**** BEGIN LOGGING AT Thu Jan 07 03:00:55 2021 Jan 07 03:41:07 mangix: https://buildmaster.aparcar.org/phase2/#/builders/32/builds/1if you're keen there are many more packages to fix :O Jan 07 03:41:48 I don't see any failures there Jan 07 03:51:14 aparcar[m]: if you want failures, set up an arch based builder. tons of failures there. Jan 07 04:03:36 mangix: na I'm fine, just saying in case you want some buildbot like testing Jan 07 04:10:49 mangix: https://github.com/openwrt/openwrt/pull/3742 this looks to good to be true? Jan 07 04:11:48 the commit is correct. cmake should not be attempted to be build with ccache as ccache itself depeneds on cmake Jan 07 04:11:55 *built Jan 07 04:12:56 it was a change that was missing in the ccache update commit Jan 07 04:13:20 it also happened to not fail for reasons explained in the commit Jan 07 05:51:00 >KGB-0< https://tests.reproducible-builds.org/openwrt/openwrt_x86.html has been updated. (100.0% images and 98.2% packages reproducible in our current test framework.) Jan 07 07:50:50 mangix: merged Jan 07 15:07:05 >KGB-1< https://tests.reproducible-builds.org/openwrt/openwrt_ar71xx.html has been updated. (99.2% images and 98.4% packages reproducible in our current test framework.) Jan 07 16:58:35 luci does not show wireless status correctly on master Jan 07 16:58:58 (commit ecded5b) Jan 07 16:59:20 Maybe it's related to iwinfo and rpcd updates? Jan 07 16:59:24 jow: Jan 07 17:17:39 dengqf6_: well, I didn't do the updates, I hope dangole tested LuCI before pushing them Jan 07 17:18:16 dengqf6_: but yes, it most likely is related to these updates as there's been no related changes to LuCI recently Jan 07 17:27:41 jow: we may need to trigger the buildbots to rebuild luci to update the dependency of the package Jan 07 18:00:44 jow: I've seen 2 reports on IRC from people who used LuCI to configure swconfig tagging and somehow after untagging an unrelated port the interface configuration was changed from e.g. eth0.1 to eth0 , and at least in one case it was verified that it wasn't the CPU port that was untagged according to /etc/board.json. Jan 07 18:12:32 Borromini: ping Jan 07 18:35:32 dwmw2_gone: ping Jan 07 18:51:45 rsalvaterra: pong Jan 07 18:52:37 Borromini: Looks like you're not alone. Results came early this morning, tested positive. :/ Jan 07 18:53:03 Oh, well… Jan 07 18:57:47 PaulFertser: needs a proper bug report Jan 07 18:58:29 rsalvaterra: sorry for nitpicking, but it's a SARS-CoV-2 test, I hope despite being positive you get few if any COVID-19 symptoms. Jan 07 18:59:07 jow: ok, what to include in that? Just board.json and the actions performed? Jan 07 18:59:27 would be a start, yes Jan 07 19:00:04 OK, I'll guide the reporter to do that, thanks! Jan 07 19:01:04 rsalvaterra: sorry to hear that - I hope you don't get any or few symptoms, and certainly not severe ones. Jan 07 19:02:26 ldir: A bit more fatigued today… still managing fine, thankfully. Jan 07 19:27:59 Anyone here know anything about HW packet timestamping on Ethernet interfaces? Jan 07 19:29:56 philipp64: I have second-hand knowledge after having people on my team previously who did that and talked about it. Jan 07 19:33:53 SwedeMike: trying to navigate my way through adding TAI-like timescales to SO_TIMESTAMP but it’s new territory for me. Jan 07 19:35:47 philipp64: we used it to very accurately measure how long it took for a packet to traverse the network (SLA probing). I don't know the terms you use so unfortunately I don't think I am of much help Jan 07 19:39:42 UTC is coordinated universal time… like GMT… what you’re used to seeing. It included leap-seconds. Jan 07 19:40:29 GMT is not a word to be used Jan 07 19:40:35 TAI is a linear, non-leaping atomic time. It counts SI’s (standard seconds) and nothing else… it doesn’t care about minutes, hours, days, years, or leap-seconds. Jan 07 19:40:51 PaulFertser: because? Jan 07 19:41:08 philipp64: it has no good defined meaning, it was always a mess. Jan 07 19:41:13 ah… did the brits get to keep exclusive use of it as part of Brexit? Jan 07 19:41:50 It's not used by the people working with time for many many years before brexit. Jan 07 19:43:19 It's easier to forget there was a word like that :) Jan 07 19:43:36 well, I could say “Zulu Time”… but unless you were in the US Army, that might not make sense either… Jan 07 19:45:31 Anyway, machines should use TAI internally because it resolves a lot of ambiguity. Indeed, even logging protocols should use non-leaping seconds (and milliseconds, where appropriate). Converting that to human-readable form (like local seconds, minutes, hours, days, months, years, timezone, etc) is a “presentation” issue to be handled by browsers. Jan 07 19:45:50 Agreed Jan 07 19:46:18 Given the machine has access to the information about the current amount of the leap seconds. Jan 07 19:46:20 Don’t even get me started on daylight saving time and what a pain it is to see the say time repeated twice in logs… Jan 07 19:49:04 set everything to UTC and you've won! Jan 07 19:49:23 that includes your watch and your wife. Jan 07 19:51:59 UTC can be problematic when leap seconds are added, that's the point Jan 07 20:01:22 Yup… Jan 07 20:01:50 And leap-seconds are completely irrelevant to machine time. They create problems that shouldn’t exist. Jan 07 20:11:28 philipp64: pong Jan 07 20:15:09 xdarklight: I did something with irq chip some time ago Jan 07 20:15:15 what do you want to do? Jan 07 20:16:33 Hauke: convert the Lantiq ICU irqchip to a hierarchical one. I think I have figured it out (mostly) in the meantime Jan 07 20:17:43 xdarklight: I think I looked into this some time ago Jan 07 20:17:57 noltari: ping Jan 07 20:18:15 xdarklight: the MIPS IRQ driver is alread there Jan 07 20:19:43 dwmw2_gone: hey David. per the above, I’m trying to add patches to Linux to make it more TAI-friendly, rather than UTC-centric… as someone who knows his way around the code base…. how many places does that entail? Jan 07 20:20:18 not sure; does the kernel really care? Jan 07 20:20:23 probably not a lot. Jan 07 20:20:58 Hauke: then I'll probably ask you for code-review :) Jan 07 20:24:28 xdarklight: https://github.com/hauke/linux/commits/lantiq-irq1 Jan 07 20:25:42 xdarklight: I do not know if it works and it still misses the eicu Jan 07 20:25:51 eiu Jan 07 20:26:00 adrianschmutzler: please give the commitcount patch a test Jan 07 20:26:25 xdarklight: I do not plan to contine on this now, so feel free to take this as a base Jan 07 20:26:58 Hauke: hah, I have something (partially) working already - I can use it to compare. thanks! :) Jan 07 20:27:55 dwmw2_gone: I guess the big difference is being able to select the timescale for SO_TIMESTAMP… Jan 07 20:51:46 dangole, wasn't able to pull before the revert, retrying now Jan 07 21:43:56 Hello, would anyone know what are my chances of getting openwrt on the unifi U6 Lite one day, if I were to buy one now? Jan 07 21:46:19 Antoine-: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=fb4d7a9680117a00721936c98ce41eeb2dea95c9 ? Jan 07 21:48:25 swalker: Oh great! Am I looking at the installation instruction? Does it mean it is already supported? :) Jan 07 21:52:31 swalker: Is it possible to install openwrt thanks to some exploit that could get patched in a future firmware release? Like, if I buy one in 6 months, will it probably still be possible to install openwrt on the U6-Lite? Jan 07 21:53:15 Antoine-: not seeing any other docs for it for now Jan 07 21:55:49 swalker: Ok :). Thanks a lot for the link! Jan 07 21:56:43 Antoine-: as you can see it has the installation instructions. Jan 07 21:57:32 what will or will not possible to do in the future depends on how well your crystal ball works. Jan 07 21:58:28 jow: any updates on ucode? Jan 07 21:59:53 Borromini: Ok, I thought we could already know some things. Thanks for clarifying! Jan 07 22:01:16 Antoine-: i don't know any ubiquiti devices that allow you to use the web interface to upgrade (i don't know if they have much of a web interface; i always flash openwrt first thing) Jan 07 22:01:28 i've set up three types of ubiquitis so far. Jan 07 22:04:04 Borromini: Did you do it through ssh or serial console? Jan 07 22:05:05 Antoine-: EdgeRouter 4 through serial, UniFi AC AP is TFTP I believe, EdgeRouter Lite just has an internal usb stick (!) you take out and put openwrt on Jan 07 22:22:45 Ok, thanks :) Jan 07 22:24:15 I'm really sorry if it's not the proper place to ask. The U6-Lite expects 48V, 0.32A. I have a POE-24-12W-G, that came with the UAP AC Lite. It provides 24V, 0.5A. Is it safe to use? Jan 07 22:24:58 no, that's ubiquiti's proprietary PoE stuff i think? Jan 07 22:25:26 Yes, it is Jan 07 22:25:26 48v is 802.3at i believe (PoE+) Jan 07 22:25:50 lots of older ubiquiti stuff is not PoE compliant i think Jan 07 22:26:17 Borromini: Yes, mine isn't. Won't power up with my 802.3at switch Jan 07 22:27:13 Well, then my POE-24-12W-G is a wasted piece of electronics :( Jan 07 22:27:17 Thanks for your reply :) Jan 07 22:27:46 well you might be able to get an injector? Jan 07 22:27:57 or does the U6 Lite come with one? no idea if they do. Jan 07 22:28:28 if anything else, you might sell the Ubiquiti switch if you're going to replace it, get a few bucks for it still :) Jan 07 22:29:13 The store says "In case a PoE injector is needed to power the device, compatible type can be purchased separately.", so I would guess it doesn't come with an injector Jan 07 22:30:07 I wonder who would need it since it isn't 802.3 compliant but you're right, I'll try to sell it :) Jan 07 22:37:52 lots of people into the ubiquiti ecosystem Jan 07 22:44:46 If I have 'mtd7: 00020000 00010000 "bs"', then my bs partition is accessible through /dev/mtdblock7, right? Jan 07 23:02:49 U6 Lite does not come with an injector Jan 07 23:04:04 In mine was a paper which explained it was for "saving the environment" Jan 07 23:05:18 It's apple for networking. They resemble the lightning vs. USB-C fiasko with their 802.3af vs. 24V Fries-your-device PoE. Jan 07 23:07:41 =) Jan 07 23:08:51 I hope buying their unifi6 is still a good option.. Jan 07 23:08:58 At least they are 802.3 compliant now Jan 07 23:10:31 It's solid Jan 07 23:13:56 I wished for a better upgrade-path, given the first revision of it's successor didn't have any 802.3af capabilities at all. Jan 07 23:14:38 Ok, nice :). I just need to wait for their poe injector to get back in stock I guess Jan 07 23:14:54 Just buy a used 802.3af injector from eBay Jan 07 23:15:07 no need to pay the Ubiquiti premium there Jan 07 23:15:21 a question about the "model" OpenWrt/LuCI reports: is it taken from the model value in the dts? Jan 07 23:15:30 yes Jan 07 23:15:54 i see lots of DEVICE_MODEL values in Makefiles, but that doesn't seem to override Jan 07 23:16:11 blocktrron: ok thanks. i suppose it's downright silly to introduce just a single dts to override that. Jan 07 23:16:29 Yes Jan 07 23:16:41 https://paste.debian.net/1180110/ < because this is giving me the heebiejeebies :P Jan 07 23:17:27 I was checking ebay but the ubiquiti poe injector is €8,16, which is a good price, even compared to ebay, isn't it? Jan 07 23:17:31 i guess the UBNT_E100 (CN5020p1.1-500-SCP) is coming straight from the upstream dts then Jan 07 23:17:43 ah, octeon Jan 07 23:18:00 this is a special case and done somewhere in the board init code Jan 07 23:18:43 blocktrron: ok, thanks. Jan 07 23:19:42 looks like target/linux/octeon/base-files/lib/preinit/01_sysinfo Jan 07 23:21:04 blocktrron: is there any objection to making that align with how other targets do it? Jan 07 23:21:24 e.g. boardname 'netgear,r6800' and model 'Netgear R6800'? Jan 07 23:21:42 this would require downstream patching Jan 07 23:22:02 you mean as in openwrt, or as in in my own tree? Jan 07 23:22:07 In OpenWrt Jan 07 23:22:26 ok. would a patch just modifying 01_sysinfo be accepted? Jan 07 23:22:35 Can't speak for Octeon Jan 07 23:22:40 ok. Jan 07 23:22:56 In the end, my opinion on those topics is to stick as close as upstream as possible Jan 07 23:23:37 understandable, lower the maintenance burden Jan 07 23:25:08 hmm, on a second look this might not involve downstream patching it all Jan 07 23:25:36 removing the write to /tmp/sysinfo/model should suffice Jan 07 23:26:09 ok. the EdgeRouter Lite has a DEVICE_MODEL in the Makefile, do you mean the 01_sysinfo script is overwriting that in /tmp/sysinfo/model? Jan 07 23:26:50 Makefile is only relevant for the build system (metadata JSON / menuconfig / ...) Jan 07 23:26:56 oh. Jan 07 23:28:19 So yeah, might be possible to change by adding a pretty-name to this file and saving this to tmpfs Jan 07 23:28:37 but no idea if this has side-effects elsewhere. Never had to work with octeon Jan 07 23:30:42 Borromini: DEVICE_MODEL is only relevant for the friendly name in make menuconfig Jan 07 23:31:05 and metadata JSON as stated above Jan 07 23:31:16 adrianschmutzler: ok, thanks Jan 07 23:32:23 and the /tmp/sysinfo stuff is set in 01_sysinfo, as you have found out already above Jan 07 23:32:51 for most of the other targets except octeon it's created directly from the DTS compatible Jan 07 23:33:23 like octeon's 01_sysinfo also does it for "UBNT_E300"*) Jan 07 23:33:47 where it actually does nothing so the generic sysinfo script can take over Jan 07 23:34:08 yeah, as per the comment Jan 07 23:34:33 adrianschmutzler: apologies for the realtek fuck-up, will look at it again tomorrow with fresh eyes. Jan 07 23:35:07 ah, wasn't aware that this is you. Jan 07 23:35:21 it is yes :) Jan 07 23:35:23 * Borromini hides Jan 07 23:35:34 adrianschmutzler: i did forget the DTSI indeed. Jan 07 23:35:56 however, i have switched the DTSI to the realtek macros Jan 07 23:36:03 should be easy to adjust actually, I considered doing it myself but then wasn't perfectly sure what should be put in the DTSI and what should be kept in the dts Jan 07 23:36:13 so it's much more compact than the original single-device dts Jan 07 23:36:36 but you e.g. also create the leds and keys in the dtsi but do not remove them in the dts etc. Jan 07 23:36:52 same for that gpio node Jan 07 23:36:55 ok, i'll take a look at that tomorrow Jan 07 23:36:57 * Borromini takes notes Jan 07 23:37:29 unless I'm overlooking some greater plan there Jan 07 23:37:47 no it should have been pretty much all in the dtsi. so i'll have a look at where it went sideways Jan 07 23:38:15 i flashed my gs1900-10hp actually with a newer build here and it went belly up, so that's plenty of proof :-/ Jan 07 23:38:54 well, without the dtsi include you won't have any partitions, as spi has been moved. Jan 07 23:39:10 yeah :P Jan 07 23:39:41 fresh eyes, tomorrow. Jan 07 23:39:43 goodnight! Jan 08 00:52:02 Antoine- Jan 08 00:52:23 Antoine-: You have one, yes? Could you please post photos on https://wikidevi.wi-cat.ru/Main_Page ? Jan 08 00:53:08 Internal photos if you'd like to take them. I am wondering what type of NOR is used, and the FCC page still has confidentiality for internal photos (https://fccid.io/SWX-U6LITE) Jan 08 00:54:26 If it's an 8-pin or 16-pin SOIC it is very likely that the fastest exploit will be to simply be to reflash U-Boot. Since it's an MT7621AT I don't believe it has any sort of on-chip bootloader like Qualcomm's IPQ series. Jan 08 00:55:28 The same is true of all new Ubiquiti Nanostation Locos -- you need to reflash using a Pomona 16-pin clip. Jan 08 00:56:39 I do see newer 256Mb SPI NOR coming in the 16-pin form factor which is promising. Jan 08 00:57:29 But seriously. A MT7915 in an affordable device with PoE capability is an absolute dream. Jan 08 01:14:38 I did not read the compatible closely enough. It is a mx25l25635f which is a 16-pin clippable chip. Great! **** ENDING LOGGING AT Fri Jan 08 02:59:57 2021