09:53 <miragebot> mirage/master 905ca53 David Scott: Merge pull request #835 from djs55/solo5...

09:53 <miragebot> mirage/master 5aa86d8 David Scott: Prepare to release 3.0.3...

09:53 <miragebot> mirage/master 3e3b86c David Scott: Link io-page-unix on Ukvm and Virtio too...

09:53 <miragebot> [mirage] djs55 pushed 3 new commits to master: https://git.io/vHbre

10:04 <miragebot> [mirage] djs55 tagged v3.0.3 at master: https://git.io/vHbrA

12:39 <apache2> wow david is on fire

12:39 <djs55> for one day only ;)

13:24 <hannes> djs55: this io-page-unix dependency for solo5 in mirage.ml is no good (please see my comment on your (already merged :/) PR)

13:42 <djs55> yes, I agree with your logic. Sorry about that — I'm in the process of reverting (just double-checking that everything still links across all the backends)

13:52 <hannes> it would be even more appreciated if you not only link, but also run the various backends on such invasive changes

14:42 <mato> found cafe with wifi in zurich (took a while :)) back for a bit

14:43 <mato> djs55: i've not been following the reasons for the io-page changes, but my WIP cleanups of the mirage-solo5 bindings (the branches mention in solo5/solo5#200)

14:44 <mato> djs55: ...include changes to mirage-net-solo5 to remove the (afaict) last remaining direct use of Io_page on the write path

14:44 <mato> djs55: (just fyi)

14:57 <hannes> mato: thanks for your cleanups, i do like them... (and would appreciate to get rid of all the assertions -- either return the error or fail hard in these cases)

16:02 <djs55> mato: thanks for the links. The io-page changes were mostly small things: I merged the outstanding "get_addr" PR, then changed the prefix from caml_ to mirage_ opportunistically when yallop pointed out that caml_ was risky. I changed the build system to jbuilder + a discover.ml that can query `pkg-config` via the `Configurator` module — I think that's made the build a bit simpler, although it's still more complicated than I'd like. The

16:02 <djs55> biggest change really is to the opam metadata — rather than having a depopt on "mirage-xen-ocaml" which creates a findlib package "io-page.xen" I made a top-level opam "io-page-xen" package with an explicit dependency. I was recently inspired by http://rgrinberg.com/posts/optional-dependencies-considered-harmful/

16:45 <djs55> I installed an alpine-3.6.1 to test ukvm, but when I launch it (./ukvm-bin ./speaking_clock.ukvm) it gets an EACCES from the `mprotect` immediately after reading the `speaking_clock.ukvm` file. Am I doing something obviously wrong? I'm running everything as root.

16:47 <djs55> I'll file an issue :)