| *** tristan has quit IRC | 06:14 | |
| *** tristan has joined #buildstream | 07:52 | |
| *** ChanServ sets mode: +o tristan | 07:53 | |
| ironfoot | hep | 09:05 |
|---|---|---|
| ironfoot | the build i left running is full of errors like this: https://paste.gnome.org/p1cnc63b5 | 09:05 |
| ironfoot | tristan: I can confirm that the new implementation of circular deps check works as fast as my wrong solution | 09:13 |
| *** tiagogomes has joined #buildstream | 09:23 | |
| *** csoriano has joined #buildstream | 09:26 | |
| *** tiagogomes has quit IRC | 09:28 | |
| *** tiagogomes has joined #buildstream | 09:50 | |
| *** ssam2 has joined #buildstream | 10:08 | |
| tristan | ironfoot, yeah, I'm pretty sure that is due to https://gitlab.com/tristanvb/buildstream/issues/4 | 10:16 |
| tristan | if you remember, I think that's what happens when you try to stage an artifact which doesnt yet exist | 10:16 |
| tristan | I might be wrong but pretty sure | 10:16 |
| *** csoriano has quit IRC | 10:36 | |
| *** csoriano____ has joined #buildstream | 10:48 | |
| tristan | Ok you guys have me muted | 10:51 |
| tristan | oop wrong channel | 10:51 |
| *** csoriano____ has quit IRC | 12:18 | |
| *** csoriano____ has joined #buildstream | 12:43 | |
| *** csoriano____ has quit IRC | 12:56 | |
| tristan | ironfoot, you did not push the latest defs2best output to buildstream-tests did you ? | 13:13 |
| ironfoot | tristan: just pushed | 13:17 |
| * tristan pulls | 13:17 | |
| tristan | mkay, I really have to spend some time on that UI, now I see that it's way overly verbose in some ways | 13:19 |
| tristan | ironfoot, were you able to complete a build of core.morph ? | 13:23 |
| tristan | core.bst rather ? | 13:23 |
| * tristan sees that it's building quite nicely *so far* | 13:23 | |
| tristan | and the conversion looks, pretty correct ! | 13:24 |
| ironfoot | :) | 13:24 |
| ironfoot | i haven't tried to build this latest version | 13:24 |
| tristan | I think it might complete, if not its just because we need better regexers for the build instructions | 13:25 |
| tristan | right now we're only substing $DESTDIR and $PREFIX | 13:25 |
| tristan | ironfoot, built up to flex now, blocking on source downloads :) | 13:36 |
| tristan | openssl-new is gonna fail, or it's output will be wrong :) | 13:37 |
| tristan | Has: --prefix="${PREFIX-/usr}" | 13:38 |
| tristan | Well, it's very close to being a fully functional conversion ! | 13:38 |
| *** csoriano____ has joined #buildstream | 13:39 | |
| tristan | ironfoot, hah, openssl passed because /usr happens to be the right prefix anyway, finished building curl and xz, downloading git-minimal, libtool, patch... | 13:47 |
| tristan | ironfoot, yeah confirmed, the error you linked is related to issue 4, and the first line of *that* error is showing that it's trying to stage pkg-config, but I'm certain that the build of pkg-config failed | 14:12 |
| *** csoriano____ is now known as csoriano | 14:25 | |
| tristan | ironfoot, issue #4 is fixed btw | 14:29 |
| tristan | ironfoot, this is a branch off of your branch of defs2bst: https://gitlab.com/tristanvb/defs2bst/tree/pedro-changes :) | 14:40 |
| tristan | adds a commit, which I think results in a building core.morph :) | 14:41 |
| ironfoot | this is exciting | 14:42 |
| tristan | e2fsprogs fails... | 14:45 |
| tristan | weird, I think mkdtemp is failing in the sandbox | 14:47 |
| tristan | maybe we dont have writable /tmp or smth | 14:47 |
| tristan | nope, that works | 14:49 |
| tristan | hmm | 14:49 |
| * tristan will look at it from home | 14:50 | |
| tristan | commands seem to work from `bst shell -b /build/directory elements/core/e2fsprogs.bst` | 14:51 |
| *** tristan has quit IRC | 14:54 | |
| *** tristan has joined #buildstream | 15:09 | |
| *** ChanServ sets mode: +o tristan | 15:11 | |
| tristan | ironfoot, turns out to be really silly, only failed because /buildstream/install was not already created (seems most things dont require the install dir exist) | 15:13 |
| tristan | autoconverted core.morph built ! | 15:13 |
| tristan | Now to build gnome.bst... | 15:13 |
| tristan | We need to solve a naming issue in buildstream I think | 15:13 |
| tristan | currently we have that requirement that all $(basename path/to/file.bst) be unique | 15:14 |
| tristan | and it chokes on a gtk3.bst conflict | 15:14 |
| tristan | I think we should just encode the full path in the name | 15:14 |
| ironfoot | I've noticed that one too | 15:18 |
| ironfoot | it's confusing, given that says the "name" is duplicated | 15:18 |
| ironfoot | if that "name" (filename?) is being used internally by BS, maybe we can generate one based on the entire path? | 15:19 |
| ironfoot | path_to_file.bst | 15:19 |
| ironfoot | or even with "/"s | 15:19 |
| tristan | ironfoot, that's exactly what I just said :) | 15:20 |
| tristan | hehe | 15:20 |
| ironfoot | meh, monday! | 15:20 |
| tristan | Now, with 2 one line changes, bst show gnome/strata/gnome.bst... completes | 15:20 |
| tristan | The UI is uglified by it (names get really long) | 15:21 |
| tristan | but I have to fix the UI anyway | 15:21 |
| tristan | It also has the side effect, that I have to rebuild the world | 15:21 |
| ironfoot | on the other hand, some of those strata, contain only one chunk | 15:21 |
| ironfoot | so could be replaced | 15:21 |
| tristan | Not because it changes cache keys, but the artifact lookups are changed by that | 15:21 |
| tristan | ironfoot, no, I've been wanting to correct this anyway | 15:21 |
| tristan | ironfoot, one of the huge pain in the a$$ I have about ybd / definitions and its "name", is that I see people wasting time on work that shouldnt be needed | 15:22 |
| tristan | So we dont have a name: field | 15:22 |
| tristan | And we only identify elements by their filename | 15:23 |
| tristan | No need to block patches or see people scrambling around because "Need to make this filename match the chunk name ! cause it achieves nothing but its PRETTIER!" | 15:23 |
| * ironfoot has just hit "e2fsprogs" build failure | 15:24 | |
| * ironfoot sees new commit in BS | 15:25 | |
| tristan | ironfoot, master has fix | 15:25 |
| tristan | yep | 15:25 |
| * tristan launches a build of gnome.bst :D | 15:25 | |
| ironfoot | :D | 15:25 |
| ironfoot | Benchmarking soon? | 15:26 |
| tristan | I suspect that as it stands, we can wrap up a bsp with core and try deploying an FS image | 15:26 |
| tristan | I have things to do before benchmarks, I dont want to treat benchmarks lightly either | 15:26 |
| tristan | i.e. if you mean benchmarking huge builds and testing different configurations of load balance | 15:27 |
| ironfoot | ha! core.bst built here too | 15:27 |
| ironfoot | I meant something simple, like building gnome.bst, and comparing | 15:27 |
| tristan | thats still a pretty big build but yeah; I think space for actually doing benchmarks is required | 15:28 |
| tristan | i.e. mental space, to do "just that" for at least a week | 15:28 |
| ironfoot | anyway, congrats tristan! this is looking very good | 15:28 |
| * tristan just spent an hour or two this evening and solved all the blocker bugs | 15:29 | |
| tristan | or the serious ones | 15:29 |
| tristan | I still wont have time to actually do this until next week I presume, end of feb delivery is demanding focus | 15:29 |
| tristan | heh, /me is building 3 things and fetching 4 | 15:37 |
| tristan | prefetches rock | 15:37 |
| ironfoot | hehe until you prefetch twice the same :P | 15:40 |
| tristan | ironfoot, yeah well the approach for issue 3 will be alright for that | 15:40 |
| tristan | it's also *very* rare | 15:40 |
| tristan | although it will happen for the kernel, and for python | 15:40 |
| tristan | and python is also huge | 15:41 |
| ironfoot | interesting build error: https://paste.gnome.org/p0sdn9o7o | 15:42 |
| tristan | ironfoot, didnt convert the submodules correctly for findutils maybe ? | 15:44 |
| tristan | no, well it *looks* correct | 15:45 |
| tristan | have we built *anything* that has a submodule ? | 15:45 |
| tristan | looks like it ignored the submodules though | 15:46 |
| ironfoot | looks like some from core.bst have | 15:46 |
| ironfoot | and log for "patch" build looks correct | 15:47 |
| tristan | weird | 15:48 |
| tristan | yeah I see bison passed and it has exactly the same | 15:48 |
| tristan | ironfoot, wanna compare that to a baserock build of findutils ? | 15:49 |
| tristan | ironfoot, I have a suspicion about ./bootstrap causing this | 15:49 |
| tristan | looks like ./bootstrap may be ignoring the fact we already checked out a gnulib from somewhere else | 15:49 |
| tristan | and ybd does something to make it "ok" that we've missed | 15:50 |
| ironfoot | but log doesn't say it was checked out: https://paste.gnome.org/p3tel9wez | 15:50 |
| ironfoot | for bison I can see: | 15:51 |
| ironfoot | [00:00:01] SUCCESS [bison-0] : Staging git://git.baserock.org/delta/gnulib | 15:51 |
| tristan | hmm, you are right, it didnt get staged | 15:54 |
| tristan | ironfoot, I also got: https://paste.gnome.org/p8plmks1t/vg7tau/raw | 16:21 |
| ironfoot | eugh | 16:22 |
| tristan | looks like for some reason ldd is in the wrong place, or missing a symlink (probably because build-essential was a bit more hands on) | 16:22 |
| *** xjuan has joined #buildstream | 16:25 | |
| tristan | ironfoot, anyway this is nice to see when it has colors: https://paste.gnome.org/pyu7p892a/gtset6/raw | 16:25 |
| tristan | looks like depth sorting is done well and prefetches are in order :) | 16:26 |
| ironfoot | !! | 16:26 |
| * tristan wont debug these problems today, and it's hammering all my cores, but I'll run `bst --fetchers 10 fetch gnome.bst` in the background :) | 16:27 | |
| *** jude has joined #buildstream | 16:34 | |
| jjardon | tristan: you have access to the baserock gitlab runners, right? | 17:01 |
| *** jude has quit IRC | 17:15 | |
| *** csoriano has quit IRC | 18:40 | |
| *** csoriano has joined #buildstream | 18:55 | |
| *** ssam2 has quit IRC | 19:31 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!