**** BEGIN LOGGING AT Thu Feb 25 02:59:57 2021 Feb 25 03:18:52 Well, I can say that the octeon 5.10 kernel I made builds, but panics :p I'm back to rebuilding Feb 25 03:42:02 it builds, ship it ;) Feb 25 03:43:48 ;p I went back and found all sorts of things that weren't turned on in the kernel Feb 25 03:44:09 So we will try again Feb 25 04:40:48 5.10 for octeon more or less is the same as 5.4. needs vitesse and some early edgerouter board definitions to be dropped since they're upstreamed Feb 25 04:45:04 build #10 of lantiq/xway is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/lantiq%2Fxway/builds/10 Feb 25 06:45:35 >KGB-1< https://tests.reproducible-builds.org/openwrt/openwrt_ar71xx.html has been updated. (99.1% images and 98.4% packages reproducible in our current test framework.) Feb 25 08:24:58 Grommish: for reference, 5.10 on malta works fine: https://github.com/openwrt/openwrt/pull/3881 Feb 25 08:26:18 guidosarducci: Thanks :) I think it's an issue with the PCIe bus not being enabled,but, question if you've got a min. Feb 25 08:26:36 Grommish: sure Feb 25 08:26:54 Any idea why OpenWrt woudl refuse to service DHCP requests on the LAN unless the WAN interface is up and linked? Feb 25 08:27:12 Or better, how to keep it from doing it Feb 25 08:27:38 My build space isn't near a network drop, so I do't have a WAN to give it ;/ Feb 25 08:27:46 Grommish: so... WAN unplugged == no DHCP LAN served? Feb 25 08:27:55 Right Feb 25 08:28:48 It's almost like dnsmasq is not working unless WAN is up Feb 25 08:28:50 Grommish: what have you checked so far? Does dnsmasq run? Is it listening on expected port/interface? Feb 25 08:29:14 Grommish: and is this master, 21.02? Feb 25 08:30:56 Sorry, my contact just fell out.. Snapshot from today Feb 25 08:31:05 Grommish: are you able to manually set IP and connect to LAN, or flying blind? I'll let you respond now ;-) Feb 25 08:31:37 Unless WAN is up, it's like the br-lan doesn't exist Feb 25 08:31:42 doesn't respond to pings Feb 25 08:32:23 Grommish: so you're blind. Can you connect via a wireless interface, or serial port to see into the system? Feb 25 08:32:30 I just need move it and get a good image on it, I just don't have a physical cable to feed it Feb 25 08:32:38 I've got a rollover console Feb 25 08:32:44 no wifi on the device Feb 25 08:32:57 morning guys Feb 25 08:33:08 Hey Borromini.. I bricked the shield ;p Feb 25 08:33:20 5.10 I build didn't have pcie bus enabled Feb 25 08:34:21 guidosarducci is tryingto help me figure out why dnsmasq doesn't seem to work on LAN unless the WAN port is up Feb 25 08:34:29 Grommish: to confirm, if you plug something into WAN everything works? I'm wondering if the LAN is just borked due to your PCIE setup... Feb 25 08:34:43 Grommish: that sucks. Feb 25 08:34:45 oh :( Feb 25 08:34:52 the master builds fine for 5.4 Feb 25 08:35:04 and thats the ramdisk I've got loaded Feb 25 08:35:07 if it helps any, i bricked my brand new gs108t v3 within minutes of unboxing xD Feb 25 08:35:25 * Borromini gives the installation constructions the side eye Feb 25 08:35:32 guidosarducci I'd have to go test it by moving it :p hich is what I was trying to avoid hahah Feb 25 08:35:50 It's ok though I'll go test and then tftp if I have to Feb 25 08:36:25 I just work out of a part of the house that doesn't have network, so I wifi only Feb 25 08:36:36 Grommish: I also wonder if something with the switch config got mangled. Maybe the concept of what's link-up is confused. Hard to say without know what all you changed. Feb 25 08:36:55 doesn't have a switch interface Feb 25 08:37:09 Grommish: maybe also port capture during startup, see if failsafe works too? Feb 25 08:37:11 just eth0, eth1, eth2, eth1/2 are bridged br-lan and eth0 is wan Feb 25 08:37:41 It Feb 25 08:38:07 Grommish: OK, I don't know the Shield, assumed *some* switch config was needed. Feb 25 08:38:09 its just frustrating because I only need to get it to work without the wan to flash hehe Feb 25 08:38:31 I don't have swconfig installed Feb 25 08:38:38 It doesn't ahve a switch chip Feb 25 08:38:56 build #15 of ath25/generic is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/ath25%2Fgeneric/builds/15 Feb 25 08:39:08 Grommish: can you try something silly for now and plug a lan into the WAN... :-( Feb 25 08:39:52 You mean, move th device? sure.. That isn't an issue, I was just being lazy and seeing if any one knew why it did it hehe Feb 25 08:40:59 Grommish: oh, I missed the location issue. No, that behaviour not normal... Feb 25 08:41:05 because my device, to flash, just have to mount the fat partition and move the .bin file into it directly Feb 25 08:41:23 so, if I can get ssh/scp to work, it's fixed Feb 25 08:41:50 but it won't dole out an IP and acts like it's not there unless wan has a link *shrug8 Feb 25 08:42:28 but I have console so I can gist anything you can think of looking at Feb 25 08:49:57 Grommish: just catching up.. if you have a console, perhaps try transferring your image that way? Without ssh/scp? Feb 25 08:50:45 Win10 for the main host OS.. I rebuilt the Microsoft kernel for WSL2 for USB but I've not got it working yet Feb 25 08:51:06 KNow a term that has kerrmit? Feb 25 08:51:30 I use minicom in ubuntu, but.. eh.. terraterm looked promising Feb 25 08:52:33 'morning! Feb 25 08:53:37 Quick question: how does configure a PVID with UCI, on DSA? Feb 25 08:53:59 *how does one Feb 25 08:55:07 I accept "one doesn't yet" for an answer… :P Feb 25 08:55:36 Grommish: but I doubt you have those utilities (xyzmodem) on your shield, right? I'm thinking you send an ascii-encoded image and decode on the SHield. let me think... Feb 25 08:55:51 I'm sure it has kermit Feb 25 08:55:59 I'm almost sure it have ymodem Feb 25 08:56:51 http://www.kermit.wwarthen.com/Download.htm < this one has standalone kermit for w10 Feb 25 08:57:14 * Borromini used ckermit on debian last time Feb 25 08:58:25 Yeah, I've got loadb and loady Feb 25 08:58:26 Grommish: so to be clear, your OWRT image on your shield right now includes kermit/zmoded/etc utility? Feb 25 08:58:45 No my uboot does Feb 25 08:58:59 So I can just throw up a raminitfs image to the fat partition Feb 25 09:00:23 Grommish: Ohhhh! I see know... I was thinking OWRT. You're set then... Feb 25 09:00:29 https://openwrt.org/toh/itus_networks/itus_networks_shield#recovery Feb 25 09:00:30 neat, more recent wireguard backport for 5.4 :) Feb 25 09:00:37 My debrick I wrote Feb 25 09:00:43 because I need it often enough Feb 25 09:04:08 Grommish: cool, it's nice they make it easy to update even uboot. Feb 25 09:05:47 I can't update the uboot, unfortuantely. Apparently the OED is not forthcoming Feb 25 09:06:02 OED? Feb 25 09:06:16 Sorry..ODM Feb 25 09:06:20 ok Feb 25 09:06:24 I'm workign with only one contact Feb 25 09:06:37 Ok.. So I'm transfering ymodem @ 4k/sec Feb 25 09:06:42 I don't miss those days Feb 25 09:07:48 Grommish: those days did teach you plan better since you didn't want to screw up and waste time redoing things... Feb 25 09:08:24 It also made it funny to uuencode a large binary as your finger reply Feb 25 09:08:39 :D Feb 25 09:08:42 :-) Feb 25 09:09:06 Wow… memories from the IRC in 1999. Feb 25 09:10:04 I only used it to find gopher addresses Feb 25 09:10:39 But I do IRC much Feb 25 09:10:46 err never did do Feb 25 09:12:03 Yeah, me neither. I was more interested in the web than in chatting… :) Feb 25 09:14:31 Once I get the recovery properly set up, i'll see about that 5.10 kernel again Feb 25 09:37:41 guidosarducci: You know, we can also build wolfSSL with MIPS16… ;) Feb 25 09:40:31 it's interesting how many assumptions break when building a linux C++ project on Windows Feb 25 09:42:07 mangix: like WSL? or like MS Code IDE? Feb 25 09:42:15 "What even Bill Gates can't buy? An stable operating system" ;D Feb 25 09:42:26 mangix: I have terrifying memories of Visual C++ 6.0, when I was at the university… Feb 25 09:42:58 Grommish: MinGW Feb 25 09:43:39 well, I'm sure there is a reason Feb 25 09:44:42 funniest comment i've heard about Bill Gates: what's he doing fighting COVID when he can't even fight viruses on his OS. Feb 25 09:47:55 mangix: [Win95] "32 bit extensions and a graphical shell for a 16 bit patch to an 8 bit operating system originally coded for a 4 bit microprocessor, written by a 2 bit company, that can't stand 1 bit of competition." Feb 25 09:49:30 lol **** BEGIN LOGGING AT Thu Feb 25 09:56:55 2021 Feb 25 09:57:45 build #10 of lantiq/ase is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/lantiq%2Fase/builds/10 blamelist: David Bauer , Daniel Golle , Christian Lamparter , Sander Vanheule , Stijn Segers Feb 25 09:57:45 , Eneas U de Queiroz Feb 25 10:03:27 guidosarducci: Still around? This is getting odder and odder Feb 25 11:02:10 build #788 of lantiq/ase is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/lantiq%2Fase/builds/788 Feb 25 12:20:39 ooh, interesting. ubnt-erx is kernel panic'ing during boot Feb 25 12:20:54 while dir860l on the commit is fine Feb 25 12:22:04 russell--: 5.10? Feb 25 12:30:55 no Feb 25 12:31:29 erx is ramips/mt7621 which i don't think has a 5.10 testing kernel yet Feb 25 12:33:18 I know it's mt7621, that's why I asked… I guess I'll defer playing with the Xiaomi for a while. :P Feb 25 12:34:13 Yeah, ramips is still at 5.4. Feb 25 12:42:04 it was fine as of r15817-d02088762a Feb 25 12:42:36 not fine aas of r15961-f0818706eb Feb 25 13:00:06 and, after the kernel bump to 5.4.100, problem gone again Feb 25 13:09:22 That's weird. So it's working at 5.4.100? Feb 25 13:09:30 * rsalvaterra wonders what changed. Feb 25 13:10:32 Pretty much… nothing relevant…? https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/log/?h=linux-5.4.y Feb 25 13:10:40 Just put the same changelog that lenovo did on one of bios updates... exactly one line in official changelog "updated BIOS version" Feb 25 13:16:10 build #741 of ramips/mt76x8 is complete: Failure [failed gcc] Build details are at http://buildbot.openwrt.org/master/images/builders/ramips%2Fmt76x8/builds/741 blamelist: Oskari Lemmela , Kabuli Chana , ?lvaro Fern?ndez Rojas , David Bauer , Daniel Golle Feb 25 13:16:10 , Adrian Schmutzler , Michael Pratt Feb 25 13:22:33 build #15 of bcm63xx/smp is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/bcm63xx%2Fsmp/builds/15 blamelist: Sander Vanheule , Stijn Segers , Eneas U de Queiroz , Christian Lamparter , David Feb 25 13:22:33 Bauer Feb 25 13:28:46 Build-testing the big-a** wireguard pull (ath79)… :) Feb 25 13:35:01 russell--: in my experience, sysupgrades on er-x are unreliable... Feb 25 13:35:39 rusell--: maybe it only works with kernelversion%2=0? Feb 25 13:35:48 https://bugs.openwrt.org/index.php?do=details&task_id=2713 Feb 25 13:36:00 https://bugs.openwrt.org/index.php?do=details&task_id=3206 Feb 25 13:42:32 zorun: i don't remember having problems before Feb 25 13:43:41 i've been updating them periodically remotely (although, admittedly not recently) Feb 25 13:50:24 reflashed the new working one, that boots okay too Feb 25 13:50:41 going to roll back and see if i can replicate the failure Feb 25 13:51:22 build #14 of apm821xx/nand is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/apm821xx%2Fnand/builds/14 blamelist: Sander Vanheule , Stijn Segers , Eneas U de Queiroz , Christian Lamparter , Feb 25 13:51:22 David Bauer Feb 25 14:01:30 5.4.99 (the commit before the kernel bump to .100) panics on boot Feb 25 14:04:11 https://paste.debian.net/1186919/ Feb 25 14:04:24 was the same stack trace as before Feb 25 14:05:03 but there is no change in .100 that is even remotely related ... Feb 25 14:07:29 back to r15964-4b92663f7a and it boots fine again Feb 25 14:10:33 build #782 of mediatek/mt7622 is complete: Failure [failed gcc] Build details are at http://buildbot.openwrt.org/master/images/builders/mediatek%2Fmt7622/builds/782 blamelist: Oskari Lemmela , Kabuli Chana , ?lvaro Fern?ndez Rojas , David Bauer , Daniel Golle Feb 25 14:10:33 , John Audia , Adrian Schmutzler Feb 25 14:12:58 * russell-- tries a vanilla config on the 99 commit, sees if maybe it's something in his custom config Feb 25 14:16:29 lipnitsk: ping Feb 25 14:18:35 yep, even with a vanilla config, panics Feb 25 14:19:01 well, .100 should allow relatively cheap bisecting Feb 25 14:22:12 zx2c4: ping Feb 25 14:24:34 The WireGuard pull request drops the wireguard metapackage in the VPN submenu… Feb 25 14:24:42 another theory was that sysupgrading from initramfs was somehow different than sysupgrading from a flash-booted firmware, but no difference Feb 25 14:24:59 … I don't think this is intended, is it? Feb 25 14:26:50 russell--: could also be a compression or check-summing related bug which triggers at a certain kernel size, and even just increasing the version string by 1 char was enough to fix it...? Feb 25 14:27:22 coming back to my modulo again ;-) Feb 25 14:27:23 r15964-4b92663f7a boots fine Feb 25 14:27:38 with a vanilla config Feb 25 14:27:54 maybe someone with an erx wants to try to duplicate this Feb 25 14:29:40 did you try to disable kernel config features to see if it still hangs ? Feb 25 14:30:07 given what i went through on MIPS64 with what I still blame on compiler weirdness I'm surprised that MIPS64 Linux is stable at all Feb 25 14:31:01 * dangole is talking about https://git.openwrt.org/?p=project/procd.git;a=commit;h=33b799b94c38fd2d3961b465a8114e384573a6d6 Feb 25 14:31:12 dangole: There's a lot of recent activity around MIPS64, nowadays, especially in order to make old SGi workstations work. ;) Feb 25 14:31:41 rsalvaterra: in Linux, yes. But GCC? Feb 25 14:32:26 Hmm… I would expect the Linux guys to scream at the GCC people when they find bugs in the compiler… but it may be just wishful thinking. Feb 25 14:32:38 i mean, look at the diff of that commit, that was the result two days lost with gdb in qemu-mips64, stepping through every instruction... Feb 25 14:33:38 Yeah, definitely demoralising… Feb 25 14:33:40 is " Status: , Cause and PrId" the same on certain bug classes ? Feb 25 14:33:45 (ok, you have to look at it in context to see how horrible this is) Feb 25 14:33:56 prid seems hardware Feb 25 14:34:01 but status and cause ? Feb 25 14:34:15 dangole: Have you reported it upstream? Feb 25 14:35:25 rsalvaterra: to GCC? maybe i should. i didn't really work on a more easy way to trigger it than compiling this exact piece of code (as my intention was to make that run, first of all) Feb 25 14:36:59 dangole: I found a bug in binutils (building the kernel in ARM Thumb-2 mode), reported it upstream and it was fixed very quickly. I don't know how responsive GCC is, though… Feb 25 14:37:18 with GCC and Linux it might work to switch from -Os to -O2 to test "robustness" of compiler generated code Feb 25 14:38:45 dangole: Does it take a lot of context to trigger the bug? It would be great to provide them with a basic test case. Feb 25 14:39:27 build #13 of lantiq/xway_legacy is complete: Exception [exception gitfresh] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/lantiq%2Fxway_legacy/builds/13 blamelist: David Bauer , Christian Lamparter , Adrian Schmutzler , Sander Vanheule Feb 25 14:39:27 , Stijn Segers , Eneas U de Queiroz Feb 25 14:39:28 rsalvaterra: i guess i'd have to be able to at least give a name to what i'm seeing there. i'm not. and there is also that always present chance that it's a very hidden bug somewhere else, in musl for example, causing stack corruption, which would also explain that. the question then is just why it only shows symptoms on mips64. Feb 25 14:43:39 Well, glibc is probably tested much more often than musl on MIPS64… Feb 25 14:45:09 There's a lot of nastiness in old MIPS64 CPUs (R10k, por example). Feb 25 14:59:32 build #13 of mediatek/mt7622 is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/mediatek%2Fmt7622/builds/13 blamelist: David Bauer , Christian Lamparter , Adrian Schmutzler , Sander Vanheule Feb 25 14:59:32 , Stijn Segers , Eneas U de Queiroz Feb 25 15:15:36 rsalvaterra: does anything depend on the meta package? Feb 25 15:16:03 lipnitsk: Yes, the kmod-wireguard and wireguard-tools. Feb 25 15:17:41 Probably what would make more sense would be just to rename the wireguard-tools package to wireguard and make kmod-wireguard depend on it. Feb 25 15:17:56 rsalvaterra: it can't be kmod-wireguard and I'm pretty sure I checked tools before Feb 25 15:18:12 You don't want kernel modules depending on user space tools Feb 25 15:19:06 I thought the meta package is superfluous so I removed it Feb 25 15:19:26 Ok, but the point is, now the user must know he has to select both kmod-wireguard and wireguard-tools, in order to have a funcional system. Feb 25 15:19:47 Yes and I think that's by design Feb 25 15:19:48 Previously, the metapackage took care of it for him. Feb 25 15:20:11 luci-app-wireguard still brings in both Feb 25 15:20:29 For people who use LuCI. I don't. ;) Feb 25 15:21:30 There are use cases for having tools without the kernel module. That has been brought up before Feb 25 15:21:57 So, if it's by design and I'm just getting what I deserve, I won't complain. I just wanted to make sure. :) Feb 25 15:23:46 what is that use case (just out of curiosity?) Feb 25 15:23:53 it is just for being able to generate keys or something? Feb 25 15:24:10 build #660 of layerscape/armv7 is complete: Failure [failed gcc] Build details are at http://buildbot.openwrt.org/master/images/builders/layerscape%2Farmv7/builds/660 blamelist: Oskari Lemmela , Kabuli Chana , ?lvaro Fern?ndez Rojas , David Bauer , Daniel Golle Feb 25 15:24:10 , John Audia Feb 25 15:24:17 I mean, is it a relevant usecase to complicate packaging? Feb 25 15:26:25 karlp: yeah that's one that has been mentioned once or twice Feb 25 15:26:46 karlp: I understand the point. For normal users it's the same, because of LuCI. If you're dealing with the config and building yourself, you're expected to know what you're doing. Feb 25 15:28:01 lipnitsk: Either way, I'm extremely happy with the whole series. Nice work. :) Feb 25 15:28:01 It's a trivial change to make wg tools depend on kmod-wireguard, so if there is consensus that can still be done. I don't have a strong opinion on this, but I don't think we need the meta package just for this case Feb 25 15:28:57 dangole: did you ever find another trigger for that MIPS64 bug, or just ujail? I use malta a lot so wondering how much of a problem that could be. Feb 25 15:36:30 rsalvaterra: karlp some historical discussion - https://github.com/openwrt/luci/issues/855 Feb 25 15:43:29 lipnitsk: let's make the dependency happen, or reintroduce the meta package Feb 25 15:43:47 actually, lets just make them all interdependent Feb 25 15:44:06 the separation is silly and doesnt help any user anywhere Feb 25 15:47:34 build #16 of octeontx/generic is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/octeontx%2Fgeneric/builds/16 blamelist: Adrian Schmutzler Feb 25 15:52:36 zx2c4: probably best to just bring back the meta package to not reatrict tools only users and stick it with the tools makefile? I'll work on it. Feb 25 15:56:02 lipnitsk: metapackage seems awfully complicated to me and i dont think the reasons people had in 2016 really hold these days Feb 25 15:56:30 but if you'd rather not rock the boat, i understand Feb 25 15:57:12 Honestly I don't care, I was gonna make tools depend on it but somebody on the mailing list objected Feb 25 15:59:12 zx2c4 It sounds like a corner case especially with the module being in the kernel Feb 25 15:59:31 alright lets make tools depend on it. Feb 25 15:59:43 if people object, send them my way and ill argue Feb 25 16:01:55 zx2c4: karlp rsalvaterra are we okay with wireguard-tools as the name, or do we want something named 'wireguard' still? Feb 25 16:02:34 wireguard-tools.. stick with that Feb 25 16:02:49 luci depends on tools. tools depends on kmod Feb 25 16:10:56 what do I need to revert for gcc on mips64 to build? Feb 25 16:14:21 lipnitsk: I don't mind the status quo. Feb 25 16:15:23 I'll open a PR and see where it goes. Simpler is better, and my change will be dead simple too. Feb 25 16:15:42 can't always make everyone happy ;) Feb 25 16:16:13 and I think zx2c4 deserves to have last word here, if nothing else because he is the author of the thing. Feb 25 16:17:31 guidosarducci: i'm not doing much in mips64, the ujail bug was the frist reason for a while to setup qemu for testing and hunting it down. so i can't really tell. Feb 25 16:17:44 and I agree that removing obstacles and helping user adoption is the right approach Feb 25 16:18:33 build #16 of oxnas/ox820 is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/oxnas%2Fox820/builds/16 blamelist: Adrian Schmutzler Feb 25 16:32:01 zx2c4: I'm also very happy I can now reduce the total kernel size quite a bit just by having WireGuard built-in, which was impossible with compat. ;) Feb 25 16:36:36 rsalvaterra: karlp zx2c4 https://github.com/openwrt/openwrt/pull/3929 Feb 25 16:37:01 lipnitsk: ugh Feb 25 16:37:12 you just took the route that is sure to incur more pain and suffering Feb 25 16:37:31 new PR: more bickering :) Feb 25 16:37:33 lol Feb 25 16:37:40 I would have just stealthfully tucked that change into the other PR Feb 25 16:37:46 build #547 of bcm63xx/generic is complete: Failure [failed gcc] Build details are at http://buildbot.openwrt.org/master/images/builders/bcm63xx%2Fgeneric/builds/547 blamelist: Oskari Lemmela , Kabuli Chana , ?lvaro Fern?ndez Rojas , David Bauer , Daniel Golle Feb 25 16:37:46 , John Audia , Adrian Schmutzler Feb 25 16:37:46 what else? send it straight to blocktrron ? Feb 25 16:38:02 oh, you're right - blocktron has already sort of frozen that PR hasnt he Feb 25 16:38:03 but he already picked that stuff into his tree Feb 25 16:38:10 ahh alright then Feb 25 16:38:17 let the games begin then! Feb 25 16:39:33 speaking of which blocktrron - any headway with `git merge && git push`? Feb 25 16:42:28 dangole: Oh, I see. I rely on it a lot, and also curious to see who else uses malta, as well as trying to get help to review/merge a malta 5.10 PR #3881. Feb 25 16:45:09 rsalvaterra: what's the size delta you see with wireguard built-in? Feb 25 16:45:09 guidosarducci: I still haven't measured precisely, but I'm working on it. Dog-slow machine compiling. :P Feb 25 16:46:07 Grommish: did you sort out your image update and WAN/LAN issue? Feb 25 16:47:02 rsalvaterra: if you're in a measuring mood, I'd be curious how that size changes with GCC version. Just thinking about your earlier comment... Feb 25 16:47:44 guidosarducci: From 8 to 10? Feb 25 16:53:50 rsalvaterra: sure, well, anything showing older vs newer. What do you use regularly? Feb 25 16:55:02 guidosarducci: I've been on GCC 10 for months, but be prepared for a "big" surprise. Feb 25 16:55:32 The inliner became much more aggressive by default. Feb 25 16:55:46 rsalvaterra: let me guess, giant binaries with a tiny performance gain? Feb 25 16:56:36 I haven't noticed (or tried to measure) any performance differences. Feb 25 16:57:15 rsalvaterra: so it's just the prestige of GCC 10? :^) Feb 25 16:57:37 I wouldn't say "giant", but the nominal size of an image for my WDR3600 increased by about 200 kiB. Feb 25 16:58:44 guidosarducci: Meh. I could revert to the previous behaviour, there's a switch for that (-fno-inline, IIRC). :) Feb 25 16:59:40 rsalvaterra: was that from GCC 9 to 10? Funny, that amount is close to average increase I saw from major release to major release. Feb 25 17:00:06 guidosarducci: I skipped 9. Feb 25 17:00:15 build #16 of sunxi/cortexa53 is complete: Exception [exception gitfresh] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/sunxi%2Fcortexa53/builds/16 blamelist: Adrian Schmutzler Feb 25 17:00:52 rsalvaterra: Hmmm, I wonder if a strategy for some with small partitions is to use the latest kernel with oldest compiler. Feb 25 17:02:11 guidosarducci: The compiler code generation can always be tamed, if needed. ;) Feb 25 17:02:30 Note that I build at -O2, not -Os. Feb 25 17:02:46 rsalvaterra: GCC 10 is supposed to add basic eBPF support, so I may need to try it out. Feb 25 17:04:08 rsalvaterra: well, -O2, that explains it! :) Feb 25 17:04:30 It's -O2, not -Omg… that's for the Gentoo crowd. :P Feb 25 17:06:18 rsalvaterra: I dare you to submit a GCC patch for that option... April 1 is coming up. Feb 25 17:06:42 lmao Feb 25 17:52:39 blocktrron: sent you an email with a couple of changes Feb 25 17:57:48 build #16 of pistachio/generic is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/pistachio%2Fgeneric/builds/16 blamelist: Adrian Schmutzler Feb 25 18:12:45 lipnitsk: please add a description to your change Feb 25 18:13:01 It's not clear to me, what the rationale behind this chagne is Feb 25 18:13:30 the netsupport.mk change? Feb 25 18:13:31 will do. Feb 25 18:16:24 thanks Feb 25 18:16:45 I'm a wireguard-nobody so I have no clue for real Feb 25 18:16:47 :P Feb 25 18:21:45 blocktrron: done, it was more to match other netsupport.mk kmod packages Feb 25 18:28:09 pkgadd: thanks for the link I haven't see that pull reqeust before Feb 25 18:29:49 build #309 of mpc85xx/p1010 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/mpc85xx%2Fp1010/builds/309 Feb 25 18:39:58 build #16 of bcm27xx/bcm2711 is complete: Exception [exception gitfresh] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/bcm27xx%2Fbcm2711/builds/16 blamelist: Adrian Schmutzler Feb 25 18:59:31 build #16 of bcm53xx/generic is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/bcm53xx%2Fgeneric/builds/16 blamelist: Adrian Schmutzler Feb 25 19:22:46 adrianschmutzler: i don't like maintaining I'm afraid Feb 25 19:22:53 too many things to take care of Feb 25 19:23:13 i would handle some stuff faster on my own :| Feb 25 19:23:53 adrianschmutzler: that BCM5301X patches were not 100% backports, but copy of local tree, without even Signed-off-by Feb 25 19:24:05 they were mislabeled (v5.10 instead of v.511) Feb 25 19:24:15 there is more stuff to backport Feb 25 19:24:23 1 patch was outdated version Feb 25 19:24:27 they were not refreshed Feb 25 19:25:13 and 1 filename didn't match patch (git format-patch anyway) Feb 25 19:41:35 rmilecki: sorry, what are you talking about? Feb 25 19:42:53 I miss the context Feb 25 19:46:39 build #15 of omap/generic is complete: Exception [exception gitfresh] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/omap%2Fgeneric/builds/15 blamelist: Adrian Schmutzler Feb 25 19:48:50 do you try to bump bcm53xx? Feb 25 19:53:14 https://w1.fi/security/2021-1/ Feb 25 19:54:36 adrianschmutzler: bcm53xx: enhance support for Linksys Panamera aka EA9500 v1 or v1.1 Feb 25 19:54:50 i'm testing bcm53xx update on actual hw in a moment Feb 25 19:54:55 and I'll push dts update Feb 25 19:55:32 ahh, now I get it Feb 25 19:57:06 btw, does that platform have phy triggers for WiFi LEDs? Feb 25 19:57:38 So that we could replace wlan0 by something more generic not depending on the interface name Feb 25 20:01:49 just saw that this has multiple interfaces, but this links only to one. probably one should just drop the wifi led line Feb 25 20:03:35 https://w1.fi/security/2021-1/ Feb 25 20:04:16 adrianschmutzler: i'm not sure about WiFi LEDs Feb 25 20:04:25 Hauke: please update to hostapd 2.10 when it's available :p Feb 25 20:06:44 build #12 of mvebu/cortexa53 is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/mvebu%2Fcortexa53/builds/12 blamelist: Adrian Schmutzler Feb 25 20:14:26 build #15 of apm821xx/nand is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/apm821xx%2Fnand/builds/15 Feb 25 20:19:30 adrianschmutzler: do you want me to test the octeon 5.10 patch on something specific? can test on er-4, er-12 and er8-pro (the disabled one) Feb 25 20:25:40 lemmi: start with what would be easy to recover if it doesn't boot Feb 25 20:26:22 trying the disabled one would be interesting as well, i.e. whether it still just needs the one config symbol enabled to work Feb 25 20:27:09 but generally: all tests are welcome, more tests are better, but choose what you prefer and what time allows Feb 25 20:28:05 >KGB-1< https://tests.reproducible-builds.org/openwrt/openwrt_sunxi.html has been updated. (0% images and 98.2% packages reproducible in our current test framework.) Feb 25 20:28:15 they're all easy to test as they can be bootet from usb Feb 25 20:28:54 the most important test will be the first one, as that will confirm that it works at all Feb 25 20:29:18 ok. i'll look into it this week if i don't forget. Feb 25 20:30:13 there is always the chance that there was a change in kernel that did not surface during refresh. Feb 25 20:35:46 build #12 of bcm63xx/generic is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/bcm63xx%2Fgeneric/builds/12 blamelist: Adrian Schmutzler Feb 25 20:40:39 ich habe tatsächlich auch probleme mit zwei uplinks die 1043 sind Feb 25 20:40:39 schon länger, aber ich habe das jetzt nicht an der architektur fest gemacht Feb 25 20:40:39 beide v4 übrigens Feb 25 20:40:40 aber andere 1043v4 in vergleichbarer konfiguration sind ganz normal Feb 25 20:40:40 hmm, einer von den beiden läuft aber gerade auch wieder und Feb 25 20:40:40 sorry, wrong window Feb 25 20:51:17 rmilecki: commit reads like you really had lots of fun Feb 25 20:57:13 adrianschmutzler: ;) Feb 25 21:00:08 build #12 of ipq40xx/mikrotik is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/ipq40xx%2Fmikrotik/builds/12 blamelist: Adrian Schmutzler Feb 25 21:06:12 Hmm… guys? What causes hostapd to log AP-STA-POLL-OK messages? It only does this when I have a specific machine connected. Feb 25 21:09:07 rsalvaterra: chinese IoT stuff? :P Feb 25 21:15:35 Borromini: RTL8821AE Feb 25 21:25:44 build #12 of mpc85xx/p1010 is complete: Failure [failed gitcheckout] Build details are at http://buildbot.openwrt.org/openwrt-21.02/images/builders/mpc85xx%2Fp1010/builds/12 blamelist: Adrian Schmutzler Feb 25 21:26:14 ok :( Feb 25 21:26:38 i am seeing hostapd persistently fail for some reason on a wifi reload. it doesn't seem to get killed anymore on 21.02 Feb 25 21:29:07 well just tried an R6800 and works there, but on my EAP it doesn't. weird stuff. Feb 25 21:56:22 guys seeing 'interface already in use' - 'variable 'data' does not exist or is not an array/object' on 21.02 (doesn't happen on all devices though) Feb 25 21:56:27 https://paste.debian.net/1186986/ Feb 25 21:56:40 this is on a wifi down && wifi up ^^ Feb 25 21:57:03 killing the hostapd process manually will allow me to bring wifi up again, until then it just spews what's in that paste Feb 25 22:01:11 i checked if the radio paths were right, wifi config uses the same as in the config, so that shouldn't be the issue. I'm not seeing anything named 'data' in /etc/config/wireless, i assume that's referring to the hostapd or parsing bit or sth. Feb 25 22:16:59