Irc discussions from #mirage on 27-07-2016

Written by unpurecamelbot
Classified under: irclog
Published: 2016-07-27 (last updated: 2016-07-27)

27-07-2016 14:58 dbuenzli @unpurecamelbot Bravo.

27-07-2016 15:00 yomimono Hello! It's time for our fortnightly catch-up..

27-07-2016 15:01 djwillia hi all!

27-07-2016 15:01 noddy let the catch-up....

27-07-2016 15:01 yomimono The agenda is available at https://github.com/mirage/mirage-www/wiki/Call-Agenda !

27-07-2016 15:01 noddy ... begin.

27-07-2016 15:01 yomimono Please feel free to add items that you'd like to discuss today.

27-07-2016 15:02 yomimono Here's a moment to wave if you want unpurecamelbot to record that you were here!

27-07-2016 15:02 yomimono waves

27-07-2016 15:02 avsm Group greets!

27-07-2016 15:02 noddy waves

27-07-2016 15:02 talex5 ~

27-07-2016 15:02 amirmc yo

27-07-2016 15:02 dbuenzli v

27-07-2016 15:02 djwillia o/

27-07-2016 15:03 GemmaG :)

27-07-2016 15:03 yomimono So first on the agenda is our recurring "quality and test" item.

27-07-2016 15:03 mattg waves

27-07-2016 15:04 yomimono Nobody's noted anything, but perhaps now's a good moment to mention that I broke all of our tests by releasing a version of functoria with an undeclared hard dependency.

27-07-2016 15:04 yomimono Oops.

27-07-2016 15:04 yomimono If dbuenzli or any other topkg expert could have a look at https://github.com/mirage/functoria/issues/67 , I'd appreciate it.

27-07-2016 15:04 talex5 Could we add some REVDEPS to detect that in future?

27-07-2016 15:05 avsm exciting news on that front

27-07-2016 15:05 Drup yomimono: it's not really your fault

27-07-2016 15:05 Drup I blame the topkg transition :>

27-07-2016 15:05 yomimono drup: me too :P

27-07-2016 15:05 avsm I now have a DataKit VM with GitHub integration working thanks to talex5 and samoht -- with a little easter egg here https://github.com/mirage/ocaml-github/pull/110 (look at the bottom)

27-07-2016 15:06 yomimono avsm: can you tell us why that's rad real quick?

27-07-2016 15:06 avsm So I'm working on checking that DataKit wont destroy all our repositories by using it on a few small repositories first, but then hooking in REVDEPS tests as the first thing it checks

27-07-2016 15:07 avsm the 50 minute limit is what was blocking REVDEPS before

27-07-2016 15:07 hannes github #110 is sexp support..?!??

27-07-2016 15:07 avsm This will let us write custom CI checks that: 1) directly interface with OPAM 2) do not have a 50 minute limit and 3) run on our infrastructure (so can deploy archive servers and so on) and 4) are multidistro

27-07-2016 15:07 avsm hannes: look at the "Some checks were not successful" box at the bottom

27-07-2016 15:07 yomimono I think avsm wants us to scroll all the way down and see that there's a continuous-integration/anil-testing/pr in addition to travis-ci and appveyor

27-07-2016 15:07 avsm DataKit makes it really easy to hook in custom stuff there

27-07-2016 15:08 yomimono so we might have non-terrible CI sometime soon :D

27-07-2016 15:09 avsm in fact, it live updates

27-07-2016 15:09 avsm keep an eye on that URL now...

27-07-2016 15:09 yomimono that's indeed exciting! it will be great when we can see some updates -- right now I just get a link to anil.recoil.org

27-07-2016 15:09 yomimono which is a nice website but a bit scant on details

27-07-2016 15:10 yomimono unless there's more on this, let's move on

27-07-2016 15:10 djwillia that's awesome! hello-hannes tests should be standard

27-07-2016 15:10 avsm that's pretty much it! more as we get it working

27-07-2016 15:11 yomimono great!

27-07-2016 15:11 yomimono next up we've got an item to discuss "ocamlfind predicates", which is really a discussion of this pr:

27-07-2016 15:11 yomimono https://github.com/mirage/mirage/pull/568

27-07-2016 15:12 yomimono I don't see pqwy here but maybe I should be looking for a different name on irc?

27-07-2016 15:12 noddy i already waved!

27-07-2016 15:12 yomimono oh! sorry!

27-07-2016 15:12 yomimono want to give us a quick summary?

27-07-2016 15:13 noddy yeah; i want to externalize dep resolution from the mirage tool a bit. we have a nice language where we can do stuff like separately compile against a particular interface, then link against a particular implementation later on. i'd like to use that in mirage.

27-07-2016 15:14 noddy note that this is less "linking tricks" pr and more "let's make ocamlfind know about targets, there's stuff it knows how to do" pr.

27-07-2016 15:14 avsm Its true that this does duplicate some ocamlfind functionality

27-07-2016 15:15 avsm and we do have a hard dependency on ocamlfind right now for everything else

27-07-2016 15:15 avsm so pushing the linking hacks into ocamlfind lets us focus on "migrate all ocamlfind functionality" in the future

27-07-2016 15:15 noddy it's also a dep hack; you can condition dep on a predicate in ocamlfind.

27-07-2016 15:16 avsm yeah, which could be useful to help us compile multiple C library variants, e.g. across solo5/xen

27-07-2016 15:16 dbuenzli Note that this solution actually scales with the number of backends and avoids solving problems by encoding them as package names.

27-07-2016 15:16 avsm I'm broadly in favour, with talex5's caution noted: we should encourage library authors to not depend on this functionality

27-07-2016 15:16 avsm as an abstraction mechanism

27-07-2016 15:17 noddy it's unclear to me where this can lead. it could certainly be abused.

27-07-2016 15:17 dbuenzli What are the exact dangers ?

27-07-2016 15:17 noddy it can also be nicely used to automate satisfaction of a single interface without stuff downstream being tainted by the same problem

27-07-2016 15:18 noddy people playing nasty linking tricks i guess

27-07-2016 15:18 dbuenzli It's not tricks.

27-07-2016 15:18 noddy my mind cannot anticipate all the horrid abuses of this people can come up with.

27-07-2016 15:18 avsm yeah. they're really a valid part of the toolchain from ocaml land

27-07-2016 15:18 dbuenzli It link time selection of implementation.

27-07-2016 15:18 dbuenzli its

27-07-2016 15:18 avsm and ocaml blocks you from the worst errors via CRC checks

27-07-2016 15:18 dbuenzli it's

27-07-2016 15:18 noddy in this usage, it is

27-07-2016 15:18 noddy and it this usage it's kosher imho

27-07-2016 15:19 hannes can we pass the CFLAGS this way as well?

27-07-2016 15:19 avsm the danger is in C libraries really -- should we decide on a convention where C archives have their own predicate names that are reserved for them?

27-07-2016 15:19 avsm hannes: not cflags, but we can pick a predicate name for a CFLAGS definition and always use that

27-07-2016 15:19 avsm but cant enforce it

27-07-2016 15:19 dbuenzli C libraries should be recorded in the archive implementations

27-07-2016 15:19 noddy we already have that convention

27-07-2016 15:19 noddy it's called xen_linkflags.

27-07-2016 15:19 noddy and freestanding_linkflags.

27-07-2016 15:19 avsm yeah

27-07-2016 15:20 noddy in this case, not too much is gained:

27-07-2016 15:20 noddy as ld is called directly (on all targets i checked), extra logic is needed to extract linkflags(predicate) from METAs and propagate that to the ld invocation

27-07-2016 15:20 noddy so at best, we standardize on a standard META field as opposed to inventing arbitraty ones

27-07-2016 15:21 noddy if the final linking is done via ocamlfind, then it becomes automagic

27-07-2016 15:21 noddy but still... i want ocamlfind to know about the target. it's useful.

27-07-2016 15:21 avsm I think it's worth the experiment, particularly as we continue solo5 merging

27-07-2016 15:21 talex5 Would be very good to have ocamlfind get the linker flags, not mirage, indeed.

27-07-2016 15:21 avsm we can always revert it if it goes wrong before the mirage3 release

27-07-2016 15:22 noddy it can't go wrong in itself

27-07-2016 15:22 dbuenzli Well I have a picture drawn in my notebook by pqwy in what the future looks without this, it's not pretty I can scan it if you want.

27-07-2016 15:22 noddy and if someone uses it for diabolical schemes... we can rant against them on twitter!

27-07-2016 15:22 dbuenzli (and it's for a library which shouldn't be tied to mirage, nocrypto)

27-07-2016 15:23 noddy ok, related --

27-07-2016 15:23 avsm ok, sounds like we have consensus on a merge then...

27-07-2016 15:23 noddy this pr in itself enables certain things, and one of those is shimming OS

27-07-2016 15:23 noddy i'd like to draft a signature for this and get it somewhere, say, in types

27-07-2016 15:23 noddy so that all implementations can depend on, and check against, the same sig

27-07-2016 15:24 hannes noddy: you mean integrating your mirage-os-shim back into mirage-types?

27-07-2016 15:24 noddy no, i mean spelling out the signature os several modules in OS and having that signature in mirage-types

27-07-2016 15:25 noddy and having implementations import that signature as opposed to either copying it, or symlinking (as xen->unix does in some places)

27-07-2016 15:25 talex5 Design question: should anyone except main.ml use OS directly?

27-07-2016 15:25 noddy i wrote i library that hooks into the event loop

27-07-2016 15:25 noddy s:i:a:

27-07-2016 15:26 avsm talex5: OS gets increasingly hard to maintain consistently as we add backends, so no

27-07-2016 15:26 avsm witness the lifecycle design conversation

27-07-2016 15:26 avsm just xen's suspend/resume model is sufficiently different from a hardware one

27-07-2016 15:27 talex5 avsm: OK. We should document that somewhere (if we didn't already).

27-07-2016 15:27 mattg when I was first looking at Mirage, OS seemed like it didn't fit with the rest of mirage-types and was confusing to me

27-07-2016 15:27 avsm yeah, agreed.

27-07-2016 15:28 yomimono concur, you're not the only one who's been confused

27-07-2016 15:28 avsm its an evolutionary thing as well: we started off with OS about 5 years ago and then slowly disentangled it into smaller module

27-07-2016 15:28 noddy in that case, big chunks need to be moved to mirage-types and provided as functor parameters

27-07-2016 15:28 avsm guess it's time to kick the remainder to the kerb

27-07-2016 15:29 djwillia (have to drop off early - sorry bout that - bye all!)

27-07-2016 15:30 talex5 https://github.com/mirage/mirage-platform/issues/170

27-07-2016 15:30 avsm move on then yomimono ?

27-07-2016 15:30 yomimono let's!

27-07-2016 15:31 yomimono next up is tracking issues for mirage 3.0

27-07-2016 15:31 yomimono avsm and gemmag, want to share what you've posted?

27-07-2016 15:31 yomimono I'll hit submit on my in-progress one in the meantime

27-07-2016 15:32 avsm looks like gemmag's issue got quite a lot of traffic in a short period of time

27-07-2016 15:32 GemmaG I added a list of smaller things we could tackle on mirage.io https://github.com/mirage/mirage/issues/569

27-07-2016 15:32 GemmaG which is quickly becoming a larger list :P

27-07-2016 15:32 avsm i've picked off some of the low hanging fruit (linked canopy from main page, etc) and am working on twiopam to generate changelogs more easily

27-07-2016 15:33 GemmaG I'm new to the issues etiquette, so apologies for mistakes :)

27-07-2016 15:33 avsm http://gist.github.com/bactrian has some early examples, need to skin them properly

27-07-2016 15:33 avsm GemmaG: all looks good to me, thanks for tracking the comments as well!

27-07-2016 15:33 avsm On the information architecture question, I've somewhat radically been learning the latest HTML wizardry again

27-07-2016 15:34 GemmaG As @hannes mentioned, it might require a bigger-picture thought process, but there are some quick and small changes that might make a huge difference

27-07-2016 15:34 avsm I wrote the current Zurb one in a couple of days a few christmases ago, so it needs a refresh quite sorely

27-07-2016 15:34 avsm i reckon make the quick changes now to bring things up to a reasonable level, and then we can whiteboard up a few ideas for what the new website should look like closer to the 3.0 launch

27-07-2016 15:35 avsm i'm keen to automate the library tracking as much as possible, it's just not sustainable by any manual means any more

27-07-2016 15:35 dbuenzli Note that one problem with automated tracking is that it can quickly become noise.

27-07-2016 15:37 avsm yes absolutely

27-07-2016 15:37 avsm we still need curated newsletters and such, like the OCaml weekly news

27-07-2016 15:38 avsm so the feed isn't necessarily for direct human consumption, but it needs to exist for someone (even if just the weekly news editor)

27-07-2016 15:39 yomimono personally, anything less noisy than watching opam-repository via github would be an improvement

27-07-2016 15:39 avsm seconds that

27-07-2016 15:39 avsm not that i can escape from having to watch it :P

27-07-2016 15:39 hannes as posted several moons ago to mirageos-devel, there's a bot watching tons of mirage repos https://github.com/miragebot.private.atom?token=ARh4hnusZ1kC_bQ_Q6_HUzQteEEGTqy8ks61Fm2LwA== is its rss feed..

27-07-2016 15:40 hannes (but certainly records changes, not releases)

27-07-2016 15:41 yomimono that's... also very noisy

27-07-2016 15:41 yomimono (sorry bot, it's not your fault :( )

27-07-2016 15:42 yomimono if nobody objects, I'd also like to get some eyes on https://github.com/mirage/mirage/issues/571 , which is the beginning of a tracking issue for mirage 30

27-07-2016 15:42 yomimono er, 3.0

27-07-2016 15:42 yomimono we're not skipping that many numbers

27-07-2016 15:42 hannes mirage 95!

27-07-2016 15:42 yomimono (there are large unlsted items which are already completed, among them solo5!)

27-07-2016 15:43 GemmaG I'm looking into options for a regular newsletter too - one that is more curated than the Weekly News, and combines all the info. Currently there are a few sources where things are missed. Suggestions and ideas welcome :)

27-07-2016 15:43 yomimono GemmaG - that would be a general OCaml newsletter, right?

27-07-2016 15:44 yomimono mentioning for the benefit of others here who might have more to add on that than a Mirage-specific one :)

27-07-2016 15:44 GemmaG Possibly yes - it started off being more of an OCaml Labs newsletter, but it makes sense to incorporate OCaml as a whole

27-07-2016 15:44 dbuenzli Well I think caml weekly news does a good job at that.

27-07-2016 15:45 dbuenzli Especially the ocaml mailing list is very well curated

27-07-2016 15:45 avsm yeah, so something a bit more mirage specific makes sense on those lines

27-07-2016 15:45 dbuenzli (and nowadays we have the blog posts of the planet + outsanding PR to the ocaml compiler)

27-07-2016 15:45 avsm the caml weekly news is fantastic

27-07-2016 15:46 avsm I find the Planet blog posts to be the opposite of the Caml Weekly News: automated noise that often have little to do with OCaml or Mirage

27-07-2016 15:46 dbuenzli I mean we have the blog posts of the planet in the caml-weekly news ( I don't know if they are curated though)

27-07-2016 15:47 GemmaG I found a lot of content from other sources that aren't in the weekly newsletter and the planet blog - it would be good to incorporate them too

27-07-2016 15:47 avsm so a "social" version of the Caml weekly news with interesting blog posts/links would be great, or offering to help Alan collate those and add them to his newsletter

27-07-2016 15:47 GemmaG Yes - I think getting in touch with Alan is a good plan

27-07-2016 15:47 avsm Sounds good to me! Happy to help hook in data feeds for this

27-07-2016 15:48 GemmaG July round up so far: https://gist.github.com/GemmaG/c0bc1495e521c2b3006d21e6b8f92b25

27-07-2016 15:49 GemmaG There's a lot mentioned on Twitter for example that isn't part of the weekly news/planet blogs. I don't want to repeat content and add to noise, but having one place for ALL the info would be good

27-07-2016 15:49 GemmaG :)

27-07-2016 15:49 dbuenzli But twitter is for brain farts.

27-07-2016 15:49 GemmaG hahahah :P

27-07-2016 15:49 yomimono we've gotten a bit off-track, so maybe I could suggest that anyone with more additions contact GemmaG out-of-band?

27-07-2016 15:49 avsm I eat a lot of beans

27-07-2016 15:49 avsm good plan yomimono

27-07-2016 15:49 yomimono dbuenzli: also very important cat photos?

27-07-2016 15:49 yomimono wait, I meant !, not ?

27-07-2016 15:49 GemmaG Yes - please do :)

27-07-2016 15:50 yomimono next (I think? we got a bit lost) is that solo5 got merged!! yaaaaaay!

27-07-2016 15:50 yomimono sorry we didn't get to it before djwillia had to leave :(

27-07-2016 15:50 avsm WOOP!

27-07-2016 15:50 avsm dan had to go and martin is on a well deserved vacation

27-07-2016 15:50 noddy BEST THING EVER

27-07-2016 15:50 seangrove wow

27-07-2016 15:50 avsm but i am currently trying to compile up a static mirage-www without tls to go live

27-07-2016 15:51 hannes tls works..

27-07-2016 15:51 avsm got the KVM machine up now, so it'll be on solo5.mirage.io soon

27-07-2016 15:51 yomimono which is the next agenda item! yay!

27-07-2016 15:51 avsm is entropy all working now hannes?

27-07-2016 15:51 noddy NO.

27-07-2016 15:51 hannes noddy: hmm?

27-07-2016 15:51 noddy we discussed that 10 minutes ago, no?

27-07-2016 15:51 hannes well, so I do have entropy using some branches , see mirage-dev repo... noddy has fixes for build systems which avoid some copying around

27-07-2016 15:52 avsm anyway, static website incoming first without TLS as I'm keen to have a simpler version anyway

27-07-2016 15:53 dbuenzli Who needs tls anyways the certificate system is broken.

27-07-2016 15:53 yomimono dbuenzli: ...is this leading to a concrete proposal for something interesting?

27-07-2016 15:54 dbuenzli Yes drop support for it. Things get simpler.

27-07-2016 15:54 noddy steganography all the way

27-07-2016 15:55 avsm alright, off topic... :-)

27-07-2016 15:55 yomimono looks like that's where avsm is going, so yay. if that resolves everyone's questions about the current state of entropy, let's move on

27-07-2016 15:55 yomimono I stuck a hackathon recap item on the agenda but I don't think it's super necessary unless anyone else has something they're desperate to discuss

27-07-2016 15:56 yomimono going once, going twice...

27-07-2016 15:56 noddy the food was gone too early?

27-07-2016 15:56 yomimono noted.

27-07-2016 15:56 noddy thank you.

27-07-2016 15:56 avsm ENOTENOUGHWINE

27-07-2016 15:56 noddy no, there was plenty of that. :D

27-07-2016 15:56 yomimono unpurecamelbot, got that?

27-07-2016 15:57 yomimono avsm, tell us about TWiOPAM.

27-07-2016 15:57 avsm Done already, will figure out best way to hook it in

27-07-2016 15:57 avsm (if you scroll up to the website discussion, future log readers)

27-07-2016 15:58 yomimono great. next item is mirage.io updates and improvements, which I believe we've also already discussed?

27-07-2016 15:58 yomimono in that case, any other business?

27-07-2016 15:59 yomimono hearing none, I declare this meeting a resounding success..

27-07-2016 15:59 yomimono everyone did a great job except my . key.

27-07-2016 15:59 avsm woop woop. thanks everyone!

27-07-2016 15:59 noddy \O/ \O_

27-07-2016 16:00 dbuenzli bye

27-07-2016 16:00 noddy .o| \O/ |O/