| jjardon | paulsherwood: true, will make a new patch | 00:37 |
|---|---|---|
| *** inara has joined #baserock | 03:34 | |
| *** inara has quit IRC | 03:39 | |
| *** zoli__ has quit IRC | 03:42 | |
| *** zoli__ has joined #baserock | 03:42 | |
| *** zoli__ has quit IRC | 03:46 | |
| *** inara has joined #baserock | 03:50 | |
| *** inara has quit IRC | 03:56 | |
| *** inara has joined #baserock | 04:30 | |
| *** zoli__ has joined #baserock | 07:20 | |
| *** zoli__ has quit IRC | 07:24 | |
| *** zoli__ has joined #baserock | 07:47 | |
| *** zoli___ has joined #baserock | 07:49 | |
| *** zoli__ has quit IRC | 07:52 | |
| *** zoli___ has quit IRC | 07:52 | |
| *** zoli__ has joined #baserock | 07:54 | |
| *** zoli__ has quit IRC | 08:16 | |
| *** zoli__ has joined #baserock | 08:16 | |
| *** zoli__ has quit IRC | 08:20 | |
| robtaylor | paulsherwood: systemd keeps getting bigger and bigger - getting slow to develop o! | 09:47 |
| paulsherwood | robtaylor: i wonder if it's just that, though previously (http://wiki.baserock.org/build-times/) i was seeing ~3 minutes. | 09:55 |
| radiofree | We're also building it -j1 which doesn't help | 10:04 |
| paulsherwood | that would explain it | 10:10 |
| paulsherwood | actually, it wouldn't. how come i could log build times in oct 2014 of 3mins, but systemd has had max-jobs=1 since august? | 10:13 |
| *** zoli__ has joined #baserock | 12:58 | |
| robtaylor | radiofree: why is it j1? I refgularly build head at -j8 | 13:12 |
| robtaylor | paulsherwood: systemd has got a lot bigger and slower to build recently. not *entirely* sure why, but hcakcing on it has been more painful these lsast two months | 13:16 |
| robtaylor | - assuming its been updated recently in br | 13:17 |
| robtaylor | (might be test additions? - seems a fair amount of time is spend linking tests) | 13:22 |
| robtaylor | paulsherwood: 30 mins at j1 seems right | 13:26 |
| robtaylor | time make -j8 here just gave me | 13:27 |
| robtaylor | real12m46.856s | 13:27 |
| robtaylor | user35m3.940s | 13:27 |
| robtaylor | sys1m52.657s | 13:27 |
| radiofree | robtaylor: there's a bizarre issue where doing anything > 1 causes the build to fail | 13:28 |
| radiofree | specifically with the g-ir-* stuff | 13:28 |
| radiofree | trys to use a file that hasn't been generated yet... | 13:29 |
| radiofree | only ever see that in baserock though | 13:29 |
| robtaylor | radiofree: there's g-ir-* stuff in systemd? | 13:33 |
| radiofree | robtaylor: gudev | 13:38 |
| robtaylor | should really get to the bottom of tht one, i remember ssam had a good luck at it but couldn't see why | 13:39 |
| robtaylor | maybe something do with l-u-c | 13:39 |
| *** petefoth has quit IRC | 13:54 | |
| paulsherwood | can i get another +1 on the pyinotify + fail2ban? | 14:03 |
| robtaylor | paulsherwood: not sure if my +1 has any value but you can have one from me | 14:04 |
| paulsherwood | i'll take it :) | 14:04 |
| paulsherwood | tvm | 14:04 |
| robtaylor | heh, np | 14:05 |
| * robtaylor is experimenenting with dolt and systemd, may send a patch upstream | 14:05 | |
| robtaylor | paulsherwood: you may want to have a try at adding --enable-gold=default --enable-threads to pinutils configure | 14:08 |
| robtaylor | paulsherwood: (and run a new bootstrap) | 14:09 |
| robtaylor | s/pin/bin | 14:09 |
| paulsherwood | robtaylor: you mean to improve speed for systemd build, or something else? | 14:10 |
| robtaylor | paulsherwood: yep, that should speed up all link steps, probably quite dramitically | 14:11 |
| paulsherwood | in all 3 binutils? | 14:12 |
| * paulsherwood gives it a shot | 14:16 | |
| *** CTtpollard has joined #baserock | 14:19 | |
| *** CTtpollard has quit IRC | 14:22 | |
| robtaylor | paulsherwood: probably =) | 14:29 |
| paulsherwood | http://paste.baserock.org/xipezohani | 14:30 |
| paulsherwood | i'll just try binutils, not the bootstraps | 14:30 |
| robtaylor | hmm, interesting | 14:32 |
| paulsherwood | i'm starting to wonder wtf i'm building :) | 14:34 |
| robtaylor | well, binutils has two linkers in, the old 'ld' and new 'gold'. gold is faster in a bunch of ways and also can use threads when linking | 14:37 |
| robtaylor | --enable-threads enables thread usage in gold | 14:37 |
| robtaylor | --enable-gold=default installs gold as /usr/bin/ld | 14:37 |
| robtaylor | its possible some things may fail - i'm not sure if the whole wold is gold-linkable yet | 14:38 |
| robtaylor | if that does happen, then we'd have to so --enable-gold=default --enable-ld | 14:39 |
| robtaylor | and for things that fail, you'd add LD=/usr/bin/ld.bfd before the ./configure | 14:39 |
| * paulsherwood is first validating that he can build stage1-binutils and stage2-binutils unmodified | 14:41 | |
| robtaylor | paulsherwood: good call | 14:42 |
| *** nowster has quit IRC | 14:45 | |
| *** paulsherwood has quit IRC | 14:45 | |
| *** kejiahu has quit IRC | 14:45 | |
| *** nowster has joined #baserock | 14:45 | |
| *** kejiahu has joined #baserock | 14:45 | |
| *** paulsherwood has joined #baserock | 14:45 | |
| robtaylor | paulsherwood: i think one reason compile times have increased is that now LTO may be enabled by the gcc update? | 15:01 |
| robtaylor | "The current implementation only produces ?fat? objects, effectively doubling compilation time and increasing file sizes up to 5x the original size. This hides the problem that some tools, such as ar and nm, need to understand symbol tables of LTO sections. These tools were extended to use the plugin infrastructure, and with these problems solved, GCC will also support ?slim? objects consisting of the intermediate code alone. " | 15:03 |
| robtaylor | from https://gcc.gnu.org/onlinedocs/gccint/LTO-Overview.html | 15:03 |
| paulsherwood | urgh!? | 15:05 |
| robtaylor | yeah, its enabled becasue it does produce faster binaries at the end, but it may not be the tradeoff you want in development | 15:07 |
| robtaylor | (and smaller binaries) | 15:07 |
| robtaylor | - the file sizes that are increased are the intermediates | 15:08 |
| * robtaylor finds out that systemd forces use of gold already | 15:09 | |
| robtaylor | dolt seems to gain me 25% though | 15:09 |
| paulsherwood | putting LD=/usr/bin/ld.bfd before configure doesn't seem to work | 15:13 |
| paulsherwood | http://paste.baserock.org/mihejubipo | 15:13 |
| robtaylor | paulsherwood: that's no failing becuase its havning problems with gold, it looks more like for some reason there is no /usr/bin/ld | 15:17 |
| robtaylor | paulsherwood: can you pstebin th config.log? | 15:18 |
| paulsherwood | where does morph put that? | 15:23 |
| paulsherwood | in other news, can't seem to chroot into the failed area... 'chroot: can't execute '/bin/sh': No such file or directory' | 15:25 |
| robtaylor | paulsherwood: in the build directory | 15:32 |
| robtaylor | paulsherwood: (were ever your configure was run) | 15:33 |
| robtaylor | paulsherwood: you cant chroot inro stage 1 (and poss stage 2) | 15:33 |
| paulsherwood | right | 15:35 |
| paulsherwood | robtaylor: wasn't http://paste.baserock.org/mihejubipo the config log? | 15:36 |
| robtaylor | paulsherwood: no | 15:40 |
| robtaylor | paulsherwood: its a file called config.log in your build director | 15:40 |
| robtaylor | directory | 15:40 |
| *** inara has quit IRC | 15:41 | |
| *** inara has joined #baserock | 15:43 | |
| paulsherwood | no such file, robtaylor | 15:46 |
| robtaylor | hmm | 15:48 |
| robtaylor | thatr is odd | 15:48 |
| *** inara has quit IRC | 15:50 | |
| paulsherwood | i'm tempted just to raise a story about slow systemd build and leave it at that | 15:54 |
| *** inara has joined #baserock | 16:01 | |
| robtaylor | paulsherwood: yep | 16:12 |
| * robtaylor is bemused that runninmg configure hasn't generated a config.log | 16:12 | |
| paulsherwood | i'll try it again later | 16:22 |
| * paulsherwood is wondering how systemd services are added to baserock systems | 16:23 | |
| * paulsherwood has found http://wiki.baserock.org/guides/systemd/ - which doesn't mention how to do this for baserock systems at all | 16:25 | |
| *** zoli__ has quit IRC | 16:40 | |
| *** zoli__ has joined #baserock | 16:40 | |
| *** zoli__ has quit IRC | 16:45 | |
| Kinnison | paulsherwood: essentially in the same way any other distro would -- you put unit files into the right place in the filesystem (off the top of my head I can't remember exactly where). | 19:12 |
| paulsherwood | Kinnison: thanks | 21:36 |
| * paulsherwood has discovered there are examples in morph files http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/definitions.git/tree/strata/zookeeper/zookeeper-server.morph | 21:38 | |
| pedroalvarez | More examples: http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/definitions.git/tree/distbuild/usr/lib/systemd/system | 21:41 |
| paulsherwood | ah, interesting. does that mean we have at least two different approaches for this? :) | 22:01 |
| *** ChanServ has quit IRC | 22:02 | |
| *** ChanServ has joined #baserock | 22:04 | |
| *** wolfe.freenode.net sets mode: +o ChanServ | 22:04 | |
| pedroalvarez | There is a third approach. And is include them in the chunk repository. | 22:12 |
| *** pacon has joined #baserock | 22:31 | |
| *** jjardon_ has joined #baserock | 22:50 | |
| *** jjardon has quit IRC | 22:50 | |
| *** JPohlman1 has quit IRC | 22:50 | |
| *** persia_ has quit IRC | 22:50 | |
| *** JPohlmann has joined #baserock | 22:50 | |
| *** JPohlmann has joined #baserock | 22:50 | |
| *** persia_ has joined #baserock | 22:50 | |
| *** persia_ has joined #baserock | 22:50 | |
| *** jjardon_ is now known as jjardon | 22:50 | |
| *** inara has quit IRC | 23:12 | |
| *** inara has joined #baserock | 23:13 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!