**** BEGIN LOGGING AT Thu Feb 27 02:59:58 2020 Feb 27 08:52:13 Could not generate file hash ~> http://downloads.openwrt.org/snapshots/faillogs/x86_64/packages/protobuf-c/compile.txt Feb 27 09:07:34 why doesn't buildbot throw error despite libprotobuf-c failing to compile Feb 27 09:09:15 package builds are non-fatal, if they would be, we would never have a "green build" Feb 27 09:09:48 afaik Feb 27 09:13:48 can someone recompile the package now ? or should I wait for another 24hrs to see what happens Feb 27 09:25:59 Hi people will some one post to the ML when we are macking the jump to k5.4 pleas? I would like to do one last k4.19 build to let things settle down Feb 27 09:33:29 mangix: point is avoiding libintl / libiconv runtime depends while providing minimal conversion support Feb 27 09:34:09 ashkan: stintel: it does for packages that are supposed to be included in the images, but anything extra is allowed to fail Feb 27 09:39:04 IGNORE_ERRORS=m Feb 27 09:39:58 KanjiMonster: thanks for elaborating Feb 27 09:40:11 ashkan: I don't think we can trigger a rebuild of a single package Feb 27 09:40:47 russell--: IGNORE_ERRORS="n m" - you also don't want to fail on packages that are only build time dependencies (so aren't selected as =m) Feb 27 09:42:14 btw that "Could not generate file hash" error started to appear on a lot of builds lately Feb 27 09:42:30 it is emitted by download.pl when the hash file is empty Feb 27 09:42:44 that hints at a possible problem with mkhash Feb 27 09:44:17 the current build is going to fail again Feb 27 09:45:25 wigyori: ping Feb 27 10:18:47 KanjiMonster: thanks! Feb 27 10:49:49 jow: does musl have libiconv separate? Feb 27 10:53:48 nope Feb 27 11:12:55 jow: why do we bother writing out the hash to a file just to read from it a few lines later, why not just directly read the output of the hash command? Feb 27 11:22:14 jow: your cgi-io update broke compilation on arc: http://downloads.openwrt.org/snapshots/faillogs/arc_archs/packages/cgi-io/compile.txt Feb 27 11:25:42 broke uclibc there then I guess? Feb 27 11:25:50 and arc the only user still? Feb 27 11:26:11 snps promised theyd' fix this, I take it they still haven't? Feb 27 11:26:54 KanjiMonster: iirc download.pl reads the stdout of wget/curl, hashes it block wise and writes it to mkhash's stdin as well as to the destination file Feb 27 11:27:34 KanjiMonster: reading both mkhash and wget output would reuqire mor complex logic including select/poll Feb 27 12:04:56 jow: while not directly related, trying to do both (writing the output file and creating the hash) seems to be like an unecessary micro optimization. letting curl/wget write to a file and then hash that file would allow sharing the checksumming code from the $mirror =~ s!/^file://!! path, something like https://pastebin.com/0dC07Hsg Feb 27 16:22:43 ynezz: Re: PR#2649, all of concerns have been expressed already, I don't have anything new to add. What I would like to see is a clear statement from one of Commiters that with *or* without migrating the interfaces names it won't go through. Feb 27 16:23:52 *it will *or* won't got through. Feb 27 17:06:53 if it was so easy, I would have merged it a long time ago ... Feb 27 17:10:38 I'd like to 'emulate' the procedures that a phase1 builtbot does but without actually creating a buildbot. Is there a list of commands that the buildbot uses to do the initial phase 1 build ? Feb 27 17:11:27 ldir: https://git.openwrt.org/?p=buildbot.git;a=blob;f=phase1/master.cfg;h=792f9b3a55a2d8ebae1917c56716cc0d699f9172;hb=HEAD Feb 27 17:12:16 https://git.openwrt.org/?p=buildbot.git;a=blob;f=phase1/master.cfg;h=792f9b3a55a2d8ebae1917c56716cc0d699f9172;hb=HEAD#l465 is where it starts Feb 27 17:15:11 ok thanks, I'll pull that apart - I basically want to build all targets under macos to see what falls out :-) Feb 27 17:15:29 :D Feb 27 18:24:57 I recently succeeded in running os x catalina in virtualbox Feb 27 18:25:08 was thinking about setting up a builder Feb 27 18:25:35 performance will probably be bad but still better than nothing Feb 27 19:43:53 karlp: ping Feb 27 20:36:18 for the life of me i can not get the ath10k card with a QCA9890 chipset to load the firmware https://paste.ubuntu.com/p/8VSNWTRXHy/ Feb 27 20:39:04 what kind of card is that? salvaged from a router or bought new? Feb 27 20:43:42 salvaged from a router Feb 27 20:43:58 that explains it Feb 27 20:44:22 retail wlan cards have their calibration in a small eeprom/ otp onboard the PCIe card Feb 27 20:44:39 but router manufacturers save the few pennies for that eeprom Feb 27 20:45:02 they store the <64 KB calibration data in the main system flash Feb 27 20:45:57 unless you can also get the original calibration data from the router you've salvaged the card from - and store it where OpenWrt looks for it, it'll remain dead in the water Feb 27 20:46:23 it's individual to each device Feb 27 20:48:00 i have that Feb 27 20:49:46 i i have a flashrom dump i took of the main flash Feb 27 20:50:20 good Feb 27 20:50:46 /lib/firmware/ath10k/pre-cal-pci-0000:01:00.0.bin is the first thing you'll need Feb 27 20:53:15 which was probably using a different PCIe bus number on the old router Feb 27 20:56:59 jow: yes? Feb 27 20:57:25 karlp: wanted to ask about that libubox testcase but remembered it was umdns Feb 27 20:58:17 ynezz: ping Feb 27 21:02:06 # Definitions for calibration data Feb 27 21:02:07 export ATH_CAL_MTD_DEVICE=1 Feb 27 21:02:07 export ATH_CAL_MTD_NAME=ART Feb 27 21:02:07 export ATH_CAL_MTD_ADDR_START=0x0000 Feb 27 21:02:07 export ATH_CAL_MTD_ADDR_INTERVAL=0x4000 Feb 27 21:07:34 jow: yup, just /etc/init.d/umdns restart will give the "command failed" error from the set_config with an array. Feb 27 21:07:52 ubus call umdns browse will "succeed" but always be empty. Feb 27 21:11:57 what kind of stuff do I need to run to get something to display there? Feb 27 21:12:09 is there a simple way to announce a service on linux? Feb 27 21:16:45 hrm, honestly, there's always been enough on my local network. Feb 27 21:17:00 rhytmbox advertises an itunes compatible thing, Feb 27 21:17:18 avahi-publish maybe? never tried I'm afraid. Feb 27 21:18:06 an older version of openwrt will advertise the ssh service form dropbear. Feb 27 21:19:35 $ avahi-publish-service -s myservirce _hoho._tcp 4000 "blah=wop" "neat=foo" Feb 27 21:20:56 ah great, exactly waht I needed Feb 27 21:21:00 yeah it works Feb 27 21:21:07 with the patched libubus Feb 27 21:21:32 yes, I've been running with that patch from the ticket for a week or so. Feb 27 21:24:46 pkgadd i have extracted the boarddata_0.bin from the rom dump Feb 27 21:26:28 also and variety of files like boardData_QCA9984_CUS238_5G_v1_003_FCC_US.bin Feb 27 21:26:47 you need to check what the uci-defaults would do on the donor router to read out the calibration Feb 27 21:27:17 assuming the donor router would be supported by OpenWrt Feb 27 21:27:26 its not Feb 27 21:28:14 that makes it more 'fun', as you basically have to find out how to decode it from ART Feb 27 21:37:26 i cant just put in on the file system under a certain name? Feb 27 21:37:47 20:50:46 < pkgadd> /lib/firmware/ath10k/pre-cal-pci-0000:01:00.0.bin is the first thing you'll need Feb 27 21:37:56 you may need more Feb 27 21:38:07 k Feb 27 21:42:27 if you have another ath10k based router, it may help to compare what's happening there Feb 27 21:42:45 as in, an already OpenWrt supported one Feb 27 22:05:40 jow: great! I moved libubox from "known issues" to "bug fixed" :) https://openwrt.org/releases/19.07/notes-19.07.2 Feb 27 22:17:05 zorun: wasn't a bug before, it was only a regression, you canm just drop it. Feb 27 22:19:15 karlp: well, it was a bug in 19.07.1, which is now fixed in 19.07.2 Feb 27 22:19:44 unless I got things wrong Feb 27 22:20:38 hrm, depends when the libubox got merged, I _think_ that was post .1 actually, Feb 27 22:23:45 it was merged about one week before the .1 release (20 January vs end of January) Feb 27 22:23:48 git describe --tags --contains f8902d1ae680ed88f0886c2acd453f1a5dfb468b Feb 27 22:23:49 v19.07.1~28 Feb 27 22:24:24 ah, fair enough. I have my own tree that just merges from 19.07 branch every now and again Feb 27 23:06:42 whats up with the imagebuildes? I'm still getting the IW bug and can"t build anything Feb 28 01:27:53 jow: fwiw, the commit sha in the Fixes on that libubox commit can't be right, the description doesn't match the hash. (it should be the one before) Feb 28 01:28:14 not a big deal, I'm much happier with the fix existing :) **** ENDING LOGGING AT Fri Feb 28 02:59:57 2020