06:10 <jbjnr> I've just noticed that someone asked about pmix and I was pinged. Is this still important (Wish we could use something like slack instad of irc, so I'd see this stuff)

06:11 <jbjnr> mdiers[m], and ms[m] I guess were asking

06:12 <jbjnr> fwiw - pmix is the x version of pmi the process management interface that mpi uses to startup. when N mpiranks are launched, they need to know how to connect to each other and pmi is the library used for that. if you have an error in pmi, it usually means that mpoi is somehow misconfigured

06:13 <zao> I have several opinions on the clusterhug that is PMI, PMI2, PMIx, and UCX :D

06:13 <jbjnr> hard to know what is actually wrong without hands on the machine and testing

06:13 <zao> Be happy you're not using Intel MPI, every releases has a new implied dependency on particular versions of UCX and PMI, requiring you to specify all sorts of env-vars and pray.

06:14 <jbjnr> I feel the pain

06:14 <jbjnr> I have similar problems

06:14 <zao> I thankfully have an old beard handling this at my site, but my personal build box is screwed.

06:15 <zao> Extra fun when you involve SLURM as it also deals with PMI to coordinate rank spawning.

06:15 <jbjnr> That's very sad. I think I might be an old beard now

06:16 <jbjnr> (Now that I have a beard, and I'm old)

07:36 <mdiers[m]> <jbjnr "I've just noticed that someone a"> thanks, my problem has been solved. was a missing name resolution in the container.

07:36 <jbjnr> ok

11:13 <zatumil> sphinx-build assert parentSymbol fails in sphinx/domains/cpp.py, line 7375 with git clone git://github.com/stellar-group/hpx --branch 1.5.0 && cmake hpx -DHPX_WITH_DOCUMENTATION=ON && make docs

11:17 <zao> If it matters, what versions of the tools do you have?

11:27 <zatumil> # Sphinx version: 3.2.1# Python version: 3.8.5 (CPython)# Docutils version: 0.14 # Jinja2 version: 2.10

11:30 <zatumil> http://paste.debian.net/plain/1162606

11:44 <zao> The build_env container seems to have Sphinx 3.1.0, Python 3.7.5, docutils 0.16, Jinja2 2.11.2

11:45 <zao> Not sure if any of those are significant differences.

11:47 <zao> Who is the resident docs wrangler, ms[m] ?

11:48 <ms[m]> zao, zatumil yeah, I suppose that's me

11:49 <ms[m]> I haven't made a very big effort to check with which versions the docs build... zatumil if you want an environment that works for sure the only one I can offer is the versions in our ci docker image

11:49 <ms[m]> this worked at the time that the image was last built: https://github.com/STEllAR-GROUP/docker_build_env/blob/aeb508895b6a4b17ec80bc3f7fe925933181746b/hpx_build_env/Dockerfile#L60

11:49 <ms[m]> but the sphinx version may have changed

11:50 <ms[m]> I'd first try breathe==4.16.0 and if that doesn't work I'll dig up the actual versions of sphinx etc. that were installed in the image

11:50 <zao> (I've pulled build_env:ubuntu and am trying the docs build right now)

11:50 <zao> (in singularity, but should be the same :D )

11:51 <ms[m]> the image is stellargroup/build_env:ubuntu btw

11:51 <ms[m]> yep ^

11:51 <zao> I got past the libs/static_reinit/api one that expoded.

12:07 <ms[m]> \o/

12:13 <zao> Installed 3.2.1 in there, still built.

12:32 <ms[m]> so it's breathe that's still causing problems...

12:35 <zao> Upgraded breathe, broke it.

16:01 <bita> hkaiser, do you have time now to join the meeting?

16:03 <ms[m]> hkaiser: btw, I meant to ask ages ago but always forgot: is the reason that the tag_invoke "member functions" for the various CPOs are friend functions just so you get to define them inside the struct definition?

16:09 <hkaiser> ms[m]: that's so they don't bleed into the outer namespace keeping lower the number of overloads the compiler has to deal with

16:15 <ms[m]> hkaiser: huh, ok, cppreference was not entirely clear on what happens when you do that :P I got the impression that it would be equivalent to defining it outside the class in the same namespace, but I had missed the visibility aspectes (cppreference doesn't contradict that though, and considering they're private it makes sense)

16:16 <hkaiser> ms[m]: yes, friend functions are injected into the outer namespace, but I think the lookup starts with the scope of the CPO so it will find those earlier

16:19 <ms[m]> oh c++... thanks hkaiser

16:26 <K-ballo> they aren't private if they are friends

17:12 <gnikunj[m]> hkaiser: yt?

17:20 <K-ballo> this is running on rostam, right? https://cdash.cscs.ch/buildSummary.php?buildid=125421

17:21 <K-ballo> does it build the PR, or does it build the merge?

17:23 <parsa> K-ballo: according to the commit number looks like it's building the PR

17:24 <K-ballo> good catch!

17:45 <hkaiser> gnikunj[m]: here

17:45 <gnikunj[m]> hkaiser: see pm please

17:45 <hkaiser> gnikunj[m]: I can't meet at 1pm today, sorry

17:46 <gnikunj[m]> we had a meeting today as well?

17:46 <hkaiser> just saying ;-)

17:46 <gnikunj[m]> :D

20:34 <parsa> hkaiser: i wish i could write that fast

20:34 <parsa> :)

20:44 <hkaiser> parsa: I'm almost done

20:49 <hkaiser> parsa: does the text make any sense to you?

20:49 <hkaiser> I'm afraid it's a bit too technical

20:50 <hkaiser> parsa: I think I'm done

20:50 <parsa> it's great. i don't think we should touch it for now

20:51 <parsa> many thanks!

20:51 <parsa> i might need to add one or two paragraphs to set up a bit more of context

20:52 <parsa> a bit of a story for what futures are and what agas roughly does

20:52 <hkaiser> parsa: sure, please do