07:56 <reynir> is the fortnightly thingie today?

08:05 <reynir> (yes)

14:54 <reynir> o/

14:57 <yomimono> hey reynir! :D

14:58 <reynir> hey yomimono!

14:58 <TImada> Hi all!

14:58 <yomimono> hi!

14:59 <yomimono> there's a puzzle an old text adventure that requires you to have, simultaneously, items called "tea" and "no tea"

15:00 <not-reynir> Hello from a mirage+qubes VM o/ (through qrexec) :D

15:00 <yomimono> cool!!

15:01 <not-reynir> Ea is a girl's name here, maybe they meant not-ea ^^

15:02 <yomimono> not-reynir: is your IRC client repo in a state where you want to publicize it?

15:02 <yomimono> if so, you should tell us where it is!

15:02 <not-reynir> Not really, but maybe I'll make a code dump just to get some feedback. There's some... ugliness I don't know how to handle better

15:03 <yomimono> you mentioned some functor applications you weren't happy with on masto IIRC

15:06 <yomimono> oh! also it's 16.05 BST on an alternate Wednesday so I think it's probably time for a catchup session

15:06 <yomimono> there is an agenda at https://github.com/mirage/mirage-www/wiki/Call-Agenda

15:06 <yomimono> it's not too late to add stuff to it!

15:06 <reynir> Yes. It's a functor that you have to pass a whole lot of modules to, but I guess mirage magic can help/hide this. But the last module is basically passing the Stack.t and such so the IRC library can connect

15:08 <yomimono> hm, maybe some discussion on mirageos-devel wouldn't go amiss for ideas on handling that

15:08 <yomimono> (I am the last person to give good advice on avoiding ugly functors)

15:09 <reynir> code dump: https://github.com/reynir/ocaml-irc-client/tree/mirage

15:11 <yomimono> reynir: the only thing that seems weird to me here is the T module, which seems to include only stuff you already get through other modules?

15:11 <reynir> yomimono: Hmm, I thought you didn't get the Pclock.t value from the Pclock module

15:12 <reynir> I think the Pclock.t value is passed to the Main.start function - maybe I missed something

15:13 <yomimono> I trust your memory more than mine, but I agree that there should be a nicer way to do that

15:13 <hannes> oh hi!

15:15 <yomimono> hi hannes!

15:15 <reynir> Regarding my item on the list: I think it would make sense if you could generate a .merlin file from a given config.ml in mirage. Besides being convenient, I think it's not always clear what packages will be used for your unikernel. You can of course check the generated opam file, but it will often include more packages AFAIK

15:16 <reynir> I can maybe attempt implementing this with some guidance ?

15:16 <hannes> reynir: imho yes, mirage merlin (!?) should be able to generate a .merlin just from config.ml!

15:17 <hannes> (there should imho as well be the possibility to generate a .ocamlinit from/after mirage configure -t YYY)

15:17 <yomimono> reynir: Pclock includes Mirage_device.S which includes a type t

15:18 <yomimono> also, I would like .merlin autogeneration :)

15:18 <hannes> reynir: I believe the merlin generation should be a separate phase (or how you call it), next to build and configure

15:19 <reynir> yomimono: Yes, it's a value of type t I'm missing - sorry, I was confusing

15:19 <hannes> which means that at least functoria (and maybe mirage) will need some added functionality

15:20 <reynir> Ah

15:20 <hannes> I'm happy to answer questions about the bits and pieces i understand there, though it's designed and implemented by drup who may have more insights

15:21 <reynir> What files would be a good place to start staring at? :)

15:22 <hannes> in functoria, app/functoria_app and app/functoria_command_line ... you could follow on along the lines how the .opam is generated..

15:25 <reynir> Thanks

15:29 <reynir> Ok, I don't really have more to say about that right now. I will probably ping you and/or Drup later :)

15:31 <yomimono> thanks, reynir! :)

15:32 <yomimono> nothing more on the agenda. anything else from anyone?

15:33 <yomimono> (if not: reminder that the next MirageOS hack retreat will be in early October, and more details are available at retreat.mirage.io)

15:34 <hannes> is anyone maintaining io-page?

15:34 <reynir> Oh, I just noticed that mirage.io seems to be a bit broken - some js doesn't load. The layout looks a bit funky as a result

15:34 <yomimono> hannes: I will have patches to it because of various hacking I'm doing in the Xen layer, so that question is relevant to me too :(

15:35 <reynir> Oh, also some css (makes more sense)

15:35 <yomimono> AFAICT the closest person to a maintainer is djs55 but I don't know whether his interest in that library is ongoing

15:35 <yomimono> reynir: hm, indeed!

15:35 <reynir> xen hacking \o/

15:36 <yomimono> reynir: yes, if you want to make unikernels that crash on boot, I'm your hookup

15:36 <hannes> yomimono: the build system is in a sad state, I suspect... I _think_ it should instead use whatever bigstringaf etc. does -- define a single library, and C stubs, and then some hacks to get xen+freestanding cross-compilation going on (checkseum/digestif use the same hacks iirc)

15:36 <yomimono> hannes: yes, I agree that the build system there needs some attention, it's quite the accretion of strange things at the moment

15:37 <hannes> but I still fail to understand how this whole building and linking works, e.g. in bigstringaf... on a freestanding/xen target we should somehow _avoid_ the for-unix-compiled C-stubs (but I've not enough knowledge on how to do that properly)

15:38 <yomimono> I think there are various features in dune which are supposed to solve this, but I don't understand them sufficiently to comment further

15:39 <yomimono> I only barely understand how the existing processes work

15:40 <yomimono> (although I do understand that well enough to document it at this point, I think, and perhaps that's worth doing)

15:40 <hannes> I'm close to installing my very first author-signed package \o/ (sorry, not very much MirageOS related)

15:40 <reynir> yomimono: what repos should I check out for crash-on-boot unikernels? :)

15:41 <reynir> hannes: \o/ \o/ \o/

15:41 <yomimono> hannes: but awesome!

15:41 <yomimono> reynir: if you're really interested I can give you a set of ~10 pins

15:42 <yomimono> and then you can experience the joy of seeing a PVH unikernel spew a bunch of debug output and then immediately crash before launching your application

15:42 <hannes> yomimono: while I understand what mirage - the command-line utility - is doing in respect to linking, I fail to understand how ocamlbuild/ocamlfind/dune play in here...

15:43 <yomimono> hannes: ah, sorry, thought you were talking about library/flag handling here, which there is supposed to be some help for in the form of variants somehow

15:44 <reynir> yomimono: I don't think I'll have too much time to hack on it, but if it's not too much trouble I'd like to check it out!

15:45 <hannes> yomimono: yes, that's what is done... using these ocamlbuild/ocamlfind flags "xen_linkopts" / "freestanding_linkopts", plus some pkg-config magic to retrieve a ld, and some additional ldflags...

15:45 <yomimono> reynir: I'll push my WIPs and send you an `opam pin` list

15:46 <hannes> but what I'm missing is whether "normal" stubs are embedded into cm(x?)a(s?) / how linking works out on unix with ocaml

15:46 <hannes> with the rise of opam2, we should also rethink the pkg-config stuff we accumulated, and maybe have switch-specific variables!?

15:47 <hannes> but maybe that is a topic for a different call..

15:48 <reynir> afk o/

17:55 <Drup> ah, just saw I have been HLd

17:56 <Drup> I agree with hannes, it should not be part of configure

17:57 <Drup> reynir: You'll have to find an appropriate verb :3

17:58 <Drup> As for the implementation, Look at how describe works, it should be pretty similar

18:25 <reynir> "mirage magic" because merlin and it's magic

18:25 <reynir> :D

18:25 <reynir> Thanks, Drup!

18:26 <Drup> You might have to implement a function in functoria_graph

18:26 <reynir> (I wasn't aware of "mirage describe")

18:26 <Drup> This is exciting, I think I have git ownership for the whole file right now

18:27 <reynir> uh-oh

18:30 <Drup> Oh, no, thomas added comments and hannes added result types :D