**** BEGIN LOGGING AT Thu Apr 09 02:59:57 2020 Apr 09 03:40:00 There's an one-year-old fstools bug: https://bugs.openwrt.org/index.php?do=details&task_id=2231 and an old patch for it: https://patchwork.ozlabs.org/patch/1082599/ Apr 09 03:42:34 and this GitHub PR author tries to get dev's attention: https://github.com/openwrt/openwrt/pull/2906#issuecomment-610951636 Apr 09 03:43:20 could someone maybe take a look into it? :) Apr 09 05:20:14 gch981213: i'm actually using that patch on my x86 + f2fs 19.07 device Apr 09 05:20:26 it's so i can do a second overlay w/f2fs Apr 09 05:20:45 works fine imo :) i agree it should be looked at Apr 09 06:24:54 gch981213: my guess is that noone really cares or remember how fancy things like extroot work :P Apr 09 06:25:51 gch981213: i found countless bugs in fstools & USB / hotplug / block(d) - i thought there were many users testing that with e.g. USB drives Apr 09 06:59:19 blogic: morning Apr 09 07:03:41 anyone an idea on how to add files to the created rootfs but on a "per profile" basis? Apr 09 07:56:55 can someone point me to code that builds build_dir/target-arm_cortex-a9_musl_eabi/root-bcm53xx ? Apr 09 07:57:10 it happens right after generating package indexes Apr 09 07:57:38 there is something doing mkdir -p $ROOTFS/tmp Apr 09 07:58:08 and then doing opkg --offline-root --force-postinstall --add-dest root:/ --add-arch all:100 --add-arch arm_cortex-a9:200 install Apr 09 07:58:31 that opkg command comes from include/rootfs.mk Apr 09 07:58:36 but I can't find a caller Apr 09 08:00:06 rmilecki: target-dir-% rule in include/image.mk Apr 09 08:01:00 nbd: that opkg_target (from image.mk) includes -f $(mkfs_cur_target_dir).conf but I don't see that used in my "make V=s" log Apr 09 08:02:05 gch981213: so I've looked at this github PR and do not understand it Apr 09 08:02:12 rmilecki: ah, sorry. that was only for the device specific rootfs dirs Apr 09 08:02:17 you're looking for package/Makefile Apr 09 08:02:22 $(curdir)/install Apr 09 08:02:35 gch981213: fstools was fixed but not updated? or fstools fixed, updated in master but not backported to 19.07 ? Apr 09 08:02:36 nbd: thanks! Apr 09 08:03:41 jow: Neither. The fix got no response for a year. Apr 09 08:04:12 gch981213: your comment seemed to imply that it was applied to fstools.git Apr 09 08:04:40 m4t: may we add your Tested-by ? Apr 09 08:04:56 for https://github.com/openwrt/openwrt/pull/2906/files#diff-4dd79a2f1257c5e3dfb18dc78bfdb552 Apr 09 08:05:13 jow: Oops. I mean it should be applied to fstools first and we make a package bump in openwrt. Apr 09 08:05:48 the patch is annoiyng to review Apr 09 08:06:44 but as I understand it, it restores the ability to discover non-ubi rootblocks when fstools was built with ubi support? Apr 09 08:07:20 since right now the code will *only* handle ubi once built with ubi support Apr 09 08:12:55 https://bugs.openwrt.org/index.php?do=details&task_id=2396 looks interesting too Apr 09 08:13:00 could explain various wifi issues Apr 09 08:17:18 * gch981213 reads fstools code for the first time Apr 09 08:21:24 jow: wow, looks interesting Apr 09 08:22:50 nbd: hi! sorry for barging in but are you aware of the partial erase issue with recent kernels? It seems it no longer works Apr 09 08:28:27 jow: the first portion of the patch fixes rootfs discovery in check_extroot Apr 09 08:28:28 but what's going on in main_extroot? Apr 09 08:28:28 main_extroot mounts local overlay to check /etc/config/fstab. The patch makes it fall back to check config in /tmp/overlay. Apr 09 08:43:25 f00b4r0: i wasn't aware of it, but unfortunately i don't have time to work on it at the momnet Apr 09 08:44:10 nbd: ok, well at least now you know ;) I tried to look at the code but honestly I don't understand enough of it nor the changes between 4.14 (working) and 4.19 (not working) to figure out what's going on Apr 09 08:47:04 aha! overlay is already mounted to /tmp/overlay before extroot routine is called. Apr 09 08:49:17 f00b4r0: 4.19 is nothing... you should see 5.5 Apr 09 08:49:37 or 5.6 Apr 09 08:49:47 rmilecki: i'd rather not :} Apr 09 08:50:45 this commit blows up our patch: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=46b5889cc2c54bac7d7e727a44d28a298df23cef ("mtd: implement proper partition handling") Apr 09 09:24:51 Hello. Has anybody posted any patches for saving debug information on a build while actually stripping what goes into the firmware yet? Apr 09 09:24:55 I need that :) Apr 09 09:25:27 As the saying goes, "necessity is the mother of all 'get off your ass and do it yourself'" Apr 09 09:26:54 Also, we need to add a global framework for enabling -flto! This works with SO many programs now -- even firefox uses it by default (at least with gcc9) Apr 09 09:27:39 I don't mean enable -flto globally, so much as a flag for each package that says if it works or not and they auto-enabled. Apr 09 09:29:26 what would be the best way to recompile an initram image if only the DTS has changed? Apr 09 09:33:29 acalvo: I have a hack for that somewhere. Apr 09 09:33:56 you can use the original firmware sysupgrade, compile the dts into a dtb and then inject it. Let me see if I can find my hack Apr 09 09:34:39 amazing, thanks Apr 09 09:35:21 first you compile your dts with this: $kd/scripts/dtc/dtc -O dtb -o $dtb $dts Apr 09 09:35:27 where $kd is your kernel directory Apr 09 09:37:40 hmm, I think you'll have to unpack the sysupgrade and uncompress the kernel first :( I'll pastebin this part of the script Apr 09 09:37:54 thanks Apr 09 09:41:19 staging_dir/host/bin/patch-dtb is the program that sticks the compiled dtb into the kernekl image Apr 09 09:41:45 http://dpaste.com/1G5Q8F9 Apr 09 09:42:24 That is part of how I build an image w/o using the OpenWRT build -- I essentially had to reverse engineer OpenWRT's build process a few years ago because I was thrown into this project without any driection!! Apr 09 09:42:38 but if you can to the last few steps in reverse and then forward again you should be able to do what you want to Apr 09 09:42:59 (you don't need most of those variables, I just put them all for reference) Apr 09 09:43:40 So going backwards, you can ignore the padjffs2, then use binwalk to separate out the components of your sysupgrade image Apr 09 09:43:57 once you have the offsets, use dd to separate out the chunks Apr 09 09:44:50 finally, once you unlzma the kernel image (or presujming that's how it was compressed) you can run patch-dtb! Apr 09 09:50:14 acalvo: do you have binwalk intsalled? Apr 09 09:50:22 I think I can simplify this for you Apr 09 09:51:34 dansan: people usually keep the build_dir and run gdb + gdbserver with a script that's part of OpenWrt. Apr 09 09:52:01 dansan: so binaries in build_dir are all unstripped, if you keep it all, you have all the debug that was enabled. Apr 09 09:52:42 Hi all, I'm a bit confused how ubox logd integrates with syslog. For example the functions openlog, syslog, closelog I assume are part of libc. How do those functions know to commuicate and how to commuicate with ubox logd? Apr 09 09:53:32 dansan: thanks! Apr 09 09:56:48 PaulFertser: Well maybe this has changed since 18? I know, I'm still in the stone age. But I thought everything in the build dir was stripped before it was even copied to the package's .pkgdir. Amd I wrong? Apr 09 09:57:07 acalvo: I'm trying to give you an example, but it's not working yet (a working example is always nice :) Apr 09 09:58:13 dansan: build_dir/target-mips_24kc_musl/hostapd-wpad-mini/hostapd-2019-08-08-ca8c2bd2/wpad: ELF 32-bit MSB executable, MIPS, MIPS32 rel2 version 1 (SYSV), dynamically linked, interpreter /lib/ld-musl-mips-sf.so.1, with debug_info, not stripped Apr 09 09:58:21 acalvo: ah hah!! Apr 09 10:00:42 dansan: I do not think it was ever stripped there Apr 09 10:01:15 acalvo: an example!! http://dpaste.com/2VH1R11 Apr 09 10:01:35 PaulFertser: cool. Is that from 19 or the HEAD? Apr 09 10:01:49 dansan: it was HEAD back then Apr 09 10:02:03 ok Apr 09 10:03:09 dansan: https://openwrt.org/docs/guide-developer/gdb?s[]=remote&s[]=gdb Apr 09 10:04:06 Now I just have to troubleshoot strange message, ever seen it before? http://dpaste.com/2KW555M Apr 09 10:04:43 PaulFertser: oh crap, I think I forgot to submit my gdb python patch -- you need it for kernel debugging now Apr 09 10:04:54 has anybody else added a python support patch for gdb yet? Apr 09 10:06:58 PaulFertser: oh sweet! I haven't used remote-gdb yet. It apparently offers support to even restart programs. I usually just run gdbserver and then if I need to re-run it, go back to my board and start gdbserver again Apr 09 10:07:29 dansan: looks like a mismatch between your toolchain version and musl startup or it's for a wrong arch. In any case, that's something not default about your build config. Apr 09 10:08:25 hmm, ok, thanks. I can't figure out what I changed that would cause this. But I haven't rebuild this firmware tree in a few monthys... Apr 09 10:09:55 Or probably somehow it's calling wrong "as" assembly translator? Apr 09 10:10:01 I wonder if this could be the -mips16 -minterlink-mips16 flags, I may have messed with those recently Apr 09 10:10:06 yeah, that's more likely Apr 09 10:10:32 Hm, -mips16 is probably doing that to you. Apr 09 10:11:18 As if gcc generates misp16 assembly but as is not aware Apr 09 10:11:29 I never quite figured out where that belongs and doesn't, or why it shows up in the "extra optimizations" flag (forgot the exact name of it) Apr 09 10:12:30 I suggest you just follow OpenWrt default config in this regard, as it's getting auto-build tested regularly. Apr 09 10:12:33 oh shit, that's what I changed!! oops Apr 09 10:12:47 :blushes: Apr 09 10:13:25 good point! Apr 09 10:13:30 "as it's getting auto-build tested regularly." Apr 09 10:18:32 acalvo: any luck? Apr 09 10:18:49 I'm curious :) Apr 09 10:48:15 jow: am working on a reply to your forum 'extract qdisc stats' question Apr 09 11:38:33 ldir: great :) Apr 09 11:38:52 the good news is that tc has json output Apr 09 11:39:46 ldir: yeah, I guess its more a question of what a good graphical representation of the stats would be Apr 09 11:39:59 bar charts, pie charts, curves, tables etc. Apr 09 11:40:09 I'm not into that topic at all Apr 09 11:40:57 dunno, well if I explain stuff and then suggest what I think are key parameters we can go from there. But at the moment I guess all the stats might as well be written in Mandarin :-) Apr 09 11:43:19 bumping ramips/mt76x8 to 5.4 caused image oversize for 4M devices Apr 09 11:43:47 don't think anybody still cares about 4M devices tbh Apr 09 11:43:47 gch981213: I'd say disable image building for them Apr 09 11:44:01 gch981213: users could try to make smaller images with the SDK Apr 09 11:44:06 erm the ImageBuilder Apr 09 11:44:24 jow: already did that :) Apr 09 12:17:48 sweet!! -g3 gets me a 28GiB build_dir :) Apr 09 12:44:38 jow: I've replied and left a whole load of worms for you to think about :-) Apr 09 12:53:17 dansan: still working with it, need to adapt to my environment Apr 09 13:02:49 jow: sure. btw this is the original patch https://patchwork.ozlabs.org/patch/1082599/ Apr 09 13:08:52 oh nvm, i see gch981213 sent the patchwork link too :) Apr 09 13:11:36 acalvo: cool. I was tyring to give you the commands to run from within an openwrt build tree -- but you probably don't have one of those... dang, sorry Apr 09 13:12:02 acalvo: I dunno what type of lzma compression is used on the kernel -- I couldn't use my /usr/bin/lzma to decompress it Apr 09 13:13:11 PaulFertser: ok, for 18.06.6, it has definitely installed executalbes into staging_dir/target-xxx/root-ramips because it's 780Mi instead of the usual 250Mi and it's too big. So going to have to hack something to get those stripped Apr 09 13:14:07 A co-worker has started the process of porting our crap on to 19 -- there's SO much to clean up on our end (for our proprietary crap) and so much to clean up and try to get merged in main line too Apr 09 13:14:50 lol, I've ordered a bigger SDD because I ran out of space trying to build with CONFIG_DEBUG! :) Apr 09 13:15:17 dansan, you should probably get an old style HDD for such Apr 09 13:22:21 Redfoxmoon: Well, I didn't actually "run out of space", I still have old HDD raid space Apr 09 13:22:44 but the difference in speed when I have to do this a lot and is great Apr 09 13:23:16 I even tweaked the damn spi nand flash drivers to shave an extra 9 seconds off of the boot time, partially because I flash and reboot these things so muhc Apr 09 13:23:34 oh dear. Apr 09 13:23:37 lol! Apr 09 13:23:49 I'm amazed that it has no problems yet Apr 09 13:24:02 I'm sure I've flashed it several hundreds of times Apr 09 13:24:43 how big is your SSD? Apr 09 13:24:51 What I *really* wish we had was a working, n-tier file system cache thing that uses more than one store, so I can have HDD --> cheap SATA SDDs --> expensive M.2 SDDs Apr 09 13:24:56 NOR flash is usually rated for 10k or even 100k cycles. So several hundreds is far from the limit. Apr 09 13:25:23 PaulFertser: true, but when you do it so many times... ones brain starts to wonder :) Apr 09 13:25:27 dansan: bcachefs :) Apr 09 13:25:35 SAS disk raid? :P Apr 09 13:25:36 I just wish it would be upstream already Apr 09 13:25:39 Borromini: Well, I have a 400G sata ssd and small 30G optane Apr 09 13:25:52 stintel: but doesn't that only do two tiers? Apr 09 13:26:28 Redfoxmoon: oh no, just software raid Apr 09 13:26:48 dansan: it can do multiple disk groups so more than 2, but there is just foreground_target (for writes), background target (for cold data) and promote target (for hot data) Apr 09 13:27:14 Redfoxmoon: I think I built it using mdadm and then lvm on top of that, for volume management -- but I guess LVM will do the software raid too now. Apr 09 13:27:22 :-) Apr 09 13:27:39 dansan: but! you can set those targets per-file with xattr Apr 09 13:27:40 dansan: i've 'benched' my nvme SSDs (970 EVO) against my S-ATA one (850 EVO), the NVMe barely makes a dent compared to the S-ATA SSD when it comes to compiling OpenWrt. Apr 09 13:27:45 stintel: but only two targets? just foreground and background? Apr 09 13:28:01 stintel: ooooooh! Apr 09 13:28:07 dansan: foreground, background, promote. but the per-file with xattr makes it theoretically unlimited Apr 09 13:28:15 it's quite neat Apr 09 13:28:28 unfortunately quite experimental too Apr 09 13:28:59 but Kent is doing a great job imo, and usually pretty quick to look into issues Apr 09 13:29:20 I've been using it for some years and lurking in #bcachefs on OFTC for some time as well Apr 09 13:29:34 I tried to use bcache when I built this system in mid 2018 and it definitely wasn't ready yet -- I need a working system for work. However, I kind of feel I don't have an excuse for trying it on an experimental volume Apr 09 13:29:36 I'm running it in production, because that's always the best test :D Apr 09 13:29:51 stintel: Better or worse experimental than btrfs? ;) Apr 09 13:30:14 Monkeh: pff, these days it's about the same. I had a deadlock on my workstation's btrfs raid1 root yesterday Apr 09 13:30:23 Monkeh: ever since 5.0 it's going downhill really Apr 09 13:30:32 stintel: I had my rootfs implode last year Apr 09 13:30:43 Very upsetting, fsck and related basically pointless Apr 09 13:30:46 stintel: hehe, not very busy in there :) Apr 09 13:30:47 Went back to ext4. Apr 09 13:31:00 interestingly enough I have a 6x2TB BTRFS RAID6 fs that has been going for years Apr 09 13:31:15 only crapped out on me once, and could recover all but most recently written data with btrfs restore Apr 09 13:31:21 I'm not touching btrfs unless I learn it much more deeply Apr 09 13:31:29 now I would like to switch the metadata to raid1c3 Apr 09 13:31:30 Well, workstation had a few odd moments in its life, including the PSU exploding Apr 09 13:31:34 but for that I need 5.something Apr 09 13:31:35 So I guess it was all too much for it. Apr 09 13:31:49 and 5.* is really not something I would currently dare to use, for valuable data Apr 09 13:31:53 yes I have backups, but still Apr 09 13:32:04 It just seems to have no capability for recovery from metadata corruption Apr 09 13:32:06 and the box is 2000km away, which is currently virtually unreachable due to corona Apr 09 13:32:37 Monkeh: that's why I like raid1c3 for metadata but to have that I'll have to introduce a ton of other new bugs too :P Apr 09 13:32:39 stintel: is #bcachefs on OFTC where it's being discussed? I hang out in #kernelnewbies there Apr 09 13:32:52 dansan: sorry, it's #bcache my bad Apr 09 13:32:57 ty@! Apr 09 13:33:16 stintel: I've reverted to ext4, may ponder zfs. Apr 09 13:33:48 well my head has difficulties switching back to something that doesn't support checksumming and reflink etc Apr 09 13:34:04 I've got the same sort of bad taste reiserfs gave me Apr 09 13:34:35 Sorry, murderfs Apr 09 13:34:36 ok, I need to write a hack script to strip these executables, build my firmware and debug this program I wrote Apr 09 13:34:37 :) Apr 09 13:34:42 Monkeh: exactly! Apr 09 13:34:52 I abandon it after that Apr 09 13:34:53 * Borromini has his laptop's root on btrfs and snapshotting through snapper Apr 09 13:35:01 I'm actually looking into gitlab ci to set up automated bcachefs testing Apr 09 13:35:24 dansan: I went earlier after it exploded violently from being created on a non-zeroed drive Apr 09 13:35:24 stintel: the cicd team is a bunch of fking clowns, I'm sorry to say Apr 09 13:35:26 this would reduce the time Kent has to spend on that himself Apr 09 13:35:34 dansan: gitlab ci/cd team you mean ? Apr 09 13:35:41 stintel: yeah Apr 09 13:35:48 well I've been trying a bit since last night, it's frustrating Apr 09 13:36:15 stintel: I wasted SO much time trying to get a simple fix in. They have hired some very inexperienced kids to maintain, but they don't KNOW they are kids and trying to tell ancient programmers like me stupid sh*t Apr 09 13:36:18 I have to run my own runner because the ktest thing Kent published on github runs tests in a VM and that doesn't fly well with docker Apr 09 13:36:30 dansan: well, it's ruby. says enough for me Apr 09 13:36:42 oh!! Apr 09 13:36:52 wait, I'm thinking of the gitlab-runner component -- that's Go Apr 09 13:36:59 ah Apr 09 13:37:06 well similar hipsterism involved ;) Apr 09 13:37:23 I wasted at least 60 hours trying to get a fucking patch through and finally told them I won't work on their project Apr 09 13:37:48 goddamn idiot wouldn't take polite education Apr 09 13:37:49 anyway, my problem now is that it generates dynamic project dir: export CI_PROJECT_DIR=$'\''/srv/glr/builds/oCG4kmyz/0/stintel/bcachefs'\'' Apr 09 13:38:17 wait, that was my attempt to override it Apr 09 13:38:27 it creates a dir /srv/glr/builds/oCG4kmyz/0/stintel/bcachefs.tmp Apr 09 13:38:34 and then tries to cd in /srv/glr/builds/oCG4kmyz/0/stintel/bcachefs Apr 09 13:38:40 which does not exist Apr 09 13:38:43 Ray Paik -- that's his name. goddamn clown Apr 09 13:39:03 guess I could look at the code Apr 09 13:39:06 https://gitlab.com/gitlab-org/gitlab-runner/-/merge_requests/1211 Apr 09 13:39:10 lol! Apr 09 13:39:13 ok, I'll shut up now Apr 09 13:39:14 also the curl -L .... | sudo bash is not very confidence insprining Apr 09 13:39:24 (instructions to install the runner) Apr 09 13:39:38 good thing we have throw-away VMs and stuff so I can somewhat live with that way of installing Apr 09 13:43:27 wait, I thought staging_dir/target-mipsel_24kc_musl/root-ramips is where files were staged prior to generating squashfs, am I mistaken? Apr 09 13:44:36 oooh, maybe that's staging_dir/target-mipsel_24kc_musl? Apr 09 13:46:16 Oh, I see! So build_dir/target-xxx/root- is where it's staged to build squashfs and staging_dir/target-xxx/root- is everythning, rather it's built as a module or not Apr 09 14:05:41 dansan: https://forum.gitlab.com/t/gitlab-runner-as-non-root-does-not-clone-repo/21935/5 *facepalm* Apr 09 14:06:21 stintel: yeah, unfortunately they're a bunch of fking clowns. And I really don't like that because I want to support some competition to m$ Apr 09 14:09:07 oh well at least I'm one step further now Apr 09 14:10:22 and if I can use gitlab.com with a local runner instead of maintaining an entire OVH CDS (which is a pita on its own) to do my autobuilds, it could do for now Apr 09 14:13:06 stintel: are you happy with ovh? Apr 09 14:13:54 Borromini: it's not exactly user-friendly but I made a workflow that autobuilds image for most of my devices, I even had staged auto-sysupgrade in it for my RPi0W's Apr 09 14:14:18 Borromini: and staged automated updates of my gentoo systems (binhost first, then the hosts that grab binpkgs from the binhost) Apr 09 14:15:51 neat Apr 09 14:16:26 and this is really important because doing all that stuff manually takes so much time I rarely get to any real work anymore Apr 09 14:17:23 :) Apr 09 14:17:31 well you could switch distros ;) Apr 09 14:17:57 thanks but no thanks. I got so used to the flexibility that any other distro isn't going to work for me anymore Apr 09 14:18:47 hehehe Apr 09 14:20:10 last time I worked for a customer with Ubuntu I ended up rebuilding source packages to fix bugs or update versions (to fix bugs) Apr 09 14:20:27 if I really have to mess with rebuilding stuff like that, gentoo is just much more friendly for that Apr 09 14:20:37 :) Apr 09 14:20:50 fair enough. i went down that rabbit hole with arch though and i'm not doing that anymore Apr 09 15:59:46 well, I'm really thankful for GitLab, saved me a _lot_ of time already, if something goes wrong, I've sources at hand Apr 09 16:47:07 ynezz: I'n not going to argue about the usefulness, but using it so far is proving to be a pita Apr 09 17:53:32 ynezz: I know it's a pitb but could you please merge this last JSON patch before it is activated by the buildbots? Apr 09 18:03:15 anyone know if any ath11k APs will have upstream OpenWrt support anytime soon? Apr 09 18:55:32 christ that gitlab is a pos Apr 09 19:12:29 greearb: the new ipq807x target currently does not support any real device yet - and prices are a bit difficult. there seems to be some interest in the xiaomi ax3600 (just because it's the cheapest of the bunch), probably down the road also the Netgear routers. but I don't know of anyone working on any device so far Apr 09 19:13:55 ath11k isn't packaged up so far either Apr 09 19:13:56 I can probably get free hardware for someone who has time to work on this, maybe some budget too. Apr 09 19:14:20 and I think the ipq807x target also misses quite some essential drivers beyond just ath11k Apr 09 19:20:04 @gch981213: the dir-860l router get "LZMA ERROR 1 - must RESET board to recover" from the k5.4 snapshot squashfs images ( https://forum.openwrt.org/t/optimized-build-for-the-d-link-dir-860l/948/1083?u=xabolcs ). a local build without the "CONFIG_ALL_KMODS=y" fixes the problem, but lowering the dictionary bits doesn't. could you run-check some other mt7621 board? thanks! Apr 09 19:21:17 it would be great if ipq807x would take off, it's great hardware from what I've seen so far (and wireless throughput is really amazing) Apr 09 19:23:08 my ipq807x device is to arrive tomorrow Apr 09 19:23:18 which one? Apr 09 19:23:26 Huawei AP7060DN Apr 09 19:23:34 ah, nice thing Apr 09 19:23:48 yeah, 10GbE PoE-PD and console port Apr 09 19:24:00 I just hope the 256MB RAM listed on techinfodepot is incorrect :) Apr 09 19:25:04 it's shocking to see wired ethernet being the limiting factor, rather than the wireless link (I got 931 MBit/s between qcn5054 and BCM43684) Apr 09 19:25:29 yeah well that's why such a device only makes sense with >1GbE uplink imo Apr 09 19:25:44 256 MB RAM sound unlikely, but yeah, for an AP it might be... most other devices seems to have 1 GB RAM Apr 09 19:26:22 puh and 30W of maximum power... that is a heating plate not an AP Apr 09 19:26:26 well we'll see. it's a tad bit expensive to immediately open it up and start messing with it but I'm going to try anywa Apr 09 19:26:39 decke[m]: expect around 20 watts idle Apr 09 19:27:14 ...and active cooling on many devices Apr 09 19:27:22 nice - have just bought an infrared heater with 950W ... I should have bought a few of these AP instead ... Apr 09 19:27:52 lol Apr 09 19:28:35 active cooling on an AP? no comment ... sounds like 1st gen hardware Apr 09 19:29:15 the wireless cards get hotter than the SOC itself Apr 09 19:46:38 xa Apr 09 19:46:39 x Apr 09 19:46:43 sorry! Apr 09 19:46:58 tab completion for names obviously doesn't work when people are gone >_> Apr 09 19:50:42 :D Apr 09 19:54:15 Thanks I did not even know that tab completion for nameswas a thing. Apr 09 19:54:21 names* Apr 09 19:55:00 Borromini: stintel Apr 09 19:55:06 * Tapper grins Apr 09 19:55:08 heh :) Apr 09 19:55:10 Apr 09 19:55:28 Tapper: haha Apr 09 19:55:32 happy to help lol. Apr 09 19:55:37 doesn't seem like it would be useful with a screen reader Apr 09 19:55:51 night guys! Apr 09 19:56:59 DonkeyHotei: it is because things I tipe ore in one tab and what you write back is in a separate tab. When I want to coppy and paste then I have to crontal tab be tween them. Apr 09 19:57:13 Control* Apr 09 19:57:43 bor later. Apr 09 19:57:58 look at that I broke it! Apr 09 19:58:37 O yeah he left! lol Apr 09 19:58:39 case in point Apr 09 19:59:22 DonkeyHotei: I did not hit the tab key that time. Apr 09 19:59:54 But you have to check what it has populated. Apr 09 20:01:41 seems to me that with a screen reader, checking what it populated would be more trouble than just typing it from scratch Apr 09 20:35:08 gch981213: MT7621-based Netgear R6220 won't boot (nand-based). Unfortunately, I don't have serial access to that device, sorry. Apr 09 20:36:57 Tapper: what screen reader are you using? Apr 09 20:38:52 mrkiko nvda mate Apr 09 20:42:29 Tapper: great Apr 09 20:44:20 Tapper: I never heard about Instantbird Apr 09 20:44:25 I knew miranda existed Apr 09 21:20:04 mrkiko: maybe blocktrron has a serial on a r6220 and can help Apr 09 21:32:29 https://bugs.openwrt.org/index.php?do=details&task_id=2986 Apr 09 21:32:45 so the initial feedback for DSA seems to be a bit underwhelming :) Apr 09 21:33:28 I thought tagging on DSA is just bridging port netdevs together? Apr 09 21:33:39 e.g. lan1.100 Apr 09 21:33:51 shouldn't that automatically speak vlan 100 on lan port 1 ? Apr 09 23:14:55 can someone with C knowledge please explain me this ubus function? https://git.openwrt.org/?p=project/usign.git;a=blob;f=main.c;h=ef47b282a7ccec7f02d8d42f927f2239fffae467;hb=HEAD#l70 Apr 09 23:15:03 *usign Apr 09 23:17:55 it composes a 64-bit value in cpu byte order from 8 individual octets in network byte order Apr 09 23:18:47 or were you asking HOW it does this? Apr 09 23:23:32 No that answered the question, thank you Apr 09 23:23:54 I reimplemented it with a for loop :S Apr 09 23:30:32 shouldn't there just be something like a ntohll() function? Apr 09 23:33:18 this is openwrt, it should be untohll()... ;) Apr 09 23:36:50 shouldn't be openwrt-specific Apr 09 23:37:06 gcc should just have something Apr 09 23:39:40 https://github.com/aperezdc/signify/pull/29 Apr 09 23:39:55 whoever is good at C (not me) please give me some advise of the code Apr 09 23:40:10 I ported signify over to OpenWrt and it works, however this one feature is currently missing Apr 09 23:40:32 DonkeyHotei & karlp if you have a few minutes please help out Apr 09 23:44:26 you are reading base64-decoded data directly into a struct on the stack, which is just asking for a buffer overflow Apr 09 23:49:38 DonkeyHotei: please comment Apr 09 23:50:10 DonkeyHotei: was just joking :) Apr 09 23:50:30 karlp: ? Apr 09 23:50:44 about the u* versions of things, nothing more than that Apr 09 23:51:43 oh. Apr 09 23:58:09 DonkeyHotei: thanks! I though i just followed the same steps as the rest of the code... Apr 09 23:58:21 I don't see any other verification at other parts Apr 09 23:58:51 does the rest of the code rely on base64 too? Apr 09 23:59:52 DonkeyHotei: I think everything yea Apr 10 00:09:58 ok, looks like other code takes care of that, but you do use printf where nothing else seems to. are you sure you want to write that to stdout? Apr 10 00:31:05 DonkeyHotei: Well that's how usign does it Apr 10 00:31:20 But know, I'm never sure about anything regarding C 🙂 Please give me an advise what to do Apr 10 00:41:19 if more than one file is specified, you emit a usage message and still use one of the files Apr 10 00:50:30 is that intentional? Apr 10 01:13:20 mrkiko: shouldn't r6220 got oversized kernel and fail to build in the first place? Apr 10 01:15:52 I'm considering building more drivers as modules to shrink main kernel. Some u-boot fails to extract big kernels and some nand devices got 2M kernel partition which doesn't fit. Apr 10 01:26:27 DonkeyHotei: After printing the usage the program exits, so no file is used Apr 10 01:32:13 aparcar[m]: the way i read the code, that is not so Apr 10 01:36:15 https://github.com/aperezdc/signify/blob/master/signify.c#L89 Apr 10 01:38:47 DonkeyHotei: So if there are more than one files provided, it prints the usage and quits Apr 10 01:39:01 at least that's my understanding Apr 10 01:40:36 missed that, sorry Apr 10 01:47:09 hrm interesting Apr 10 01:47:56 With my Archer C7v2, I could get wireguard working. Same config doesn't work on DSA Turris Omnia Apr 10 01:47:59 hmmmmmmmmm Apr 10 02:01:29 DonkeyHotei: np, tanks for checking! Apr 10 02:31:24 mrkiko: nm. I just noticed that r6220 got 4M kernel partition. Apr 10 02:35:16 jow: That vlan config is supposed to work. But someone captured packets and noticed that it receives vlan packets just fine but vlan tag is missing when sending packets. Apr 10 02:38:03 jow: same functionality can also be achieved using bridge vlan filter and vlan filter works fine. but we have no uci config for this yet. Apr 10 02:44:34 gch981213: curious, have you gotten wireguard with dsa working? Apr 10 02:47:40 I'm also curious about your SPI frequency comment on the mailing list: "@dengqf6 80MHz and 75MHz both rounds down to 73.33MHz anyway. There's no difference." Apr 10 02:49:09 git grep spi-max | grep -c 80000 returns 10 for the ramips directory Apr 10 02:51:22 aparcar[m]: !!seckeyfile + !!pubkeyfile + !!sigfile != 1 Apr 10 02:51:28 those are a lot of !! :) Apr 10 02:58:55 mangix: I think it is some magic to make thee values to bools **** ENDING LOGGING AT Fri Apr 10 02:59:57 2020