**** BEGIN LOGGING AT Sat Sep 21 03:00:00 2019 Sep 21 06:31:56 rr123: I believe, that donations should be improved in the upcoming months Sep 21 06:33:10 rr123: manual -> it's opensource project, so anyone is more then welcome to join the effort in any part :) Sep 21 06:34:06 it was discussed during the meeting as well, notes could be found under "Create a manual" topic Sep 21 08:35:34 any idea about a circleci build failure i'm getting? "Package ninja is missing dependencies for the following libraries: libstdc++.so.6", but afaict the dependency is there: https://github.com/openwrt/packages/pull/10043#issuecomment-533779538 Sep 21 08:38:48 maybe some SDK hiccup? Sep 21 08:39:26 you tell me ;) Sep 21 08:39:50 it's not because of libstdc++ vs libstdcpp i assume? Sep 21 08:39:52 I mean, you've probably build tested your changes inside the buildroot where it worked Sep 21 08:40:12 but the CI tests are done with the SDK, which might behave differently Sep 21 08:40:25 ofc, runtime host/target is fine too Sep 21 08:40:41 hm ok Sep 21 08:41:17 that DEPENDS:=$(CXX_DEPENDS) is all i have to do for that dependecy, right? Sep 21 08:43:37 what if you add that dependency explicitly? Sep 21 08:43:41 dhewg: that error can also mean that the software was not cross compiled and linked against host linbraries Sep 21 08:43:58 DEPENDS := $(CXX_DEPENDS) +libstdcpp Sep 21 08:44:01 dhewg: or that a host-compiled artifact was accidentially packaged into the ipk Sep 21 08:44:23 or that the source tarball was unclean to begin with (upstream sources sometimes contain precompiled object files or similar) Sep 21 08:45:31 tarball seems clean Sep 21 08:45:46 I'd say cross compilation failed, then Sep 21 08:46:14 but i can run `ninja` just fine on the arm target Sep 21 08:46:16 weird Sep 21 08:46:41 how is this supposed to work? https://github.com/openwrt/packages/pull/10043/files#diff-a4d06ad0c75a4dd311db1c69e561998cR73 Sep 21 08:46:46 maybe it's worth mentioning that a cross compiled ninja is build with the host ninja Sep 21 08:47:06 yeah, that's exactly that ^ ;= Sep 21 08:47:08 ;) Sep 21 08:47:16 the Ninja macro uses $(2) as command Sep 21 08:47:18 its that define in ninja.mk Sep 21 08:47:33 yeah, for env vars Sep 21 08:47:33 ah, $(2) is probably meant for providing override env vars Sep 21 08:47:54 right, which is used for "DESTDIR=foo ninja install" Sep 21 08:48:14 well, I guess you have to debug or trace ninja, see what it actually is doing Sep 21 08:48:38 I am sure the new super cool and awesome hipster build system come with extensive documentation and debug capabilities Sep 21 08:48:49 hehe Sep 21 08:49:22 so you'd that it's unlikely a sdk issue? Sep 21 08:49:25 is there some way to have ninja output the gcc etc. invocations? Sep 21 08:50:06 yeah, i pass all that on Sep 21 08:50:07 if so, I'd say unconditioanlly enable that in a debug commit, force push to the pr, see what circleci is doing under the hood Sep 21 08:50:09 make V=sc package/ninja/compile Sep 21 08:50:42 sounds like a plan, i'll try that Sep 21 08:51:39 I wonder if the uploaded artifacts can be seen somewhere Sep 21 08:51:58 above that build failure is "Entering directory '/home/build/build_dir/feeds/base/package/libs/uclibc++'" Sep 21 08:52:08 maybe its just that, lemme try that libc++ Sep 21 09:06:43 dhewg: I can recreate that failure by using the SDK Sep 21 09:08:06 that's good so far Sep 21 09:08:18 never done that, i always build the tree myself Sep 21 09:10:34 yeah, i can reproduce too with uclibc++ Sep 21 09:14:08 for linking against uclibc++ you need to tell the compiler explicitly to not link against libstdc++, and you don't do that anywhere Sep 21 09:19:28 yeah Sep 21 09:19:32 its using CXX=arm-openwrt-linux-muslgnueabi-g++ Sep 21 09:19:54 its supposed to use the g++ with the -uc suffix Sep 21 09:20:11 which include/uclibc++.mk overwrites Sep 21 09:20:40 yet in devel/ninja TARGET_CXX_NOCACHE doesn't have the -uc suffix Sep 21 09:24:56 hardcoding it doesn't even compile "fatal error: unordered_map: No such file or directory" Sep 21 09:25:20 so i'll just hard depend on libstdcpp Sep 21 09:25:41 dhewg: any reason you use TARGET_CXX_NOCACHE and not TARGET_CXX? Sep 21 09:27:47 meson supports ccache in another way, it has to be wired up differently (i think, haven't yet looked into it) Sep 21 09:28:53 changing TARGET_CXX_NOCACHE to TARGET_CXX makes it (try to) compile against uclibc++ Sep 21 09:30:25 hm right, include/uclibc++.mk doesn't overwrite both vars Sep 21 09:33:50 but it looks like it won't compile against uclibc++ anyway Sep 21 09:33:52 src/hash_map.h:94:10: fatal error: ext/hash_map: No such file or directory Sep 21 09:34:06 uclibc++ doesn't seem to support that Sep 21 09:34:58 yeah, from commits in ninja i assumed it does Sep 21 09:35:09 anyway, i just force pushed with DEPENDS:=+libstdcpp Sep 21 09:47:53 nice, that worked Sep 21 09:48:38 and looking at meson/ccache, it looks like it works with the usual CC="ccache cc", i'll try without _NOCACHE Sep 21 10:04:29 dhewg: TARGET_CXX_NOCACHE not being overwritten by uclibc++.mk when ccache isn't enabled can be considered a buildsystem bug though Sep 21 10:06:27 yeah Sep 21 10:16:40 where are the 18.06 and 19.07 snapshots for downlaod located? Sep 21 10:18:19 Hauke: http://downloads.openwrt.org/releases/18.06-SNAPSHOT/ resp. http://downloads.openwrt.org/releases/19.07-SNAPSHOT/ Sep 21 10:19:33 KanjiMonster: thanks Sep 21 10:27:37 KanjiMonster: I added the links to the wiki Sep 21 10:31:19 AFAIK they werent added there for some purpose (forget the reason) Sep 21 10:45:58 I put them next to the build bot links: https://openwrt.org/infrastructure#buildbot Sep 21 10:46:33 installing kernel modules later will not work with these images **** BEGIN LOGGING AT Sat Sep 21 11:26:08 2019 Sep 21 19:50:43 dhewg: ping Sep 21 22:52:50 Hello openwrt-devel, could anyone commit https://patchwork.ozlabs.org/patch/1157844/ ? it's a really simple cherry-pick for 19.07 **** ENDING LOGGING AT Sun Sep 22 03:00:26 2019