06:41 <simbergm> diehlpk_work: ah, shame, hope he feels better soon!

06:41 <simbergm> jgolinowski: awesome!

06:41 <simbergm> waits in suspense to see if it's jbjnr

06:45 <simbergm> it's not :)

06:45 <simbergm> jgolinowski: thanks for taking care of that, hope he gets it running

06:45 <simbergm> that person also opened an issue for hpx itself, he seems to be wanting to run it on arm

06:47 <simbergm> but our exception handling must still be somewhat broken if that's the output one gets for not having started the runtime

07:30 <jgolinowski> simbergm, yeah I was thinking that (about the error message)

09:06 <jbjnr> simbergm: yt?

09:15 <simbergm> jbjnr: here

09:16 <jbjnr> option 1 - do nothing (fine by me), opt 2 - make numa allocator a cmake option (too many options already), opt 3 waste a lot of time trying to find out why it fails on certain machines then go with opt 1 or 2 anyway. opt 4 can we disable it easily for some platforms some other way???

09:17 <jbjnr> opt 4 seems hard since we don't know why it fails.

09:17 <jbjnr> so tat leave opt 1 or 2. I'm leaning towards opt 2

09:18 <jbjnr> we add a HPX_WITH_EXPERIMENTAL_FEATURES (or similar) and put all the slightly avant-garde stuff under control of that

09:18 <jbjnr> simbergm: ^^

09:19 <simbergm> jbjnr: I think I agree with your options

09:20 <simbergm> 1 would be fine by me if we get the dashboard green

09:21 <simbergm> I like the flag for experimental features

09:22 <simbergm> it seems like ali is sick at the moment so we'd have to wait to see if we can change something on rostam

09:22 <jbjnr> It does raise the problem of when we move something out of experimental into main. if it's turned off on dashboards by default, we'll never really know when it's safe to move to main

09:23 <simbergm> with 4 we'd just have to make sure we still test those

09:23 <simbergm> exactly that, we can't have them experimental and then not test them, that'll just break without us noticing

09:24 <simbergm> I'd make the flags per feature though, not one global flag

09:24 <simbergm> possibly an additional global flag so that we can turn all experimental features on easily for testing

09:25 <simbergm> it would be a nice way of rolling things out without giving all the

09:26 <simbergm> guarantees about not breaking the api etc while something is still experimental

11:18 <jbjnr_> simbergm: the problem is that if we add an experimental feature flag, then we have to test i, then our dashboards are not green, so why bother - we can have non green dashboards by doing nothing

11:19 <jbjnr_> who's talking in the meeting today.I was preparing and doing course all week, so if you have something interesting you want to talk about, you do it.

11:24 <simbergm> jbjnr_: I can do it even though I don't necessarily have something interesting

11:25 <simbergm> good point about the flag, it doesn't help in making the dashboard green

11:25 <simbergm> let's ask ali what he can do

11:26 <simbergm> I'm also pretty okay with releasing the allocator as it is (even though the dashboard isn't entirely green) because it seems to still do something sane

11:26 <simbergm> i.e. it doesn't segfault or crash in some other way

11:26 <simbergm> it still gives a chunk of memory, just not quite where you want it

11:27 <jbjnr_> what we really need is a green dashboard, which means not running the test on the dashboard - is it work running the test on the dashboard, but exclucing it from the red/green status?

11:27 <jbjnr_> is it worth^

11:34 <simbergm> no

11:34 <simbergm> let's see if we can disable hugepages first

11:35 <simbergm> if we can't do that we'll need to change the test

11:35 <jbjnr_> I can make the test return pass if no segfaults happen, but then it isn't a very interesting test

11:36 <jbjnr_> it was cleverly designed to actually test the memory binding.

11:36 <jbjnr_> so it's doing it's job

11:36 <jbjnr_> I wonder if I balse up the test somehow

16:07 <aserio> jbjnr_: Will you be joining the meeting?

16:08 <aserio> diehlpk, diehlpk_work: ^^

16:08 <diehlpk> I will be late

16:08 <diehlpk> Can make it at 11;20

16:09 <diehlpk> aserio, Other meeting running late

16:23 <diehlpk> aserio will be there in 10min