**** BEGIN LOGGING AT Wed Jan 22 03:01:07 2020 Jan 22 04:03:51 skipping 4.19 would be interesting. upstream got rid of the single queue I/O schedulers (finally). 5.4 also would allow Turris MOX support. Jan 22 08:16:27 Hauke: still a LTS makes more sense to me Jan 22 10:35:25 jow: any idea what's wrong here? https://github.com/openwrt/packages/pull/11089 Jan 22 10:36:00 There's no zip package being generated by the SDK, but it works just fine locally. Jan 22 12:10:27 ucert (master, 19.07) seems to be unable to parse its own output after you do an ucert -A -c "image.ucert" -x "image.sig". The resulting "ucert + signature" file is rejected as an invalid certificate, thus rendering sysupgrade impossible on 19.07 and master if CONFIG_SIGNED_PACKAGES is enabled. Jan 22 12:10:35 is this a known issue? Jan 22 12:16:05 (sysupgrade -F works, but that's beside the point) Jan 22 12:30:16 blogic: i think it is/will be LTS, just hasn't been announced yet Jan 22 12:30:37 that's also what hauke meant when mentioning greg and android Jan 22 13:09:08 hmh: this looks like a bug, please report it on https://bugs.openwrt.org Jan 22 14:43:32 wow, there's a first (afair): ath10k-ct + HTT fw: [2315981.927293] ath10k_pci 0000:00:00.0: firmware crashed! (guid n/a) Jan 22 14:48:21 greearb_: https://www.linux-ipv6.be/OpenWrt/ath10k-ct.txt - is this useful to you? if so I'll log it on https://github.com/greearb/ath10k-ct/issues Jan 22 15:11:02 zorun: looks like it, but given the damage it does, I find it quite strage that it survived for more than a day in both master and 19.07 since it basically renders anything signed impossible to sysupgrade without -F. Thus, I am ensuring it is not some local issue here, first Jan 22 15:11:52 hmh: signing is still not used officially Jan 22 15:11:57 and not enabled by default Jan 22 15:12:03 persoanlly I enver even tried it yet Jan 22 15:12:13 so it does not surprise me at all that it is broken Jan 22 15:12:44 should be marked BROKEN or DANGEROUS, then. That said, I am having a lot of trouble *disabling* it in the first place with ucert still enabled. Jan 22 15:13:10 but since it looks like ucert itself is broken, might as well continue using just usign in our local stuff :-) Jan 22 15:13:21 * hmh is still trying to root cause this a bit, to ensure it is not a local problem. Jan 22 15:13:38 I wouldn't trust ucert just yet Jan 22 15:13:46 given that even usign failed to calculate proper checksums Jan 22 15:42:57 ucert usage with blob_parse_untrusted() is utterly broken, since it cannot deal with chains. it has never been tested, evidently, since it simply CANNOT work. Jan 22 15:43:05 yeah, I am dropping that for good. Jan 22 15:43:07 will report the bug. Jan 22 15:43:56 (the defect is actually in libubox, it does not make any sense at all to use a buffer that can have one OR more blobs, and refuse to process if the size of the buffer is not EXACTLY the size of a message) Jan 22 15:44:16 which is actually *worse*. well, whatever. Jan 22 16:11:22 adrianschmutzler: the mikrotik branch looks OK Jan 22 16:11:30 not seeing any errors and ethernet works every time Jan 22 16:11:38 adrianschmutzler: thanks for your work also :-) Jan 22 16:11:59 * hmh hmms, libubox does not say if it wants a blob containing blobs (in case it is right and ucert is crap) or if it wants a series of blobs back-to-back (with padding). actually, it lacks docs all around :-( Jan 22 16:12:24 well, not crap. just never worked in the first place :-( Jan 22 16:36:25 meh, FS#2762 fwiw. Which the system set as priority "very low" Jan 22 16:41:09 and FS#2763 Jan 22 16:58:32 stintel, looks like I've seen that one before, very rarely. Looks like mem corruption and/or race related to stations disconnecting Jan 22 16:58:56 unless you can find a way to reproduce it reasonably often, probably not worth bothering about Jan 22 16:59:14 greearb_: ok, I have no clue what caused it Jan 22 17:01:03 xback: I've changed my mind regarding the lzma patch, I will send a v2 today. So, if you merge the mikrotik stuff, just don't merge the lzma commit (the latest one), I will then do that some days later. Should be tolerable if initramfs won't work there for two days. Jan 22 17:01:42 greearb_: apparently it happened right after my laptop disconnected due to inactivity Jan 22 17:01:52 that looks right from the backtrace Jan 22 17:02:00 ok Jan 22 17:02:12 if I see it again I could mess around to try and reproduce it Jan 22 17:02:14 I am updating bug checking in that code, maybe if it hits again on the new FW I will make progress **** BEGIN LOGGING AT Wed Jan 22 17:14:38 2020 Jan 22 17:32:34 error or the day: Jan 22 17:32:36 The command "exit 0" failed with exit status "1" Jan 22 17:34:16 * karlp laughs Jan 22 17:39:06 jow: urk :) Jan 22 17:44:29 blogic: 4.19 is already LTS Jan 22 17:44:47 the very long term LTS (5 years) are normally announced later Jan 22 17:45:36 and there is an android common kernel with 4.19: https://android.googlesource.com/kernel/common/+refs Jan 22 19:16:57 hmm, just updated a snapshot image today and got Jan 22 19:17:02 /etc/rc.common: ./collectd: line 182: syntax error: unexpected end of file (expecting "}") Jan 22 19:24:10 so I added the "}" at the end of /etc/init.d/collectd and it will complete without error... Jan 22 19:24:49 but somethings not right Jan 22 19:35:06 well, I don't know how this happend but /etc/inid.d/collectd seems to have lost lines 183 to 356 Jan 22 19:35:28 adding them back by hand brings collectd back to life Jan 22 19:36:43 the "update" was via image builder Jan 22 19:41:18 * nmrh wonder if 174 lines of code are missing elsewhere Jan 22 20:05:40 well, I guess this happend during the sysupgrade process as the file is ok in the image builder sysupgrade image Jan 22 20:07:44 funny, its also ok in /rom/etc/init.d Jan 22 20:13:55 arg, this is self inflicted never mind and sorry for the noise Jan 22 21:09:10 is there a way to specify parameters to modules as they are loaded as part of startup? - I want to specify a parameter to the ifb module Jan 22 21:10:19 the automatic start version of 'modprobe foo bar=baz' Jan 22 21:13:29 hmm, mr google says add parameters into /etc/modules.d/foo file Jan 22 21:32:43 yeah, arguments should work Jan 22 21:32:48 in modules.d Jan 22 21:34:13 I shall try shortly - will be back soon.... or not :-) Jan 22 21:36:14 ldir: looks like it is (somewhat) documented here: https://openwrt.org/docs/techref/init.detail.cc Jan 22 21:38:21 It worked. Jan 22 21:40:19 Hmm, wonder how I can make that a build default, or preserve across sysupgrade at least Jan 22 21:54:33 possibly by adding MODPARAMS.foo:=bar=baz to package/kernel/linux/modules/netdevices.mk Jan 22 21:54:37 * ldir tries Jan 22 22:02:56 * ldir succeeds Jan 23 01:00:01 are there still targets except x86 which are not using device tree? **** ENDING LOGGING AT Thu Jan 23 02:59:57 2020