Irc discussions from #mirage on 20-04-2016

Written by unpurecamelbot
Classified under: irclog
Published: 2016-04-20 (last updated: 2016-04-20)

20-04-2016 14:58 engil ohai o/

20-04-2016 14:58 lobo hi

20-04-2016 14:58 GemmaG Hi :)

20-04-2016 14:58 avsm The git logging bot is alive! Let's hope there isn't a merge conflict!

20-04-2016 14:59 engil :)

20-04-2016 14:59 talex5 Hi

20-04-2016 15:00 djs55 waves

20-04-2016 15:01 nullcat hi

20-04-2016 15:01 avsm Alright! Let's begin the IRC meeting. Agenda is at

20-04-2016 15:01 avsm Who is around? We've had a few waves

20-04-2016 15:01 thomasga \o/

20-04-2016 15:01 djwillia hi all

20-04-2016 15:01 dbuenzli \o/

20-04-2016 15:01 amirmc hi

20-04-2016 15:02 avsm First port of call is information retrieval. There were a few threads about this on mirageos-devel

20-04-2016 15:02 avsm GemmaG Hannes, do you want to say anything about this?

20-04-2016 15:03 dbuenzli Note if you want to have good documentation, you have to design it at some point. These things do not happen magically.

20-04-2016 15:03 avsm While they're typing, I'm going to jump ahead to "SSL expiration" as I'm refreshing the site: I don't think it's controversial to switch to letsencrypt certificates, so I will go ahead and do that as part of the website refresh

20-04-2016 15:03 mattg waves

20-04-2016 15:04 avsm Sorry, I realised Hannes is driving to Oxford at the moment. Information retrieval covers a few things: firstly where our individual libraries and contents are, secondly the API documentation, thirdly tool documentation

20-04-2016 15:04 avsm the mirage-dashboard project and the TROVE are taking shape, and so we'll start curating a list of projects and libraries soon on the main website.

20-04-2016 15:05 GemmaG Is the plan also to separate general mirage content from hackathon content on canopy?

20-04-2016 15:05 avsm API documentation is (as dbuenzli notes) a mess at the moment, but is due to get better due to the release of a cross-referenced documentation tool. As part of deploying it, we'll need to work through explaining some of the core concepts in the Mirage API better.

20-04-2016 15:06 dbuenzli I think people should avoid using the canopy but rather try to improve the mirage website.

20-04-2016 15:06 avsm I'm inclined to split this into two sections: the MirageOS website is where curated documentation sits, and the Canopy is the stream-of-consciousness for rapid pushes

20-04-2016 15:06 avsm It's a lot easier to edit stuff into the main website if it shows up on Canopy first

20-04-2016 15:06 dbuenzli It's not in an entirely bad shape.

20-04-2016 15:06 dbuenzli (the mirageos website)

20-04-2016 15:07 dbuenzli Remove cruft and make it more user-centered

20-04-2016 15:07 avsm Yeah, but it hasn't kept up with all the recent changes. Finding (e.g.) functoria information is an exercise in search.

20-04-2016 15:07 avsm I'm curious what the newer contributors such as djwillia or mattg think about this

20-04-2016 15:07 avsm mattg: especially with your effort to fix CLOCK

20-04-2016 15:07 thomasga well we haven't updated the wiki/doc in for a long time

20-04-2016 15:07 amirmc I think splitting things across two sites makes it harder for people to find.

20-04-2016 15:08 avsm It's not splitting across two sites: the Canopy is where notes get pushed, and then they get edited into the main website. New users go to

20-04-2016 15:08 amirmc e.g. I'm not sure we have any reference or pointers to canopy fro the current site

20-04-2016 15:08 thomasga I'm fine to integrate canopy features in (and new style maybe)

20-04-2016 15:09 thomasga can't we just replace the blog post stream by canopy?

20-04-2016 15:09 avsm that's right -- and there shouldn't be. But Canopy can reference A good example is talex5's error handling proposal, which was a PR to mirage-www (still is), but could have been edited live on Canopy instead

20-04-2016 15:09 avsm thomasga: yes, we can do that.

20-04-2016 15:09 GemmaG Perhaps some necessary changes to + a canopy style stream

20-04-2016 15:09 mattg avsm: I think when I first came across Mirage, I followed the hello world example from, and then pretty much went to working from examples in mirage-skeleton and using .mli files for reference

20-04-2016 15:09 GemmaG Yes as per above :)

20-04-2016 15:09 avsm the big flaw right now is that we don't edit the blog post content back into the docs, so users have to search blog posts for documentation (e.g. functoria)

20-04-2016 15:09 thomasga yes, we do terrible job with the wiki/doc

20-04-2016 15:09 avsm mattg: thanks, that's useful -- so a rough learning curve :)

20-04-2016 15:09 djwillia avsm: I still haven't wrapped my head around all of the packaging stuff (that I think is really important to understanding Mirage) but I'm not sure that's the website's fault

20-04-2016 15:10 dbuenzli I'm sure it is

20-04-2016 15:10 avsm djwillia: it's an integral part of a library OS, so I think an explanation of OPAM (or least a strong xref to the OPAM site) is appropriate on

20-04-2016 15:11 djwillia avsm: yes, certainly, I think needing to understand packaging is the biggest block to understanding Mirage

20-04-2016 15:11 mattg avsm : I should also mention I learnt a lot from Mindy's and talex5's blogs

20-04-2016 15:11 thomasga me too

20-04-2016 15:11 avsm Ok, so GemmaG and I will take this on as an action then with help from Eng and mort: work through the website/blog/canopy melange and figure out the workflow we want (blog posts should come out more quickly, and get edited into the main website)

20-04-2016 15:11 dbuenzli which is bad (not their prose of course)

20-04-2016 15:11 avsm Mindy and talex5's blogs are legendary :)

20-04-2016 15:12 mort___ (yo)

20-04-2016 15:12 GemmaG avsm: sounds good :)

20-04-2016 15:12 dbuenzli this knowledge should be consolidated in the mirageos website

20-04-2016 15:12 avsm Yes, agree with dbuenzli -- they did great posts due to the lack of documentation but it should be consolidated by now in the central docs.

20-04-2016 15:12 avsm Ok, closing out Information Retrieval. Any last comments?

20-04-2016 15:12 djwillia avsm: I mostly learned (and am still learning) from doing terrible terrible hacks before understanding packaging :)

20-04-2016 15:12 mato Random comment, I think the process of creating the Solo5 port (and integrating it with Mirage) could be good material for writing a blog post.

20-04-2016 15:13 mato "How to port Mirage to a new platform"

20-04-2016 15:13 thomasga would be great if Engil can help to fix/improve CSS

20-04-2016 15:13 avsm mata djwillia : yes definitely! It also covers the ukvm work which is an area we've not talked about much in the blog (aside from the MiniOS ARM port that talex5 did)

20-04-2016 15:13 djwillia mato: yes, that's a great idea because it's forcing us to really understand what a mirage "platform" is

20-04-2016 15:13 engil thomasga: well why not

20-04-2016 15:13 engil :)

20-04-2016 15:13 mattg avsm: I can offer as some intro material, once i've merged mort__'s fixes

20-04-2016 15:13 thomasga cool :p

20-04-2016 15:13 avsm mato: could you create an issue on as a reminder to do this blog post...

20-04-2016 15:14 avsm Alright, full agenda, so moving onto next topic: Quality and Test

20-04-2016 15:14 mato avsm: will do.

20-04-2016 15:14 thomasga we need more tests :p

20-04-2016 15:14 avsm Good news: OCaml 4.03 is due out this week with an incredible new inliner that speeds up some mirageos kernels by 20-30% with a simple recompiler

20-04-2016 15:14 djwillia avsm mato: we may want 2 blog posts one on ukvm, on on porting mirage to new platforms

20-04-2016 15:14 avsm bad news: yallop keeps finding showstopper bugs in 4.03 ;-)

20-04-2016 15:14 avsm djwillia: (agree on more small blog posts)

20-04-2016 15:14 mort___ (that's good news really...)

20-04-2016 15:15 avsm however, 4.03 is converging, and we have done really good work collectively on moving to PPX from camlp4

20-04-2016 15:15 avsm so aside from a port of the runtime to mirage-platform, we're pretty much set to move to 4.03

20-04-2016 15:15 avsm Has anyone taken a look at a 4.03 runtime for mirage-platform/xen? If not I'll take a look soon

20-04-2016 15:15 thomasga NO MORE CAMLP4!

20-04-2016 15:15 thomasga I've fulfilled my life goal

20-04-2016 15:15 yomimono cheers

20-04-2016 15:15 dbuenzli NO MORE PPX

20-04-2016 15:16 avsm Bulk builds via Docker are in good shape: we have a comprehensive set of containers hosted on the Docker Hub for every conceivable combination of Linux distros at -- read it and weep :)

20-04-2016 15:16 thomasga dbuenzli: that's my next life goal

20-04-2016 15:16 dbuenzli you have a new life goal

20-04-2016 15:16 thomasga exactly :-)

20-04-2016 15:16 avsm Leave camlp4 alone! It has served us well. A moment of silence for our old friend

20-04-2016 15:16 thomasga HAAAAA

20-04-2016 15:16 mort___ belch

20-04-2016 15:16 djs55 is it really gone? I still feel its presence… somewhere

20-04-2016 15:17 thomasga ppx his little brother is now there

20-04-2016 15:17 avsm So overall, the last few months of work should pay off very soon -- 4.03 is around the corner, ppx extensions make life easier, and we have an increasing amount of automated build infrastructure.

20-04-2016 15:17 avsm But we need to finish the leap as there are still a few stragglers that don't work.

20-04-2016 15:17 avsm For example: OASIS

20-04-2016 15:17 avsm (with ocaml 4.03)

20-04-2016 15:17 avsm is a giant elephant in the room

20-04-2016 15:17 mort___ what's the elephant?

20-04-2016 15:17 thomasga can we just remove oasis instead of porting it to 4.03?

20-04-2016 15:18 dbuenzli Let's be realistic what's the actual problem ?

20-04-2016 15:18 djs55 in the defense of oasis, it does make windows porting easier

20-04-2016 15:18 avsm one of the dependencies fails; I'm primarily interested in knowing if someone wants to take fixing it in opam-repository as an action

20-04-2016 15:18 avsm I'm sure it wont be hard

20-04-2016 15:18 dbuenzli Why not ping the author ?

20-04-2016 15:18 avsm no response

20-04-2016 15:19 avsm Ok, I'll take on the action then, and either fix it or bug someone else to fix it :P

20-04-2016 15:19 avsm Closing out quality and test unless there are any other updates?

20-04-2016 15:19 dbuenzli A message to caml-list could do no ?

20-04-2016 15:19 dbuenzli It's a pretty important part of the current eco-system

20-04-2016 15:20 avsm Sure. I'd rather get it unblocked in the short term though. A patch would be just as good as an email

20-04-2016 15:20 avsm And yes, it's very widely used indeed in our libraries.

20-04-2016 15:20 avsm Ok, let's take the OASIS topic offline for now.

20-04-2016 15:20 avsm Moving onto "Improved Logging and Errors"

20-04-2016 15:20 companion_cube in the defense of oasis, it's awfully convenient

20-04-2016 15:20 avsm talex5 has been working on this...

20-04-2016 15:21 talex5 I split out a logging-only patch to Functoria, which is now merged. The corresponding PR to Mirage isn't merged yet:

20-04-2016 15:21 talex5 I split out a logging-only patch to Functoria, which is now merged. The corresponding PR to Mirage isn't merged yet:

20-04-2016 15:21 talex5 I split out a logging-only patch to Functoria, which is now merged. The corresponding PR to Mirage isn't merged yet:

20-04-2016 15:21 talex5 I split out a logging-only patch to Functoria, which is now merged. The corresponding PR to Mirage isn't merged yet:

20-04-2016 15:21 talex5 I split out a logging-only patch to Functor, which is now merged.

20-04-2016 15:21 talex5 It's slightly ugly; if someone can suggest a more elegant solution please do :-)

20-04-2016 15:21 avsm is someone cutting and pasting into this meeting??

20-04-2016 15:21 talex5 :-(

20-04-2016 15:21 dbuenzli Stuttering client.

20-04-2016 15:21 talex5 Sorry. Using OS X cut-and-paste.

20-04-2016 15:22 talex5 I split out a logging-only patch to Functoria, which is now merged. The corresponding PR to Mirage isn't merged yet:

20-04-2016 15:22 avsm Any blockers to the mirage/mirage merge?

20-04-2016 15:22 avsm Someone break talex5 out of his temporal loop

20-04-2016 15:22 yallop mirage/mirage merge is quite difficult to say fast

20-04-2016 15:22 talex5 Hoping @Drup can suggest an alternative.

20-04-2016 15:23 talex5 Once that's merged, I'll make PRs to add logging to the various libraries.

20-04-2016 15:23 avsm Ok, getting a release out with this next week would be useful to unblock people

20-04-2016 15:23 thomasga +1

20-04-2016 15:23 talex5 Then I'll resubmit the original patches to improve the error reporting and we can discuss that next time.

20-04-2016 15:23 avsm Yeah, fixing up libraries is going to take a while so I'm keen to get the mirage/mirage tool sorted

20-04-2016 15:24 avsm What would you like to do with your mirage-www PR

20-04-2016 15:24 avsm the long standing discussion document about error reporting mechanisms

20-04-2016 15:24 talex5 Needs updating.

20-04-2016 15:25 GemmaG Got some great trip reports on the hackathon: Just waiting on a couple more details for last entry. Can add to canopy.mirae

20-04-2016 15:25 GemmaG bah wrong button

20-04-2016 15:25 talex5 I don't know if there's full agreement yet, but I'm happy to update it to say that result types should be used for expected runtime errors and exceptions for bugs.

20-04-2016 15:25 avsm we appear to have cut and paste issues this week

20-04-2016 15:25 GemmaG hahahahahaha

20-04-2016 15:25 yomimono I'm very pro-result types & rresult so I'm on board with that

20-04-2016 15:26 dbuenzli \o/

20-04-2016 15:26 avsm thanks talex5 -- establishing logging standards across libraries is right at the bottom of the stack, so it feels like a good one for the website refresh as well

20-04-2016 15:26 dbuenzli @talex which exceptions ?

20-04-2016 15:26 dbuenzli Nevermind we can discuss that there

20-04-2016 15:26 avsm yes yomimono -- and 4.03 has the Result.t defined in the standard library! This is more amazing than breaking away from camlp4

20-04-2016 15:27 avsm Ok, closing out logging -- rest can be discussed on the respective issues, and its making good progress.

20-04-2016 15:27 yomimono thanks talex5 for continuing to push on this :)

20-04-2016 15:27 avsm Next onto a slightly random but annoying bug: the lack of strtod

20-04-2016 15:27 avsm a number of libraries are hitting this due to use of floating point, and in particular Cohttp

20-04-2016 15:27 dbuenzli We already discussed this last time.

20-04-2016 15:27 yomimono Did anyone do anything as a result of that discussion?

20-04-2016 15:27 dbuenzli Just use the strod.c file that is boot-rpi-ocaml

20-04-2016 15:27 thomasga I remember that Daniel said he wanted to fix this

20-04-2016 15:27 yomimono didn't

20-04-2016 15:28 avsm Yeah, no progress since 2 weeks ago I believe, except has anyone looked at how important locale dependence is?

20-04-2016 15:28 dbuenzli I didn't say I wanted to fix it

20-04-2016 15:28 dbuenzli I just told people what to use.

20-04-2016 15:28 thomasga tututut I have a perfect memory :p

20-04-2016 15:28 avsm It could be significantly easier if we grab a BSD library for this (e.g. OpenBSD's)

20-04-2016 15:28 dbuenzli The file there is what used to be used in JavaScript runtime until they switched to grisu

20-04-2016 15:28 thomasga (was just joking, you indeed pointed out to the right files)

20-04-2016 15:28 avsm dbuenzli: ah, that's most useful to know, thanks.

20-04-2016 15:29 avsm Ok, let's move discussion of this one to the issue as well. It would be good to bump this up to get fixed in the mirage-platform support for 4.03, so I'll take a look

20-04-2016 15:29 avsm it's at least easy to test by running a cohttp server

20-04-2016 15:29 avsm or yojson, or most things...

20-04-2016 15:29 yomimono or calling float_of_string

20-04-2016 15:30 avsm I think we've made progress -- used to crash if we touched the floating point registers in xen mode :)

20-04-2016 15:30 avsm Ok, onto the next topic...

20-04-2016 15:30 avsm Mirage 2.8.0 released along with new version of TCP/IP, and most important, this was yomimono's first Mirage release!!!

20-04-2016 15:30 avsm Three cheers for yomimono for doing a full release, and implicitly, by not messing it up!

20-04-2016 15:30 GemmaG woop :)

20-04-2016 15:30 thomasga \p\ /o/

20-04-2016 15:30 mort___ cheers

20-04-2016 15:30 engil \o

20-04-2016 15:30 avsm What's happening there yomimono ?

20-04-2016 15:30 djwillia yay

20-04-2016 15:30 yomimono I accept plaudits in the form of kitty cuddles

20-04-2016 15:31 yomimono 2.8.0 was pretty minor but involved a change to mirage-types -- we now have an icmp module type and corresponding separate module implementation in tcpip.

20-04-2016 15:31 yomimono you can now easily make unikernels with whatever icmp behavior you want, which will make it quite a bit easier to make routers &c with our tcpip stack.

20-04-2016 15:31 avsm Does this make NAT easier?

20-04-2016 15:31 dbuenzli superbe

20-04-2016 15:31 yomimono potentially, yes.

20-04-2016 15:32 talex5 Yeah, would be really useful to do NAT of ICMP messages (e.g. ping) for mirage-firewall.

20-04-2016 15:32 yomimono nods

20-04-2016 15:32 avsm awesome. Hannes informs he is working on the TCP/iP test suite against freebsd as well, so that (combined with the AFL fuzzing work that yomimono did at the hackathon) puts us in good shape for more effective tcp/ip testing too

20-04-2016 15:32 dbuenzli avsm is the tcp/ip problem under embargo ? I saw that you retracted the topic from the call agenda

20-04-2016 15:32 thomasga great!

20-04-2016 15:32 avsm dbuenzli: yes

20-04-2016 15:33 avsm One outstanding PR in tcpip is ipv6

20-04-2016 15:33 avsm we have the code merged now in the stack, but not exposed to configuration

20-04-2016 15:33 dbuenzli (I think this should be made more clear in the future)

20-04-2016 15:33 avsm Does anyone have a particular interest in ipv6 and would like to learn more / hack on it?

20-04-2016 15:33 yomimono I'm going to have to a bit as part of ongoing tcpip restructuring

20-04-2016 15:34 yomimono it's part of pulling all the parsers/printers out and exposing them

20-04-2016 15:34 avsm yomimono: ack. I have access to public IPv6 addresses on the Rackspace Cloud, so point me if you need a VM that is routable for you to test stuff (or deploy)

20-04-2016 15:34 yomimono although if anyone else wants to volunteer I encourage them :)

20-04-2016 15:34 avsm we can't deploy Xen unikernels on Rackspace yet, but some tuntap thing should work.

20-04-2016 15:34 yomimono it actually should be doable with ec2 as well

20-04-2016 15:34 avsm IPv6 on EC2?

20-04-2016 15:35 yomimono think so

20-04-2016 15:35 avsm I thought they didn't support it except via the load balancer service

20-04-2016 15:35 yomimono also I have my own xen machine, so nbd

20-04-2016 15:35 yomimono maybe I'm wrong about ec2; irrelevant to most people anyway, sorry

20-04-2016 15:35 avsm Ok, please ping yomimono or me if you are interested in IPv6!

20-04-2016 15:35 avsm Next, onto the hackathon: a roundup blog post is flying around. What's happening GemmaG?

20-04-2016 15:36 GemmaG Yes! Got some great trip reports on the hackathon: Just waiting on a couple more details for last entry. Can add to canopy and

20-04-2016 15:36 thomasga (I've read that before)

20-04-2016 15:36 GemmaG hahahaha

20-04-2016 15:36 GemmaG I'm proud of that paragraph :P

20-04-2016 15:36 avsm deja vu -- I didn't click on the link before, and it's great to see pictures of sunshine and hacking there too :)

20-04-2016 15:37 avsm looks great, that'll be our next blog post on the main site hten

20-04-2016 15:37 avsm If anyone else does have updates on their trip that aren't there, please do ping soon

20-04-2016 15:37 avsm Ok, moving onto libraries then, just to get a check on them

20-04-2016 15:37 avsm mirage-dashboard came first: is the author here?

20-04-2016 15:38 avsm there was some mailing list traffic, and it's making progress. I'm going to do a unix-socket deployment (as I can do that quickly) next week hopefully

20-04-2016 15:38 avsm Having something up describing our libraries is very useful, and the TROVE was a little too low-level and always out of sync

20-04-2016 15:38 avsm Moving onto Canopy: engil, what's happening with it?

20-04-2016 15:39 engil slowly progressing

20-04-2016 15:39 engil RSS, tags are here

20-04-2016 15:39 engil we still need to handle client caching properly

20-04-2016 15:39 avsm I've got a local deployment -- is anyone else using it themselves?

20-04-2016 15:39 engil dinosaure, hannes

20-04-2016 15:40 avsm Ah yes! Hannes has been blogging again --

20-04-2016 15:40 engil hannes has made very good progress on making it run on Xen

20-04-2016 15:40 avsm Great stuff -- mort___ I'm going to move over to it as soon as I sort out my build scripts; maybe a target for too since you also run a unikernel homepage

20-04-2016 15:41 avsm Next: ARM images for Alpine Linux to replace the builds

20-04-2016 15:41 mort___ yes

20-04-2016 15:41 avsm Progress has been made here, but is waiting on a decent build box.

20-04-2016 15:41 mort___ been slowly workign on dockerising all the random crap i have for … canopy next

20-04-2016 15:41 avsm Good news: we have a fast multicore arm64 machine in the computer Lab that can do the builds.

20-04-2016 15:41 avsm mort___: what is the bad news? :)

20-04-2016 15:42 mort___ re arm build box — have (I think) got the new firmware in place on the box and on its BMC

20-04-2016 15:42 mort___ in doing so, the rootfs got moulinexed

20-04-2016 15:42 mort___ (thanks avantek "support"!)

20-04-2016 15:42 avsm (we have reason to believe that we are in fact the first people in the world to ever buy an ARM64 machine from this particular supplier)

20-04-2016 15:42 mort___ i've now booted off usb successfully and mounted the old filesystems — the uefi fs and the boot partition

20-04-2016 15:43 mort___ i've not got further yet in either fixing things up so it boots off its hdd again, or doing a complete reinstall

20-04-2016 15:43 avsm thank you for working through all the tedious installation issues mort___! The good news is that one it is up, we have fast builds for both ARM32 and ARM64.

20-04-2016 15:43 mort___ i will hopefully find time tomorrow morning to spend on that

20-04-2016 15:43 avsm Ok, moving onto... camlp4: any remaining libraries?

20-04-2016 15:44 avsm aside from the oasis issue, is anyone aware of other blockers for libraries they need that hold them back to <OCaml 4.03?

20-04-2016 15:44 thomasga last time I tried, Irmin was not compiling with flambda

20-04-2016 15:44 avsm We have fixed URI recently, but Cohttp actually still has a constraint issue for OCaml 4.03 I need to fix (somewhere in the PPX build chain). I'm compiling right now.

20-04-2016 15:44 thomasga hopefully it should be better now

20-04-2016 15:44 avsm thomasga: woah, when was this? Many bugs have been fixed in the last 2 months by mshinwell, yallop, diml, chambart

20-04-2016 15:45 thomasga last compiler hacking meeting

20-04-2016 15:45 thomasga the compilation just didn't terminate

20-04-2016 15:45 avsm if anyone wants to try out ocaml 4.03, now is an excellent time to do so: opam switch 4.03.0+trunk or docker run -it ocaml/opam:debian-7_ocaml-4.03.0_trunk_flambda will get you a working environment

20-04-2016 15:45 djs55 camlp4-free cstruct hasn't been released yet. It looks like ocaml-pcap hasn't been merged and released. Perhaps it's time to merge and release those 2?

20-04-2016 15:45 thomasga would be useful to test it with 4.03's candidate if anyone has time

20-04-2016 15:46 avsm djs55: thanks, added to action items. ocaml-pcap should be straightforward to release.

20-04-2016 15:46 mort___ i know i have an issue against ocaml-pcap that i need to fix

20-04-2016 15:46 talex5 Are there any bulk builds for 4.03?

20-04-2016 15:46 avsm thomasga: I've kicked off a docker build of irmin head on my laptop just now too

20-04-2016 15:46 mort___ (via djs55 / yomimono i believe)

20-04-2016 15:46 avsm mort___: do you want to take care of the ocaml-pcap release then?

20-04-2016 15:46 mort___ sure

20-04-2016 15:46 yomimono ocaml-pcap 0.4.0 is a fairly large change from 0.3 FWIW, users will need to make code changes for sure

20-04-2016 15:46 mort___ is there a document describing release process anywhere? (as a n00b :)

20-04-2016 15:46 avsm mort___: thanks! talex5: I've done partial bulk builds, but they took > 500GB and so I ran out of space

20-04-2016 15:47 mort___ else i'll bug yomimono as the most recent releaser :)

20-04-2016 15:47 avsm mort___: yomimono: just bump up to ocaml-pcap 1.0 then. semantic versions are cheap :)

20-04-2016 15:47 avsm ok, if everyone who mentioned libraries could do a refresh that'd be useful. I'll be tracking down the Cohttp one which unblocks a number of libraries. 4.03 will be out real soon now

20-04-2016 15:48 avsm Onto mirage-rump, which I believe mato has been looking at.

20-04-2016 15:48 avsm pinging mato pinging mato

20-04-2016 15:49 avsm also, a semi hijack to congratulate djwillia on releasing ukvm as well, as mentioned in last call

20-04-2016 15:49 yomimono yaaaaaay! \o/

20-04-2016 15:49 djwillia mato has been fantastic with helping get solo5 as a proper backend

20-04-2016 15:49 avsm mato and djwillia have been fixing up issues (better CPU portability, better OPAM packaging) which is awesome!

20-04-2016 15:49 thomasga yay!

20-04-2016 15:49 djwillia yes, a big thank you to mato

20-04-2016 15:49 avsm full support from me for getting solo5 in-tree, the qemu support is something we've wanted for years!

20-04-2016 15:49 thomasga +100

20-04-2016 15:50 thomasga that's pretty cool work

20-04-2016 15:50 avsm mato has fallen asleep at his keyboard from his heroic coding efforts, so we'll move onto the next library

20-04-2016 15:50 djwillia thanks!

20-04-2016 15:50 avsm which is...

20-04-2016 15:50 seangrove Thanks you djwillia!

20-04-2016 15:50 avsm what for it...

20-04-2016 15:50 avsm TOPKG!

20-04-2016 15:50 avsm how's it going dbuenzli ?

20-04-2016 15:50 thomasga drum roll

20-04-2016 15:50 dbuenzli It's going.

20-04-2016 15:50 dbuenzli When it's done.

20-04-2016 15:50 avsm Background: this is an effort to provide a simple package description format that is "opam native" to avoid duplication of metadata

20-04-2016 15:51 dbuenzli But I'm doing only this.

20-04-2016 15:51 dbuenzli But I'm slow because of you.

20-04-2016 15:51 thomasga of me?

20-04-2016 15:51 djs55 Would more coffee help?

20-04-2016 15:51 dbuenzli no avsm

20-04-2016 15:51 avsm alright, no more Cambridge formal halls for dbuenzli until topkg is released

20-04-2016 15:51 thomasga avsm: stop slowing down Daniel!

20-04-2016 15:51 dbuenzli He got me to join ocaml lab's slack

20-04-2016 15:52 avsm but seriously, I'm ready to port Cohttp to topkg whenever you give the signal :)

20-04-2016 15:52 avsm Any other libraries of note that anyone would like to highlight that have been released recently?

20-04-2016 15:52 avsm dinosaure got really nice results from using flambda on Decompress/Zlib

20-04-2016 15:52 dbuenzli Yes seriously, I hope to be finished at least at the beginning of may. I'm tired of the build stuff and want to do a little bit of other things

20-04-2016 15:53 avsm dbuenzli: that sounds like an excellent timeframe to me!

20-04-2016 15:53 thomasga yay!

20-04-2016 15:53 dbuenzli I also have many other things in the pipeline that are waiting for it (e.g. carcass)

20-04-2016 15:53 dinosaure hi all :}

20-04-2016 15:53 thomasga I'm really looking forward to change all of my build scrips

20-04-2016 15:53 avsm I realised some people don't know this, but you can get the packages sorted by date at to find out what has been merged recently

20-04-2016 15:53 dinosaure I forget the meeting :d

20-04-2016 15:53 avsm dinosaure: hello! 5 minutes left :-)

20-04-2016 15:54 avsm any updates on your libraries that you'd like everyone to know?

20-04-2016 15:54 thomasga avsm: i prefer the 2nd "most popular" package in that list

20-04-2016 15:54 avsm while dinosaure is typing, thomasga : any update on new ocaml-git and irmin releases?

20-04-2016 15:54 dinosaure yes, I will improve the performance of and after that I will try to compute a gzip content

20-04-2016 15:55 thomasga there are few breaking changes

20-04-2016 15:55 thomasga and perf improvment

20-04-2016 15:55 avsm dinosaure: this is a really good test case for the new flambda mode in 4.03, so looking forward to that

20-04-2016 15:55 thomasga they now both use logs and rresult

20-04-2016 15:55 thomasga and fmt

20-04-2016 15:55 engil thomasga: you had the time to investigate the stacktraces I gave you ? :}

20-04-2016 15:55 thomasga engil: no

20-04-2016 15:55 engil :}

20-04-2016 15:56 avsm We're now also using Irmin in this very meeting to log and push them to Canopy, so I'm glad that Irmin's performance has been improved so it can scale as we type more

20-04-2016 15:56 engil ok :)

20-04-2016 15:56 thomasga might maybe release them anyway as the API is nicer (even if they are still some issues in git push/pull with the smart HTTP protocol)

20-04-2016 15:56 engil avsm: well to be honest the bot is just keeping everything in memory and will push later

20-04-2016 15:56 avsm That concludes the agenda that everyone wanted to discuss. Any other business people want to discuss in the closing minutes that weren't on there? Things you want tested or otherwise highlighted?

20-04-2016 15:56 thomasga @talex5 did a lot of work, I haven't done anything

20-04-2016 15:57 thomasga i want to see unikernels in space

20-04-2016 15:57 dbuenzli me in snow

20-04-2016 15:57 dbuenzli on my alitmeter

20-04-2016 15:57 avsm in that case thomasga you should come to this talk today!

20-04-2016 15:58 avsm alright, sounds like we're done.

20-04-2016 15:58 thomasga bye!

20-04-2016 15:58 yomimono "the xxi century" - nice

20-04-2016 15:58 avsm In two weeks OCaml 4.03 will be out; have fun trying it out everyone!

20-04-2016 15:58 dbuenzli bye

20-04-2016 15:58 GemmaG Enjoy the talk!