:59 blocktrron: I've just pushed an image template for lzma-loader. It can be used by adding $(Device/uimage-lzma-loader) to device image recipe. Apr 14 04:32:29 Hi people just booted my wrt3200acm on Kernel 5.4 all seems to be ok so far thanks for all your work. Apr 14 04:33:05 blocktrron: Let's make these commits one by one because we need testing on actual devices. Apr 14 05:05:00 greearb: where is the information exported? Apr 14 05:49:38 Lifting lzma dictionary size limit and enabling lzma-loader for all mt7621 routers saves 15KB space. Is it worth to do so? Apr 14 05:51:40 There's a risk: we will break devices if u-boot can't load uncompressed image. I guess no vendor bothers to remove this capability. Apr 14 05:52:46 meanwhile 15KB isn't a significant amount of space. Apr 14 06:07:50 greearb: find / | grep ath10k_t results in nothing Apr 14 06:14:31 hrm I found that hwmon7 has ath10k_hwmon listd Apr 14 06:14:53 cat temp1_input shows a consistent 61000 Apr 14 06:14:57 wonder if that's legic Apr 14 06:15:01 *legit Apr 14 06:16:10 ah it sometimes goes to 62 Apr 14 06:16:13 seems like it is Apr 14 06:23:57 gch981213, extra space is always worth, there is 7549 packages with installedsize less than 15k Apr 14 06:25:01 according to https://openwrt.org/_media/packages_dump_tab_separated.zip Apr 14 06:25:20 may be more, but Calc refuses to load all Apr 14 06:47:08 ok. RFT patch on using lzma-loader for mt7621 sent. Apr 14 06:52:34 Pepe: do you guys use your own buildbot for packages? Apr 14 06:54:23 AGH I just lost ssh access to my router Apr 14 07:10:44 mangix: you need a buildbot? Apr 14 07:11:49 no I'm just wondering. On their build servers, mpd is built. On the OpenWrt ones it is not. Apr 14 07:12:18 at least I think they have one... Apr 14 07:13:25 mangix: sorry I'm missing some context here Apr 14 07:15:49 aparcar[m]: https://downloads.openwrt.org/releases/faillogs-19.07/i386_pentium4/packages/mpd/mini/compile.txt Apr 14 07:16:14 i'm spinning up a local build to see if I can replicate Apr 14 07:16:33 check out jows 2.4.1 branch Apr 14 07:21:26 mangix: https://github.com/aparcar/buildbot/pull/3 Apr 14 07:21:54 by default it builds all targets which takes quite some time... Apr 14 07:41:27 jow: when caching the buildbot, should I cache build_dir and staging_dir? Apr 14 07:41:31 or just one of them Apr 14 07:49:02 PaulFertser: hello! thanks yo your help i sucesfully booted my device and bother ethernet ports plus the wifi interface are working Apr 14 07:50:27 PaulFertser: rather than tftp, i'm woking on a script that exploits the root command injection and uses the mtd binary that the oem firmware is using in order to write on tha flash Apr 14 07:52:22 i believe that, even with the fact that i'm chaning the partition structure, i could just split my new image in six parts with the sizes of the al partition and write themt his way Apr 14 07:54:34 stintel: there's an argument here to rename kmod-sched-core to kmod-sched and kmod-sched to kmod-sched-extras Apr 14 07:55:37 is there a jshn aware person in da house ? Apr 14 07:56:34 aware yes Apr 14 07:56:39 but guru, no Apr 14 07:58:58 I'm trying to iterate over an array of key, value pairs but can't work out how to do so Apr 14 08:00:03 aldoaldo: indeed you can. I still recommend to research if the bootloader has some trick to allow tftp uploading without serial interaction (for recovery purposes etc) Apr 14 08:00:07 I have an array of something called 'tins', each index of tins contains a number of key, value pairs. Apr 14 08:01:41 PaulFertser: thank you, i'll try both ways just to learn to play :) Apr 14 08:02:11 The example code at https://openwrt.org/docs/guide-developer/jshn iterates over a key that contains an array of values (I think) Apr 14 08:25:34 ldir: https://paste.ubuntu.com/p/fGcGjfbgfp/ Apr 14 08:25:47 extracted part from some config script Apr 14 08:26:52 ldir: json https://paste.ubuntu.com/p/BD4xk83KV5/ Apr 14 08:27:44 Guys ? when will be available https://github.com/openwrt/packages/tree/master/utils/docker-ce package ? 20.08 ? Or is possible to backport it to 19.XX ? Thanks Apr 14 08:29:21 i don't think packages are backported Apr 14 08:30:11 ynezz: thx, will try Apr 14 08:34:25 PaulFertser: is it posible that the leds are now directly controlled by the ath9 driver? I have /sys/class/gpio/export and /sys/class/gpio/import but i don't have any gpiochip* Apr 14 08:34:57 aldoaldo: one (or probably two?) leds can be controlled from ath9k, yes. Apr 14 08:34:58 but i have /sys/class/leds/ath9k-phy0 and the power, lan, and wan leds eems to be already working Apr 14 08:35:15 i have yet to check the wifi but i think that too Apr 14 08:35:26 there are left two "RSSI0" leds left Apr 14 08:35:58 aldoaldo: but not having any gpios avalable for exporting certainly means you didn't enable generic gpio in your DTS. Apr 14 08:37:40 @Borromini oh, and is there any eta for 20.XX ? Apr 14 08:38:43 muhaha no mate Apr 14 08:39:10 the jump from kernel 4.19 to kernel 5.4 is still ongoing. Apr 14 08:39:42 PaulFertser: and you're of course right, thank you again Apr 14 08:39:54 aldoaldo: :) Apr 14 08:40:32 oh noe, so I guess its better to run openwrt in docker instead of running docker in openwrt .. Apr 14 08:48:23 depends on the use case I guess :) Apr 14 08:48:53 is OpenWrt *in* docker actually useful, networking wise I mean? Apr 14 08:49:08 I could imagine that its comparatively much effort to wire all that traffic in and out again Apr 14 08:49:51 aparcar[m]: staging_dir Apr 14 08:50:11 however the workspace is cleaned before each run anyway Apr 14 08:59:15 jow: fair. I though to check out GitLab CI for staging builds as it seems easier than extendin the current buildbot Apr 14 08:59:42 it works quite well, following the same steps as buildbots, except some of it's perks like kmod archives https://gitlab.com/aparcar/openwrt/-/jobs/509700462 Apr 14 09:01:25 aparcar[m]: FYI https://gitlab.com/ynezz/openwrt/-/blob/staging/.gitlab-ci.yml Apr 14 09:01:40 ynezz: https://gitlab.com/aparcar/openwrt/-/blob/staging/.gitlab-ci.yml Apr 14 09:02:36 well, feel free to reinvent the wheel, I'm just trying to save you some time Apr 14 09:04:09 ynezz: no don't get me wrong :) I had the gitlab-ci file from a long time ago and just extended it to create the json stuff, your approach looks better, as usual Apr 14 09:05:19 ldir: sorry my mum called but I see ynezz did help already Apr 14 09:26:25 aparcar[m]: well, your usual response to anything is either "why not something newer" or "why not something else", but great that there's something else now which is doing the same Apr 14 09:26:32 :) Apr 14 09:29:21 I have to increase the start param of my daemon to survive a reboot. Is somethint wrong with just setting it to 99? Apr 14 09:30:37 no, but sounds like a symptom of a race condition Apr 14 09:31:03 daemon needs umdns Apr 14 09:31:06 and umdns is at level 80 Apr 14 09:33:34 jow: true true, I'll look into staging tree builds for buildbot Apr 14 09:34:34 jow: "buildmaster: update to v2.4.1, buildslave: update to Debian 10" nice, when do you plan to deploy those? :) Apr 14 09:34:43 ynezz: when I have time Apr 14 09:34:53 probably not within the next week Apr 14 09:34:53 ah, so it's ready? Apr 14 09:34:59 cool, great Apr 14 09:35:00 more or less Apr 14 09:35:06 thanks! Apr 14 09:36:05 jow: aparcar[m]: staging is quite tricky, because its not that easy to find out from the changes what target to build in order to verify the "correctness" Apr 14 09:36:52 I don't know if it's possible to have some local, in the tree file which would tell buildbot to build certain target, user selected Apr 14 09:37:09 .buildbot.yml with TARGET=ath79/generic or such Apr 14 09:37:33 or build more targets (sometimes needed as well) Apr 14 09:38:30 maybe a filter based on the first words of the commit messages? if none starts with tools, they could be cached. Apr 14 09:38:47 implying strict rules on commit messages Apr 14 09:39:19 or based on git file changes, similar how the openwrt/packages.git ci works Apr 14 09:39:45 I tihnk commit message tags make sense Apr 14 09:40:09 something like Affects: ar71xx/generic or Build-Test: mvebu Apr 14 09:40:19 or git notes? Apr 14 09:40:32 not sure waht git notes are Apr 14 09:41:23 they're "ideal" in theory, except virtually unused, so might as well not exist ;) Apr 14 09:41:54 ah, additional object metadata Apr 14 09:41:59 sure, why not Apr 14 09:42:18 maybe with some script in https://git.openwrt.org/?p=maintainer-tools.git;a=summary to simplify managing them Apr 14 10:47:46 Is LXC still available only with custom build ? Apr 14 10:51:45 jow: BTW that Debian version bump in will set default gcc version to gcc (Debian 8.3.0-6) 8.3.0 (current is 6), is that intentional? Apr 14 10:52:06 s/bump in/bump in buildslave/ Apr 14 10:53:06 don't take me wrong, I'm in favor for this bump, just asking :) Apr 14 10:54:56 need to check in more detail later Apr 14 10:55:31 I usually used the oldest available gcc in the corresponding stable debian to ensure best portability of the resulting SDKs and IBs Apr 14 11:50:05 gch981213: I did read your comment in the ML regarding the fact that if uncompressed kernel grew to 5MB in 32-MB ram devices, then there won't be any room for user-space. I assumed the kernel would relocate itself and overwrite unneeded data when decompressed, but I guess I am missing actually lots of things here. can you explain this a little bit more in-depth? Thanks!! Apr 14 12:12:03 mrkiko: You're right. Apr 14 12:13:11 gch981213: so this is unrelated to lzma-loader, the point is just 5.4 is too big for 32 mb ram devices Apr 14 12:14:07 mrkiko: Not yet but I bet it will come soon :) Apr 14 12:17:39 on one hand the status light seems to indicate that it's done booting, on the other, the switch didn't come up. Apr 14 12:20:46 just throw that 32MB RAM / 4MB flash devices in the trash, it's 2020 people Apr 14 12:22:37 gch981213: does it make sense to look at /proc/iomem and sum up kernel code + data + bss from there to see how much space it takes currently? I'm on 4.19.91 and get 5635 kiB on my ath79 device right atm. meminfo reports free 3692, available 5228 Apr 14 12:24:51 stintel: TP-Link is really good at using VxWorks and saving memory space. Apr 14 12:25:49 but we can't with latest linux :P Apr 14 12:25:51 stintel: my device does what I want from it (I guess I'd only like to be able to run my own recursive DNS resolver) would be sad to trash it... Apr 14 12:26:47 totally agree with you guys Apr 14 12:27:30 gch981213: and, lzma-loade in ramips will solve the problem for NOR devices only, right? and, why do we need different bootloader for DIR-860? sorry, didn't read all the history Apr 14 12:28:04 PaulFertser: Honestly I'm just too lazy to choose a different text_base for our lzma-loader in ramips :P Apr 14 12:30:55 mrkiko: It also solves problems for nand devices. I found out yesterday that lzma-loader doesn't have to read flash. It's also possible to let u-boot copy compressed kernel with it into memory. Apr 14 12:32:23 all other devices use uimage header and should be able to load uncompressed loader with compressed kernel together into memory. Apr 14 12:32:46 DIR-860L uses seama header and I don't want to compress kernel twice. Apr 14 12:34:15 so I prefer to let u-boot load a compressed lzma-loader only and let lzma-loader read kernel from flash. Apr 14 12:44:43 gch981213: thanks!! If you want me to test your changes on a R6220 let me know... even tough I'll have to reflash stock u-boot; or maybe blocktrron can help us inthis? Apr 14 13:25:06 mangix, you good on the ath10k-ct temp readings? Apr 14 13:48:55 gch981213: I'll confirm your patch is working on a R6220 and a nanoHD later today Apr 14 14:10:17 dengqf6: vlan_filter should not be enabled right? If it turn that on on the bridge interface, it kills all the traffic and only a reboot can fix that. Apr 14 14:25:45 jow: fwiw, uboot-sunxi _does_ respect git-src package overrides, just that you need to used "make package/boot/uboot-sunxi/{clean,compile}" (which does correspond to it's pðarth) Apr 14 14:26:18 make package/uboot-sunxi/{clean,compile} will find a package, but blow up attempting to compilearm code with x64 gcc or something, Apr 14 14:26:33 which is where i got waylaid Apr 14 15:10:07 Rene__: that's what my another patch fixes Apr 14 15:12:34 Rene__: https://lore.kernel.org/netdev/E1jEB0y-0006iF-5g@rmk-PC.armlinux.org.uk/ Apr 14 15:42:58 The user & group dropdowns in luci_statistics, exec module are empty, code looks like o = ss.option(widgets.UserSelect, 'cmduser', _('User')); Apr 14 15:43:18 any ideas what is wrong? Apr 14 15:44:45 ldir: missing acl rules most likely Apr 14 15:45:43 ldir: checking Apr 14 15:52:07 ldir: fixed https://github.com/openwrt/luci/commit/1dad678379c9a1914b0cb8c339d93732518c3226 Apr 14 15:52:23 ldir: you can also edit the json file on your router and issue "/etc/init.d/rpcd reload" Apr 14 15:55:55 thank you Sir Apr 14 15:57:55 jow: luci-app-ksmbd is missing uci permission Apr 14 16:32:40 PaulFertser:ok, got my sunxi booting from spi flash, but I odn't get jffs overlay, Apr 14 16:33:14 in that forum post you linked, I don't see how they're getting that either, even if they do make a "sysupgrade.bin" Apr 14 16:36:12 karlp: the forum post doesn't mention some kernel options unfortunately. Apr 14 16:36:18 karlp: please pastebin your boot log Apr 14 16:37:51 hrm, this is just usererror I guess, I wrote the initramfs-kernel.bin to spi flash, so it's loaded it properly, got the _rest_ of the spi machinery working, but not the right rootfs Apr 14 16:38:31 I'm trying to neatly add the sysupgrade parallel to existing sdcard as options, without just hardcoding it all like that forum user Apr 14 16:46:27 karlp: it would be nice to have it properly supported indeed Apr 14 16:47:09 well, it's gonna be weird, the spi flash isn't mounted by default, so it has to be a second board version. Apr 14 16:47:45 overlays ... don't seem super ready to go in our buildroot yet, and we'r eusing uimage, not fit imges for sunxi, so can't use that either. Apr 14 16:47:58 still, gonna keep at it, Apr 14 16:48:44 do you knkow what determines whether "target images" includes sysupgrade? Apr 14 16:50:11 karlp: just what one specifies in IMAGES := Apr 14 16:50:48 flash only writes at 72kB/sec, so takes a little while :) Apr 14 16:54:03 gotta figure out why the wifi's gone too, last time I was poking this with an initramfs and I had the wifi and bluetooth working, but don't get either now :) Apr 14 17:54:09 who wants to watch some stuff about wifi6 https://commscope.webex.com/commscope/j.php?MTID=m91ed3557289c97fae9c49a3a49049275 Apr 14 17:58:44 I'll go heckle. Apr 14 18:06:21 build #119 of sunxi/cortexa7 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-19.07/images/builders/sunxi%2Fcortexa7/builds/119 Apr 14 18:43:38 I enjoyed heckling. Apr 14 19:26:34 blogic: ping Apr 14 20:20:06 why does the build system requires ncurses? Apr 14 20:21:14 make menuconfig is ncurses based Apr 14 20:24:48 pkgadd libncurses5-dev sounds like it includes headers which are then compiled into something, however I can run menuconfig without compiling anything Apr 14 20:25:25 then remove gcc and try again :) Apr 14 20:26:26 dont forget to run `git clean -xfd scripts/config` as well Apr 14 20:27:23 ynezz: so menuconfig is actually compiled code every time? wow Apr 14 20:27:30 no Apr 14 20:29:57 ynezz: so without running menuconfig, it wouldn't be required - right? Apr 14 20:30:08 it is still required Apr 14 20:30:22 menuconfig, defconfig etc. are all the same executable iirc Apr 14 20:30:26 it's just invisible, doesn't follow verbose(V) convention Apr 14 20:30:31 make -s Apr 14 20:30:49 so it happens, but you cant see it, magic Apr 14 20:31:03 $ ldd scripts/config/mconf Apr 14 20:31:03 linux-vdso.so.1 (0x00007ffeb179e000) Apr 14 20:31:03 libncursesw.so.6 => /lib/x86_64-linux-gnu/libncursesw.so.6 (0x00007f62271a9000) Apr 14 20:31:06 libtinfo.so.6 => /lib/x86_64-linux-gnu/libtinfo.so.6 (0x00007f622717a000) Apr 14 20:31:09 libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f6226fb7000) Apr 14 20:31:15 libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007f6226fb2000) Apr 14 20:31:15 /lib64/ld-linux-x86-64.so.2 (0x00007f6227241000) Apr 14 20:33:23 thank you Apr 14 20:37:16 The imagebuilder does not seem to require ncurses, can I add a ifndef in prereq-build.mk or are there special cases where the imagebuilder would require it? Apr 14 20:38:18 doesn't IB let you use menuconfig to select what to put in it? Apr 14 20:38:20 isn' Apr 14 20:38:22 t that the point? Apr 14 20:38:54 karlp: the SDK does, IB only allows to use make image PACKAGES="foobar" Apr 14 20:39:10 hrm... Apr 14 21:19:50 Is it enough to have "&gpio {status = "okay";};" in the DTS file in order to get the /sys/class/gpiochip* devices? Apr 14 21:20:06 Or are there other requirements Apr 14 21:22:35 ok.... so device names can't have _ in them. that's a great fantastically visible trap to fall into Apr 14 21:55:11 jow: would you be okay with a JSON merging step like this? https://github.com/aparcar/buildbot/pull/3/commits/2ac45d104c2a349711efb0d482841c3ea14d4a09 Apr 14 22:02:09 karlp: it should be vendor_model anyway Apr 14 22:51:49 * mangix hates how broken glibc Apr 14 22:51:49 is Apr 15 00:36:57 what makefile target actually gets uboot built? Apr 15 00:42:09 is 'interpreter' option still valid for uhttpd? was not on openwrt uhttpd page and not used anywhere it seems Apr 15 00:44:28 make defconfig again after touching things "unrelated" yay. because the uboot entries are HIDDEN Apr 15 00:50:47 ldir: what timezone are you in? Apr 15 00:51:11 he's uk based, Apr 15 00:54:07 karlp: thanks! Apr 15 01:11:29 ok, got as far as booting to a "no root partition" but I have a squashfs sysupgrade image built at lest. progress Apr 15 01:20:08 karlp: may I ask what you're working on? Apr 15 01:22:08 spi flash support for a sunxi board. Apr 15 01:22:32 the layers and layers of magic redirection is un pleasant **** ENDING LOGGING AT Wed Apr 15 02:59:57 2020