**** BEGIN LOGGING AT Sat Sep 15 03:00:26 2018 Sep 15 04:53:11 any ideas why i'd suddenly stop being able to send commands over minicom? Sep 15 04:53:48 8n1 hardware/software flow control off, can receive fine and also stop the bootloader, but i can't send anything after that Sep 15 13:26:14 olleh Sep 15 16:27:02 In an init script, what's the most idiomatic way to get the wan interface? Sep 15 17:20:56 johnf: ifstatus wan Sep 15 20:39:36 ok I would have been looking forever Sep 15 20:39:47 but i have found something interesting Sep 15 20:39:59 the build system doesn't scale up well Sep 15 20:40:18 i just built my img on 12/32 cores on my threadripper Sep 15 20:40:22 builds ok Sep 15 20:40:30 built on 32 cores, it fails Sep 15 20:40:55 some packages can't handle it and need exceptions Sep 15 20:41:33 these race conditions aren't very easy to find, once identified, fixing usually isn't too difficult Sep 15 20:42:06 Indeed Sep 15 20:42:27 This is going to be a trollercoaster Sep 15 20:42:29 but Sep 15 20:42:32 will see Sep 15 20:55:57 I can get a successful build up to 20 cores Sep 15 20:58:11 the problem will always be the same, the number of concurrent threads just lightly nudges the probabilities of the race condition triggering or not. the situation might be quite different for someone else trying with the same level of concurrency - or it might catch you at -j4 or -j6 tomorrow Sep 15 20:58:49 interesting Sep 15 20:59:55 They're quite annoying to track down, so I generally just leave them be Sep 15 21:00:09 I can only take so much build system nonsense before cracking Sep 15 21:00:47 understandable Sep 15 21:01:25 I often run into it while building for lantiq (normal package set plus asterisk/ chan_lantiq), but it's difficult to pin point the failure - especially when you don't expect it and haven't full logging enabled (and if you toggle that, the issue often magically disappears due to slightly different timings) Sep 15 21:01:52 ughhhh Sep 15 21:02:27 I'm building for ARM EBU Armada Sep 15 21:02:55 it was just an example Sep 15 21:03:09 i understand that Sep 15 21:04:39 but if there are two architectures that are reaching such race conditions one(me) with many options enabled and you (normal + chan_lantiq) wouldn't the race condition be in the kernel itself? Sep 15 21:05:13 very, very unlikely Sep 15 21:06:22 the kernel is relatively standalone and has only few, well defined, build-dependencies - furthermore the kernel is regularly built by a f...load of different people on vastly different hardware, targetting even more diverse hardware using the most arachane toolchains Sep 15 21:06:34 yes Sep 15 21:06:39 issues like that get noticed quickly Sep 15 21:06:47 the userland is harder to tame Sep 15 21:06:52 well i didnt mean the kernel build itself Sep 15 21:07:16 but how the kernel is then tampered with by the build system' Sep 15 21:08:38 and as mentioned, my lantiq builds are fairly different from all my other target configurations by including asterisk (with is a monster truck in a race between reliant robins) Sep 15 21:09:41 who said that it would be the kernel build failing? did you debug it that far? Sep 15 21:12:45 no but once I have the toolchain built and the build system goes to cross compile for my architecture and I only select the default packages + kernel mods + luci I reach said race condition, Am I wrong to assume the race condition must be within the selected base packages, luci, or kernel mods? Sep 15 21:17:27 even that selects a quite large number of userspace packages Sep 15 21:17:47 you can't assume or guess here, you need to confirm Sep 15 21:42:44 understood Sep 15 22:17:27 thagabe: could be me. I've added a bunch of PKG_BUILD_PARALLEL:=1 to several packages Sep 15 22:17:40 make V=s will tell you what breaks Sep 15 22:48:39 make V=s doesnt run into the race condition Sep 15 22:49:02 make -jX does Sep 15 22:49:13 but.... will it blend? Sep 15 22:49:22 * karlp invokes make -j32 V=s Sep 15 22:49:26 all at the same time! Sep 15 23:10:04 fio's handling of both cygwin and windows is totally broken. lovely. Sep 15 23:10:15 exit Sep 16 01:10:51 Hello. Tell me please, is it really possible to get a consultation from the system developers? Sep 16 01:54:15 Tohin: that's not very likely in the middle of the weekend, but if you simply ask your question, perhaps someone will know **** ENDING LOGGING AT Sun Sep 16 02:59:59 2018