**** BEGIN LOGGING AT Wed Jun 12 03:00:27 2019 Jun 12 07:40:21 meh. automatically running swanctl --load-conns after strongswan was started seems problematic Jun 12 07:40:45 what an idea to require that to load the connections in swanctl.conf Jun 12 07:41:16 and new features will not be supported in ipsec.conf so it's not possible to stay with ipsec.conf Jun 12 07:41:17 damnit Jun 12 07:48:34 testing if /var/run/charon.vici exists and is a socket is apparently not enough, still getting connection refused Jun 12 07:49:23 this seems to work but I think it's ugly: https://gist.github.com/stintel/f78c18cf07d3d83090d4133aff12acf7 Jun 12 07:49:26 any thoughts ? Jun 12 07:50:19 and at the end of start_service() in /etc/init.d/ipsec I added this: [ -x /usr/sbin/swanctl ] && /etc/init.d/swanctl restart & Jun 12 07:50:45 maybe I should check for /etc/init.d/swanctl instead. but that doesn't change the ugliness Jun 12 08:02:11 hello, i want to modify package/kernel/mac80211/files/lib/wifi/mac80211.sh . how do i create a patch with quilt ? Jun 12 08:03:10 make package/kernel/mac80211/{clean,prepare} V=s QUILT=1 does not provide the shell file Jun 12 08:04:57 you just edit it there, QUILT is for patches in the patches/ directories Jun 12 08:07:26 ok, thx Jun 12 08:08:29 any other way then quilt to provide the changes as a patch at build time ? Jun 12 08:13:30 de4rst: unfortunately not, this file iirc is copied verbatim after patches got applied Jun 12 08:23:34 is /initd/early.c already patched for unprivilleged lxc ? (MS_RELATIME vs MS_NOATIME) Jun 12 09:03:54 stintel: o/ Jun 12 09:04:04 xback: \o Jun 12 09:04:18 stintel: I was just wondering if you are stll seeing the mem leaks Jun 12 09:05:25 xback: I do, but that VM is still on 4.14.109 #0 SMP Thu Apr 4 09:35:53 2019 x86_64 GNU/Linux Jun 12 09:06:29 let me do a build for this device Jun 12 09:07:09 "device" Jun 12 09:07:36 quite likely I see that on my APU2s too but they have 4GB of RAM so they never hit OOM before I flash a new image Jun 12 09:36:51 hi! is it acceptable to ask openwrt devs here? e.g. to review pull request in gh. Jun 12 09:37:29 xback: will you update the kernel to 4.14.125 soon? Jun 12 09:37:49 hauke: currently finishing compile-tests Jun 12 09:37:59 Hauke: i'll probably push within an hour or so Jun 12 09:38:11 krd_: just ask - this is normal here Jun 12 09:38:18 xback: ok that should then "fix" FS#2305 and FS#2297 Jun 12 09:38:39 i'll check those 2 and update the patches accordingly Jun 12 09:38:52 the fix is still wrong, but the error is not happening any more Jun 12 09:41:26 krd_: now please ask your question... Jun 12 09:42:59 krd_: "asking to ask" on IRC is usually discouraged (an extra back-and-forth is performed for no reason). Jun 12 09:44:22 i still doubt whether it is necessary to divert someone. :] Jun 12 09:44:30 please do it Jun 12 09:45:54 i've fixed extra duplicate in libnghttp2, but library is anyway is installed as 0644. is it ok, or i should fix that too? Jun 12 09:46:56 krd_: what would the fix consist of? Jun 12 09:48:41 currently, library (and symlink) is installed via $(INSTALL_DATA), proposed change is to install via $(CP), since `install' dereferences symlinks. Jun 12 09:50:27 xback: I updated the kernel in 17.01, I have to do some more tests and then we can push it and probbaly do a minor release Jun 12 09:50:49 Hauke: thanks Jun 12 09:51:13 I didn't dare to touch the mac80211 incompatibilities :) Jun 12 09:51:44 I fixed them for lede 17.01, in upstream backports I just remove that part Jun 12 10:17:46 Hauke: https://git.openwrt.org/?p=openwrt/staging/xback.git;a=commit;h=7fdef072a2bff72462fda77b53bc244c5098fac4 Jun 12 10:19:59 de4rst: unfortunately not, this file iirc is copied verbatim after patches got applied Jun 12 10:20:02 +. Jun 12 10:31:52 There is no ImageBuilder for the adm5120/router_be target, bug? https://downloads.openwrt.org//snapshots/targets/adm5120/router_be/ Jun 12 10:37:32 aparcar[m]: the target is dead and not getting built by the buildbots anymore, we can probably remove the artifacts Jun 12 10:40:54 jow would appriciate. I'm getting a hang on removing target related stuff, could you point me somewhere on how to remove those artifacts? Jun 12 10:50:39 lynxis https://gitlab.com/gitlab-org/gitlab-ce/issues/53552 Jun 12 10:58:30 Hauke: I'll need another hour. spotted an error during compile-testing mvebu .. Jun 12 10:58:44 aparcar[m]: the adm5120 target was removed Jun 12 10:58:52 xback: looks good Jun 12 11:13:36 Hi everyone! I have a question concerning the SDK. Is it possible to somehow put the list of package dependencies into the Makefile but don't actually build them? I'm trying to build a package which has only shell scripts in it and it requires certain packages to be installed, so there is no need to build them. Jun 12 12:01:37 ao_: I think you want DEPENDS:= +foopackage1 +foopackage2 so it depends on foopackage1 being installed Jun 12 12:03:02 maybe EXTRA_DEPENDS works for that, not sure Jun 12 12:03:49 ldir: yes, it makes these packages installed, but also makes them compile, which I want to avoid Jun 12 12:04:04 I don't understand the 'so there is no need to build them.' Jun 12 12:04:44 KanjiMonster: haven't found any docs regarding EXTRA_DEPENDS, will try it, thanks! Jun 12 12:05:10 your new package (set of scripts) relies on some functionality provided by other packages....so those other packages need to be installed...which surely means they need to be built at some point. Jun 12 12:06:27 ldir: if I put a package into the "DEPENDS", the SDK builds these packages and their dependencies as well. I don't need to build them because they are part of the standard OpenWrt repository and can be installed directly via opkg Jun 12 12:07:23 if I build something myself, I don't want to do any opkg stuff.. so yes, ofcourse they will build then Jun 12 12:08:00 just let them get built, stop trying to work around a fairly artificial problem. Jun 12 12:08:13 also your package not saying it depends on something else is also fail... user can't know what is needed to install even if used opkg Jun 12 12:10:08 this is about runtime only dependencies, which we don't really support yet (but would be nice to have) Jun 12 12:10:39 for most packages dependencies are also build dependencies due to headers or libraries to link against Jun 12 12:11:43 KanjiMonster: looks like you are right with EXTRA_PACKAGES. https://github.com/openwrt/packages/issues/8535#issuecomment-483859547 Jun 12 12:12:30 if there's new magic fields, please add them to https://openwrt.org/docs/guide-developer/packages Jun 12 12:14:28 EXTRA_DEPENDS, not EXTRA_PACKAGES, sorry for a typo above Jun 12 12:23:29 karlp: not sure if 2006 counts as new ... https://github.com/openwrt/openwrt/commit/aa29fdce3d35a326245ec435e8b84314461ab71e Jun 12 12:54:00 ynezz: ping Jun 12 12:54:04 ynezz: regarding https://github.com/openwrt/openwrt/pull/1943 Jun 12 12:54:30 ynezz: I didn't notice that one and in the meantime got my own patch: https://git.openwrt.org/?p=openwrt/staging/xback.git;a=commit;h=71468ccf600f07b05ea1ba6fc3175a5388ad4e0f Jun 12 12:55:19 ynezz: do you want to continue with the original PR, or can I push mine in the meantime? (and fix the changes names separately) Jun 12 12:58:58 Question regarding the makefile install for packages - am I able to copy a whole directory from local to the install path via INSTALL_DATA Jun 12 12:59:08 or do I have to tree out the directory and copy it all manually Jun 12 13:00:31 you can just use $(CP) to copy it entirely, maintaining structure Jun 12 13:00:46 oh that makes it a lot easier Jun 12 13:01:06 does that retain permissions/executable status? or should I still use INSTALL_BIN Jun 12 13:02:27 I've only got a few executables so I should be okay to just override them with INSTALL_BIN Jun 12 13:05:21 JamieSinn_, beware of github.com/openwrt/openwrt/pull/2114 :) Jun 12 13:05:23 it keeps it all. Jun 12 13:05:50 I've got a few hundred APs with awful firmware, what's the liklihood of getting Openwrt on them? Jun 12 13:06:55 From what I'm reading in that issue rockdrilla - Symlinks are broken and will create duplicates with INSTALL_DATA? Jun 12 13:07:29 yepp, exactly. dereferenced symlinks and data duplication. Jun 12 13:07:30 I'm trying to deploy about 50-60 different openvpn profiles, so they're just flatfiles Jun 12 13:07:44 no symlinks, just a ton of .ovpn files Jun 12 13:07:56 so, build package locally and look into it for inconsistencies. Jun 12 13:08:17 will do, good note Jun 12 13:09:11 JamieSinn_, give a chance to wireguard - imo, it's better solution than ovpn. Jun 12 13:09:51 I took a look at it, but our IT dept wasn't amazingly interested in having to manage/deploy another server Jun 12 13:10:09 They're on my shortlist of replacements for OVPN since OVPN-AS pricing is painful Jun 12 13:10:15 Hauke: fyi, bumps pushed to master. i'll push 19.07 and 18.06 within 30 minutes Jun 12 13:16:57 Chris_ what you got? Jun 12 13:17:23 Slimey: https://imgur.com/a/bc9oQ7n Jun 12 13:18:47 tell me more Jun 12 13:19:05 a outdoor enterprise ap? Jun 12 13:19:15 yes Jun 12 13:19:19 make/model Jun 12 13:19:37 Alvarion WBSac-2450-O-EU Jun 12 13:19:58 Although on the label it says OAP7252AG Jun 12 13:55:27 Slimey: any thoughts? Jun 12 13:57:58 qca9550 should be supported ok, but you'll have the usual concerns of gpios and leds and bootloaders I'd imagine... Jun 12 13:58:18 what radios does it have on that minipci connector? Jun 12 13:59:24 I believe there's a QCA9558 and a QCA9882 Jun 12 13:59:38 9558 2.4ghz, 9882 5ghz Jun 12 14:01:13 probably not terribly complicated then, aiui, there's plnety of them already. Jun 12 14:02:04 not complicated if you know what you're doing! Jun 12 14:03:03 I think the current firmware that they're running might even be openwrt based Jun 12 14:08:31 ao_: I'm joining late, but EXTRA_DEPENDS is exactly what you are looking for. If you take the DEPENDS line, just omit the '+', and you're good to go. Jun 12 14:09:32 You can check the results in the package's control file. The dependencies should be there for opkg to install them, and they won't interfere with the build process. Jun 12 14:19:17 cotequeiroz: Thank you! I've added a line about EXTRA_DEPENDS to the wiki: https://openwrt.org/docs/guide-developer/packages?rev=1559328051&do=diff Jun 12 14:20:21 :+1: Jun 12 14:21:50 It may be worth adding that you may not use the conditional syntax there :, but since they're processed at build time, you can use standard Makefile functions to take care of that. Jun 12 14:22:43 buildbot: I know there's a swingle kernel built for all boards on a target/sub-target, but is the root fs built on a per-board basis? Jun 12 14:44:26 I've got a package that seems to be failing to be found when using ./scripts/feeds update Jun 12 14:44:40 checking the list of packages attached to the feed returns empty Jun 12 14:44:53 its listed in feeds.conf as a src-link Jun 12 14:49:36 I'm dumb - I typo'd the path Jun 12 14:58:34 karlp: where would start getting a firmware? Send some APs to a developer and ask them nicely? Jun 12 15:00:59 Chris_: Depends on the specific device -- "asking nicely" and hardware in hand for an "interesting" device or a trivial port might work. An unsupported SoC or effectively new, undocumented board isn't a quick-and-easy task Jun 12 15:02:04 Okie dokie Jun 12 15:02:41 I'm asking as I don't know how trivial or difficult it is. I know it's certainly beyond my expertise Jun 12 15:03:22 given that it's already supported radios and soc, it should be ~straightforward, if nto trivial. Jun 12 15:03:30 (I can't help personally) Jun 12 15:03:56 do you have bootloader access at all? you can try just booting one of the existing images for a similar board even, see how far you get... Jun 12 15:04:26 I don't have a serial ttl cable, I assume the 4 pins on the board would be for that Jun 12 15:05:35 Chris_: most likely Jun 12 15:05:36 ..or at least good assumption Jun 12 15:05:41 Yes, only TX, RX, and GND need to be connected. Make sure that your adapter is 3.3 V logic -- "TTL" (5 V) or "RS232" can/will quickly fry your SoC Jun 12 15:06:12 If you've got questions about specific adapters, https://forum.openwrt.org/ can be a good place to check Jun 12 15:06:41 just need to find the right pinout Jun 12 15:07:16 esr1750 for instance seems pretty similar Jun 12 15:07:26 not sure why that's not in master though, it's in 1806... Jun 12 15:09:23 karlp: to big kernel for kernel partition Jun 12 15:10:06 ah, that old joy Jun 12 15:11:12 the same was with epg5000 but I ported it to ath79 Jun 12 15:23:53 Chris_: Your router uses a NAND flash. And the ar934x-nfc driver should be ported to ath79 before support for your router can be added :) Jun 12 15:24:35 sorry, I don't follow gch981213 ? Jun 12 15:26:16 (it's not quite as "simple" as I might have implied) Jun 12 15:26:29 ok Jun 12 15:27:19 Yep. That's exactly what I mean :D Jun 12 15:28:50 https://imgur.com/Us2uaXt Jun 12 17:30:14 Is there an easy way to override this error? Jun 12 17:30:47 https://pastebin.com/6Vckkzeq Jun 12 17:31:02 I want my package to override them (not publicly distributed Jun 12 17:33:04 I could include them in the files dir, and just deal with it that way, but I'd prefer a way to update the config files easily Jun 12 18:00:48 I've complemented ao_'s information with EXTRA_DEPENDS and PKG_BUILD_DEPENDS syntax. https://openwrt.org/docs/guide-developer/packages?do=diff&rev2=Array&difftype=sidebyside Jun 12 18:08:10 Any better place to pick up an "ask user" sh function than /lib/upgrade/common.sh ? Jun 12 19:05:24 Any "easy" way to unmount an overlay? `umount -f -t overlay /` doesn't seem to do anything Jun 12 19:06:21 (nor does pre-umount-ing /overlay) Jun 12 19:13:22 jeffsf: the easiest way is to reboot into failsafe and not mount it ;) Jun 12 19:13:56 LOL, yeah, I'm trying to fix firstboot on NAND and dealing with the case of already mounted Jun 12 19:14:27 and I recall several "complaints" that globbing of * missed .file Jun 12 19:16:05 Might just break down and use `rm -rf *` and check that it does the right thing Jun 12 19:17:01 (Probably does, as leaving behind `.fs_state` might be a *good* thing) Jun 12 19:18:03 (Ugh, more digging into Linux internals) Jun 12 19:24:43 jeffsf: in practice it's quite complicated, which is why we got an extra helper for sysupgrade to switch to pid 0(1?) to be able to unmount overlay, as it will likely fail when not done by the init process Jun 12 19:26:08 Thanks KanjiMonster -- I'll look into the "right" way to erase an overlay -- `.fs_state` bothers me in that I don't know what it is just yet Jun 12 19:28:54 `ubiupdatevol -t /dev/ubiblah` works well if unmounted, but trying to write a JFFS2 mark definitely isn't the alternative Jun 12 19:31:38 truncating the roofs_data partition is definitely the right way when unmounted Jun 12 19:32:41 Yep, clean, preserves block-erase counts, UBI BBM, ... Jun 12 19:33:32 jeffsf: maybe it would make more sense puttting a special tag file in the overlay to indicate it should be erased on boot, for which the generic boot code checks when mounting rootfs_data before using it for the overlayfs; then let jffs2reset on a "hot" system just create that file and reboot Jun 12 19:34:30 I have a mPCIe 3x3 mimo WiFi card and it's country code is set to 00 in its eeprom. Jun 12 19:34:30 Do I need to apply patches, e.g. 402-ath_regd_optional.patch, myself or is this functionality included in 19.07? Jun 12 19:34:54 Quazil: these are already enabled by default nowadays Jun 12 19:35:19 Hmmm, good thought -- trying to avoid messing with jffs2reset as its pretty jffs2-specific Jun 12 19:35:42 But I like the general approach KanjiMonster Jun 12 19:35:55 Kanjimonster: If the bands are disabled will they not appear in the LuCI GUI? (mening if I can select them that means they are valid as-far-as the driver/firmware is concerned?) Or is that not correct/reliable? Jun 12 19:37:14 jeffsf: I remember weird side effects after running jffs2reset on jffs2 systems (like once-ovewritten files now not recreatable etc) Jun 12 19:38:18 You're bringing back memories of White Russian and the mythical 30-30-30 Jun 12 19:38:20 Quazil: it means you set a country code and it replaces the eeprom one; after that they should appear in the gui Jun 12 19:38:47 Superceeds the eeprom one or actually writes to the eeprom? Jun 12 19:39:07 no modifications of eeprom at all Jun 12 19:39:31 (ty, I'm new to this deep of level of wifi so I don't know what the expected behavior is.) Jun 12 19:46:13 jeffsf: I mean I'm not even sure that I *want* to define how the system is supposed to behave when you replace/reset the filesystem contents (should services be restarted so they use the default config, how about firstboot generated files etc), so enforcing a reboot when running jffs2reset sounds like a reasonable way of enforcing a "clean" reset IMHO Jun 12 19:58:51 KanjiMonster: The more I think about it, "next boot" is the only way to robustly handle it on a multi-threaded, not to mention multi-user system Jun 12 19:59:43 jffs2reset fails miserably with NAND, in general, and doesn't "understand" UBI volumes Jun 12 20:00:35 (tries to write a 4-byte JFFS2 mark to raw MTD device) Jun 12 20:38:09 Chris_ heh sorry been in meets all day Jun 12 20:38:13 meetings Jun 12 20:38:36 no problem Slimey Jun 12 21:21:39 jow: ping Jun 12 21:26:56 jow: I got it nearly working but got the feeling the PKG_LICENSE_FILES tag causes trouble. https://paste.debian.net/1087664/ Jun 12 21:30:12 why is there gplv2 and gpl-2.0? Jun 12 21:30:28 one of those must be something freeform, not the spdx tag Jun 12 21:33:27 502 Bad Gateway on LuCI. AGH Jun 12 21:35:43 karlp: PRs welcome ;) Jun 12 21:40:34 karlp https://github.com/openwrt/openwrt/pull/2120 Jun 12 21:41:24 * karlp upvotes Jun 12 21:43:01 a PR by me that somebody likes, huh! Jun 12 21:44:30 * ldir grins Jun 12 21:45:30 it doesn't build here :) Jun 12 21:46:14 (╯°□°)╯︵ ┻━┻ Jun 12 21:46:43 and I'm not sure if the license change should be the reason for the package version bump Jun 12 21:47:12 if the license is included in the opkg (is that still just planned) then it should bump, otherwise not I guess. Jun 12 21:47:20 no harm thoughh, other than a tiny amount of rebuilding for some places. Jun 12 21:48:18 karlp I'm nearly done with that feature Jun 12 21:50:33 `git grep GPLv2` suggests like we need more housecleaning :) Jun 12 21:55:01 as it requires a release bump a simple replace wouldn't do Jun 12 21:55:23 patch send to mailing list Jun 12 21:58:26 welllll GPL-2.0 is _also_ wrong.... so... Jun 12 21:58:40 if you go by the source doc: https://spdx.org/licenses/ Jun 12 21:58:42 ynezz so when I fixed all these, I get a tshirt? Jun 12 21:59:05 GPL-2.0 was the old name, now deprecated. Jun 12 21:59:22 aparcar[m]: if you cheat, yes :) Jun 12 21:59:31 when they got rid of the "+" in the names because it confused things. Jun 12 22:00:39 yea, it's SPDX v3.5 incompatible Jun 12 22:00:48 maybe we should wait for v3.6 Jun 12 22:03:00 so you're telling me the PR I just merged & closed was pointless :-) dammit - I still closed a PR Jun 12 22:03:03 so no merge for now or rather be deprecated than entirely wrong? Jun 12 22:04:49 Totally sidepoint humour: can we pull request SPDX to.have and use forever valid lisence of: (╯°□°)╯︵ ┻━┻ Jun 12 22:06:44 olmari: yes, if you explain what it's supposed to be Jun 12 22:07:05 throwing a table Jun 12 22:07:08 ldir: someone flipping a table Jun 12 22:07:57 ooooohh Jun 12 22:08:17 ldir: terminal breaks? (Unicode is a bitch sometimes too if the case) Jun 12 22:11:21 I have a crude image that on my terminal looks something like a pig and indeed a table on the right - but I couldn't join the two together - having googled I now understand :-) Jun 12 22:11:33 "strike any key" Jun 12 22:11:40 with a hammer, preferably Jun 12 22:12:52 I don't have an 'any' key ;-) Jun 12 22:13:44 ... I know, make something idiot proof and they'll only build a better idiot Jun 12 22:14:40 new idiots, new warranty return ;) Jun 12 22:14:46 You're lucky you're not following the 12:00 flasher Jun 12 22:14:48 These 2 keys is must Jun 12 22:14:57 * olmari uploaded an image: ANY_DSCN0366.jpg (209KB) < https://matrix.org/_matrix/media/v1/download/hacklab.fi/RbzqTGMwRxhUstseCFOEVxlr > Jun 12 22:15:29 I gave up after the first 81-poat thread Jun 12 22:18:41 should there be a treewide update for SPDX? And is there a script to increase the release for treewide changes? Jun 12 22:18:51 aparcar[m]: btw it builds on macos, so I'm happy ;-) Jun 12 22:24:41 aparcar[m]: as I said, I'm really not sure if we need the package bump, as the packages are build from scratch Jun 12 22:24:42 ldir oh right, so circleci wants cash for using macosx CI. maybe a freebsd docker container could be used for testing? Jun 12 22:24:43 https://hub.docker.com/r/auchida/freebsd Jun 12 22:25:32 ldir: No experience with that, but I can simply add this container to the ci if you tell me how to install the required packages on freebsd Jun 12 22:25:48 ynezz without the bump the new opkg feature wouldn't work Jun 12 22:27:42 afaik no one has really succeeded in using FreeBSD as a build host in recent times, there were a couple of forum threads about it over the past couple of years, but lots of (tiny) problems, without real success Jun 12 22:29:05 * ldir decides it's too late so will pick this up after sleep Jun 12 22:29:13 aparcar[m]: it will, you'll just need to wait for package rebuild Jun 12 22:31:25 if it works for you localy with `make package/foo/{clean,install}` then it should work on the buildbot with the same way Jun 12 22:32:36 i hope, that nobody is going to waste flash cycles just to update package in order to be SPDX v3.5 compatible Jun 12 22:35:55 ynezz right, I forgot that the buildbot still rebuild everything even if there were no changes. another todo :) Jun 12 22:37:23 ynezz changed, however it's in ldir's staging tree :S Jun 12 22:41:24 aparcar[m]: just because the package didn't change doesn't mean the sdk used didn't change in a significant way (e.g. new gcc version) ;) Jun 12 22:41:58 I isee Jun 12 23:03:35 tools and toolchain is for every target the same, right? Jun 12 23:06:35 aparcar[m]: musl doesn't support all target arches, so some use uclibc, some glibc (see https://github.com/openwrt/openwrt/blob/master/toolchain/Config.in#L238 ) Jun 12 23:07:10 KanjiMonster good to know, thanks! Jun 12 23:08:52 aparcar[m]: also arc uses gcc8; while all other targets use gcc7 (https://github.com/openwrt/openwrt/blob/master/toolchain/gcc/Config.in#L3 ) Jun 12 23:15:48 how come arc is so avangart? Jun 12 23:19:39 the less friendly way is, they spent years thinking they didn't have to bother upstreaming thheir work, and now they're still catching up Jun 12 23:19:46 so gcc8 is the first to properly support it. Jun 12 23:34:12 thanks for the clarification Jun 13 01:10:12 * jeffsf_ frowns at |.ELF............|, hoping to stay out of C code for at least a day Jun 13 01:25:54 No `tempfile` under OpenWrt? Jun 13 01:26:57 NM, `mktemp` **** ENDING LOGGING AT Thu Jun 13 02:59:57 2019