17:28:41 #startmeeting 17:28:41 Meeting started Wed Jul 20 17:28:41 2016 UTC. The chair is jow. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:28:41 Useful Commands: #action #agreed #help #info #idea #link #topic. 17:28:56 #topic LEDE July meeting 17:29:21 nbd: It wouldn't hurt. We could possible archive to gmail 17:29:21 we can start if you're ready 17:29:34 hi 17:29:38 sorry got held up 17:29:57 np, greato have you here 17:31:11 SO the agenda proposal is here 17:31:13 agenda: http://piratepad.net/RVVihbFpUh 17:31:14 Title: PiratePad: RVVihbFpUh (at piratepad.net) 17:31:19 #link http://piratepad.net/RVVihbFpUh 17:32:21 #topic Infrastructure 17:33:10 OK - I'll bite. Flyspray? 17:33:53 no problem we can start with Flyspray 17:35:02 I have been in contact with the developers (such as they) are and they mentioned the desire for help with the product. 17:35:27 Personally, I am not all that interested in doing that, but I will continue to look into bugs, 17:35:39 report them and push fixes to them on Github. 17:36:38 Also, I would like to do more triage work on reports and have a request for all of you. 17:37:09 I would assign bugs for investigation or fixing if we had a list to topics each of us was 17:37:26 interested in handling. Otherwise things just sit there until someone picks them up. 17:38:56 makes sense to me 17:39:23 at least I know that I cannot really fix ath9k or ethernet driver bugs 17:39:36 I'm happy to deal with buildroot or packaging stuff though 17:39:59 or fix things in the projects/ realm, such as ubox and things 17:40:08 Any objects to me putting out a request (email) for topics / specialities from each developer/committer. 17:40:22 thess: sounds good 17:40:53 I'll put it up as a page on our site 17:41:07 #action inquire committer topics for better future bug triage 17:41:31 #action publish results in web.git 17:42:11 Anything else with FS? 17:42:27 thess: i like it alot 17:42:29 at some point I like to put some effort into the design 17:42:43 the layout of comments to an issue is unlucky imho 17:43:02 like shifted to the left below the menu and way narrower than the rest 17:43:07 makes it easy to overlook 17:43:22 but all in all I like it too, especially the github auth 17:43:50 Yeah, the Github auth is awesome. 17:44:26 Layout changes are pretty straight forward as it is a themed CSS (ala Drupal/Joomla) 17:44:26 KanjiMonster asked for less redundant info in the mail notifications 17:44:41 but I didn't find the motivation to redo that part yet 17:44:54 jow: Yes, still have open task on that. 17:44:58 I was lucky that the lede-bugs and lede-commits lists work as they do for the moment 17:45:07 I mostly care about the subject 17:45:32 that we can probably change easily 17:45:54 #action rework FS issue notification mail subjects to allow gmail threading 17:46:12 iirc it is an printf-like string available directly via the settings menu 17:46:22 Didn't we do that already? 17:46:41 I changed some thing already but did not yet remove the change/action type from the subject 17:47:31 it should basically always contain just the issue title without any "comment added" or "issue changed" suffix 17:47:33 With respect to layout - take a look at bugs.archlinux.org (Flyspray site) 17:48:33 thess: it has the same problem though, comments feel a bit "unstructured" and unrelated to the actual issue 17:48:57 but thats bikeshedding - lets move to the next item 17:49:17 Regarding the mirror situation 17:50:33 currently we have the problem that half a dozen different build slaves are rsyncing their artefacts to the download server at random intervals 17:50:53 on top of that the hetzner rsync performance is awful at best 17:51:43 the mirror team of the RWTH aachen sponsored us a machine we can use to act as primary download server 17:52:32 that machine should provide a more stable rsync performance compared to hetzner, downside is that it does not have ipv6 connectivity 17:52:52 are they working on ipv6? 17:53:05 it is supposed to be available "soon" 17:53:17 but as we all know this can mean next month or next decade 17:54:01 jow: howabout having master at aachen 17:54:09 and sync to hetzner and provide ipv6 there ? 17:54:23 furthermore we got another mirror sponsored from rwth aachen (aside from the yet-to-be used master machine) 17:54:56 I am for blogic's idea 17:55:01 we got asked by the people there why their mirror receives so little traffic and whether they should continue providing it 17:55:16 +1 w/john 17:55:27 they said its currently spending more bandwidth rsyncing than serving any clients 17:55:41 which brings me to the next problem giving me headaches 17:56:16 our build artefacts are extremely volatile, from the rsync pov there is little similarities between the binaries produced by two consecutive build runs 17:56:24 we should investigate why 17:57:29 example here: http://phase2.builds.lede-project.org/builders/mipsel_74kc_dsp2/builds/134/steps/packageupload/logs/stdio 17:57:58 weird 17:58:03 date stamps, signatures ? 17:58:12 lynxis: did we not have reproducable builds ? 17:58:22 thess: possible, it needs to be figured out 17:58:42 so what did I do so far; 17:59:06 - I lowered the TTL of downloads.lede-project.org to 5 minutes so that we can quickly switch to the new host 17:59:39 - I assigned an alternative name to the current rsync server and reconfigured the buildbots to sync there, to not interrupt things during the switch 18:00:04 blogic: reproducible builds: I'm working on that one. will finish it tonight. 18:00:09 - I prepared HTTP, HTTPS and RSYNC on the new machine 18:00:36 What I need to figure out yet is how to make the rsync transfers atomic 18:01:54 my plan was to investigate pre-xfer and post-xfer hooks in rsyncd 18:02:44 the basic idea is to acquire an exclusive file lock when rsyncing to the download master and let upload requests from builders stall until that master transfer is through 18:03:11 jow: just need way to cleanup lock after failures/timeouts, etc. 18:03:34 how is debian doing those things? 18:03:36 the current (old) download master will serve 1) as staging area to gather and consolidate the buildbot uploads 18:03:44 2) as ipv6 mirror 18:04:00 also ensure that other builders that wait for the lock won't get killed because of a percieved lockup as well 18:04:28 KanjiMonster: that is part of the investigations 18:04:33 lynxis: they don't 18:05:03 at least from the things I found in the internet 18:06:11 is it possible to upload to a tmp directory and when it is finished just change the link from the old one to the new one 18:06:25 KanjiMonster: the idea was to use some kind of best-effort flock, basically wait for something like 30s on the exclusive lock then continue anyway 18:06:59 KanjiMonster: that should already dramatically reduce the likelyhood of rsyncing partial uploads to the download master (assuming the staging<->master connection is reliable and fast) 18:07:35 Hauke: that would be my backup plan 18:07:38 if we upload it to a tmp directory which changes for each upload we do not have any locking issues 18:08:54 jow: is --delay-updates not enough? 18:09:31 nbd: no, it cannot guarantee the consistency of the entire tree, e.g. if the Packages index fits to the .ipk files 18:10:26 the other point I want to discuss quickly is the traffic distribution 18:10:48 jow: how about this: hard link the dl dir to a new one, rsync to that one, then rename it and get rid of the old one 18:10:51 as I wrote above we got complaints from the rwth that their lede mirror is underutilizied 18:11:34 jow: just turn it into the master 18:11:41 and point downloads.lede... there 18:11:45 that should utilize it :-) 18:12:08 blogic: the one they complained about is not the future master machine but their main ftp 18:12:58 we've got five instances from them by now, three buildslaves and two download machines 18:13:10 does opkg use the mirros or only the master? 18:13:17 opkg only uses the master 18:15:13 jow: So as blogic said - make it the master 18:15:49 ok 18:15:49 I never thought I'd hear of someone complaining about not consumming bandwidth 18:16:13 #action promote RWTH machine to new download master 18:16:48 SHall we switch to the code updates? 18:17:25 +1 18:17:32 okay 18:17:32 OK 18:17:36 #topic Codebase updates 18:17:50 Not much from me here, a few very minor commits aside 18:18:03 I mostly merged random PRs and a few patchwork items 18:18:19 same here 18:18:34 I'm working on omap to become compatible with beaglebone black cape's 18:18:39 updated lots of image building code with mayor collateral dammge 18:18:52 but nbd fixed most of it now 18:18:53 lynxis: Great! 18:19:00 ipq seems stable now 18:19:29 is this todo list up to date: https://lede-project.org/todo.html ? 18:19:29 apart from the mange bootargs thingy hanging on certain cmdline contents 18:19:30 Title: Project ToDos (at lede-project.org) 18:19:38 *mangle 18:20:02 KanjiMonster: i did not follow it to be honest 18:20:03 Hauke: no, the todo needs a mjor overhaul, ideally we decide the new contents in this meeting 18:20:25 KanjiMonster: the conversation cinfused me and i did not understand what the problem was so i am waiting for someone to send a patch 18:20:59 what are the bigger topics which are missing for the next release? 18:21:27 the one big topic which is affecting the actual release mechanics is the multi device image building code 18:22:04 blogic: it seems to choke at a root= argument that doesn't match what the code is expecting, hanging the device. I guess, the code isn't the most readable. 18:22:43 KanjiMonster: as i said, i'll wait for a patch to fix it 18:22:54 KanjiMonster: i only have the eval kit so i am already chuffed we got thi far 18:23:03 jow: yes 18:23:06 from my side I'd say we could start release branches any moment as soon as we can build images containing their profile packages 18:23:10 jow: is this only the target specific code or some generic code which is missing? 18:23:28 Hauke: as far as I know the entire implementation of it needs to be done 18:23:33 nbd correct me if I'm wrong 18:23:36 jow: once that is done we should fire up the infra and push a release and use it as a testing release to get status quo and then push another release soonish 18:23:50 the image building rework in the targets is basically preparation steps to overhaul the image building 18:24:03 ok 18:24:50 is IPv6 working now? there is " 18:24:52 Review recent IPV6 changes 18:24:57 on the list 18:25:16 the idea is that we can build multiple rootfs images in one go, to let each final device image contain exactly those packages it specified in its default packages (e.g. kmod-usb in tplink units having usb ports) 18:25:38 sorry to be offtopic: does anyone know why wikidevi has been down for days? 18:26:19 Hauke: it was never really broken, it just has various quirks 18:26:25 ok 18:26:41 but all in all I'd say that it works 18:29:07 anything else here? 18:29:14 Release names? 18:29:19 ah yes 18:30:10 I'm for the pragmatic approach and would just call it LEDE 1, LEDE 2, ... LEDE N 18:30:10 I have become partial to yy.mm.minor however, it should be the actual release date and not a desire. 18:30:40 but yy.mm is fine for me too 18:30:57 I just want to not deal with the code naming anymore 18:31:32 Fine with me -- projects have names, release have version numbers 18:31:38 its tedious to find suitable names and without inside knowledge its hard to figure out which version it corresponds to 18:31:47 I never know how debian 8 is called 18:31:54 jessie 18:31:58 :) 18:31:59 and I never know what version "Gutsy Gibbon" is 18:32:03 ... 18:32:03 whoops - wrong 18:32:46 dates are fine - how about use the date we branch and the minor for releases (candidates and actual) 18:33:16 yeah, something like branchyear.branchmonth.buildnr or something like that 18:33:29 +1 18:33:53 regarding version numbering, I find the libreoffice approach for rcs sensible, where xx.yy.zz rc1 is just xx.yy.zz.1, rc2 then xx.yy.zz.2 etc, which avoids the rc -> final rename issue for opkg paths 18:34:30 why not 18:35:19 I prefere LEDE 1, LEDE 2, ... over a date 18:35:20 what I was aiming at was simply carving out some side branches that represent a release and put aside some buildbot resources to just watch those branches 18:36:03 e.g. branch master into 16.09 or lede1, then let buildbots build that whenever something is pushed there, much like the current snapshots are done 18:36:07 this has a few upsides: 18:36:11 having build bot build the branches would be very nice 18:36:23 - new users always install the latest patch version 18:36:37 hauke: the problem with LEDE 1, 2 etc. is these names are more appropriate for a product name than an engineering release 18:36:42 - rolling out fixes is as easy as cherry picking or push things to the branch 18:37:11 - no dedicated release maintainer is needed anymore 18:37:40 jow: whatever the name, your strategy is good 18:37:53 what do we do when the kernel version or config gets changed, then a full upgrade is needed 18:38:09 I like the strategy 18:38:12 Hauke: in an existing release? 18:38:37 we should simply avoid it 18:38:44 like a minor kernel version update 18:38:49 we can still just make a new release 18:39:24 its basically zero effort, just branch, adjust default config, tell buildbot to watch releasebranch++ 18:39:27 i wish names were sortable 18:39:37 e.g. 1 vs. 12 18:39:42 so we should have... 001? 18:39:45 sounds weird 18:40:06 rmilecki: sort -n can also sort 1 vs 12 18:40:14 jow: sounds good 18:40:16 yes, but some tools can't 18:40:26 e.g. i'm not so sure about nginx listing 18:40:27 so back to yy.mm ? ;p 18:40:52 yeah lets do yy.mm[.buildnr] 18:41:01 agree w/jow 18:41:02 +1 18:41:05 will be add luci by default in the stable branch builds? 18:41:11 I guess so 18:41:16 ack on both 18:41:20 yy.dd + luci 18:41:24 yy.mm ;) 18:42:01 #topic Releases 18:42:12 #agreed use buildbots to build release branches 18:42:30 #agreed use branch year and month (yy.mm) as version designation 18:42:52 this all depends on the image building though 18:43:01 last time we had no eta for that 18:43:13 nbd: can you chime in? 18:45:16 I guess no, so lets move to the openwrt merge discussion? 18:45:35 i wish to have this completed 18:45:38 and then take a decision 18:45:47 see who of is working on what 18:45:56 and agree on some schedule 18:46:03 rmilecki: this = merge or this = release stuff? 18:46:08 rootfs 18:46:10 ah 18:46:27 yeah I wish to have it completed too then carve out a release asap 18:46:33 +1 18:46:45 +1 18:48:38 #agreed multi rootfs image building needs to be finished asap 18:49:10 shall we wait for nbd? 18:49:33 I need to leave soon 18:49:40 let me call him 18:49:43 1 sec 18:49:46 meet again soon? 18:50:10 re 18:50:40 nbd: comments ? 18:50:58 just a moment, need to read backlog 18:51:30 image building is still a bit hard to predict, but i got most of the groundwork in place now 18:51:42 there is now a way to deal with ubifs and its different flash settings properly 18:51:45 using the new image format 18:51:49 i've converted some targets to it already 18:52:08 i should be able to make legacy images work with the new rootfs building code as well 18:52:48 if no unexpected issues show up, i hope to be able to get a prototype of it working in maybe 2 weeks 18:52:55 it requires more porting work though 18:53:07 to make targets use either the new image building code or at least the LegacyDevice wrappers 18:53:38 I guess we can all volunteer in converting the target parts 18:54:04 OK 18:54:59 jow: most is done 18:55:08 rt305x is missing, i forgot about it 18:55:08 got mvebu fully ported over in my staging tree 18:55:14 kirkwood definitely needs some love 18:55:26 legacy nand devices don't produce images 18:55:46 maybe related to ubi-utils 18:56:03 #action Give some love to kirkwood 18:56:58 anything else on the image building topic? 18:57:34 jow: If you need another target builder, I have have one ready to go. 18:57:35 i would take care of reproducible-build on https://reproducible-builds.org/ 18:57:36 Title: reproducible-builds.org (at reproducible-builds.org) 18:58:07 jow: only that i wish i had not had the idea :-) 18:58:51 blogic: but it is long overdue 18:58:56 jow: i know 18:59:09 ok, remerge topic? 18:59:19 ok 18:59:29 #topic OpenWrt merge discussion 18:59:48 must things that happened in the last weeks can be seen on the mailing list 19:00:44 prpl reiterated their wish for us to sit together and work out the issues 19:01:05 is there big intrest in a merge ? The discussion on the list stoped 2 weeks ago 19:01:23 Hauke: i think yes 19:01:33 personally I'm a bit lost in the topic 19:01:42 not sure waht we're still supposed to do 19:01:43 Hauke: i guess people are scared to be flamed at so they dont get involved 19:01:51 if you ask me, I don't have too much interest, i'm unfortunately getting annyoed by remaining OpenWrt ppl from time to time :( 19:01:51 jow: agreed 19:02:15 kaloz being MIA doesn't help either (at least it seems that way) 19:02:31 I helped zoltan to migrate the openwrt wiki to a non-lede machine 19:02:35 jow: i honestly would not mind, however i think we cant make it happen 19:02:37 yeah, last i heard from him he wanted to try to put together an opportunity for a face to face meeting 19:02:41 its up to the owrt folks i guess 19:02:51 but i didn't hear back from him 19:03:07 jow: the thing is we did everything thewy asked, name conditions, propose a process, enter discussions 19:04:16 KanjiMonster: what is the situation on the openwrt side? is there any more activity in irc? 19:05:00 nbd: so the paris meeting was just happy talk? 19:05:54 jow: the public channels are still rather active 19:06:32 ok 19:06:46 jow: mostly. i still do think that a face to face meeting would be quite important for getting closer to being able to get the remerge discussion going in a constructive direction again 19:06:59 nbd: the question is what should be in the end 19:07:19 personally I neither care about the openwrt name nor about the old openwrt codebase 19:07:43 and I know that we cannot handle wiki and forum without serious volunteer help 19:07:50 like I said having a F2F meeting would probably exclude some people that have problems traveling 19:08:07 well f2f could be a video conference too I think 19:08:29 I like f2f meeting and it gives us much more than a video conference 19:09:00 so I guess one thing we should figure out first; 19:09:07 someone added a meeting at the ELCE 19:09:17 I assume that many people will be there 19:09:46 is any active lede developer willing to support a remerge and willing to work on a reunified project or is someone having objections 19:09:54 I think we shold do an audio/video meeting first and then decide for F2F 19:10:00 or rather are all instead of is any 19:10:28 I would like to see us merge again 19:10:29 and, where do we want to end up, what is the desired outcome 19:11:19 jow: well, we already agreed on what we think should be the way 19:11:47 in short, have them join lede and then use the old brand in one way or another after the domain is handed over to spi 19:12:02 i doubt we would find much support for doing it the other way around 19:12:22 blogic: there is more than domain if u ask me 19:12:34 mail servers, mailing lists, irc channels, servers 19:12:36 rmilecki: yes that was the short version 19:12:37 what about them? 19:12:45 nto that I expect a full answer 19:12:48 just many things to consider 19:12:54 sure 19:13:05 mailing lists can be forwarded 19:13:16 irc channels live and die by their participants 19:13:28 jow: they are maintained by someone at OpenWrt who does random things silently 19:13:31 servers can be repurposed 19:13:46 that's why I can't figure out a good way to work with OpenWrt 19:14:39 rmilecki: but thats the exact problem, once you're outside of the openwrt core its a nebulous thing 19:15:02 jow: i never realized how much until i was not part of the core anymore 19:15:02 Does OpenWrt represent more to us than just resources. 19:15:09 its almost impossible to get a proper handle onto it, much like on the entire remerge discussion in the first place 19:15:31 thess: well, at least some of us have a histroy with it 19:15:49 thess: it's also brand name / recognition. everyone knows openwrt, but lede is "new" 19:15:49 thess: and at least some $corp had made an investment, which they are now worried about 19:16:06 thess: for me its nostalgia, a more or less strong brand name and community resources 19:16:53 but personally I do not really care about the brand 19:17:06 brand name isnt as important as most people think, people who use openwrt also know about lede, most of them. 19:17:17 thats my perception as well 19:17:33 So our choices seem to be: attract the resources, or get openwrt to be organized and managed as LEDE 19:17:46 I guess it is more about aligning with the community, everything else is replacable (in my opinion) 19:18:35 having both openwrt and lede also splits contributions; NXP chose openwrt for their initial submission of their amd64 platform 19:18:35 it would be nice to get it resolved in one way or the other 19:18:47 we have a long history after all 19:19:59 KanjiMonster: sure but is there someone acting upon those submissions? 19:20:05 as an external contributor who is not so activae in the commonity it is now hard to decide where to contribute 19:20:34 jow: wigyori is at least merging pull requests on github 19:21:22 ok 19:21:48 so what shall we do in adition to what we already did? 19:21:58 simply ask them to join us, thne share the name? 19:22:24 we can keep telling eachother that we would like to meet and discuss but I see very little momentum 19:22:49 It would help us going forward if we had a Wiki & Forum. Who manages the openwrt ones now? 19:23:03 and more importantly I do not even see any indication of discussions similar to ours going on on the openwrt side 19:23:17 last week I asked in the thread to find a date for a audio meeting 19:23:55 but got now response so far: http://comments.gmane.org/gmane.comp.embedded.openwrt.devel/40938 19:23:56 Title: OpenWrt Development List (at comments.gmane.org) 19:24:02 thess: I managed the oepnwrt wiki until recently and helped zoltan to migrate 19:24:14 that was the reason for my question if anybody is intrested in a merge 19:24:21 thess: the forum is administered by imre iirc 19:24:23 since wigyori seems to run openwrt now, maybe we just need to coordinate with him 19:24:48 (but I guess the final blocker for some things remains kaloz ...) 19:24:56 at least he is the most active person there 19:25:10 KanjiMonster: you mean because kaloz is mia? 19:25:44 thess: we do have a copy of the openwrt wiki data lying around on our download server 19:25:48 I do not want to wait for kaloz forever 19:26:04 jow: assuming there are some adminstrative things for the domain/servers for which noone else has access/the rights 19:26:15 we could also try to talk to mikwe 19:26:17 #*mike 19:26:58 thess: but not sure if we really want to host a wiki copy with s/OpenWrt/LEDE/ 19:27:17 Wasn't suggesting that! 19:27:29 jow: I guess that would be a viable workaround 19:28:58 thess: that was partly joking; its just that any kind of wiki we'd set up would contain mostly contain redundant info 19:29:20 and I didn't want to maintain the authorative documentation in the wiki but rather in git 19:29:41 so what remains is the community wiki contents and those are already well served by the openwrt wiki 19:30:33 similar to the forum, the topics discussed there are already openwrt/lede universal, its just running under the openwrt domain 19:30:50 so it likely makes no sense to start a different competing forum with the very same topic scope 19:31:49 Any way to cross-reference without pissing people off? Are the OpenWrt forum moderators amenable to LEDE topics? 19:32:36 thess: lede users on the owrt forum started to refer to usign "the drak side" because they get flamed ;) 19:32:38 that would be the big question 19:32:59 a cheap diplomatic solutiopn would be a LEDE subforum or something 19:33:08 and let openwrt treat us as a community build 19:33:15 The Wike has too much valuable data to loose. The dev doc could use some updates, etc. 19:33:25 s/Wike/Wiki/ 19:34:10 i am afraid i need to leave now 19:34:19 l8r 19:34:36 thess: gn8 19:34:50 me too, shall we summarize hise point on the public list? maybe we can facilitate some more feedback there? 19:35:13 s/hise point/this points/ 19:35:18 yes jow will you do this? 19:35:28 yes but probably not before tomorrow 19:35:35 mo problem 19:35:48 Yes, we could try a further public discussion. 19:36:27 #action Try to work out more details regarding the LEDE/OpenWrt coexistence on the public list 19:36:34 jow: Your points (topics) in a old mail message may be a good start 19:37:07 can I close the meeting log? 19:37:11 OK 19:37:13 yes 19:37:19 #endmeeting