_whitelogger has joined #mirage
<hannes> avsm: well, at least I found some subtle bugs in the tls packet parser. which is good :)
<avsm> whitequark: cool, was looking at a simple backend for irminsule, and irc logging is one
<dsheets> frontend?
avsm has quit [Read error: Connection reset by peer]
<whitequark> ok, it works: http://irclog.whitequark.org/mirage
<dsheets> die, deniability, die
<dsheets> whitequark, mostly reviewed the latest diffset. it all looks great. will probably merge later today and defer release until ctypes 0.3 goes out
<whitequark> dsheets: \o/
<whitequark> dsheets: did you decide anything about string/bytes naming?
<whitequark> btw, as the ocamlfind bytes patch has landed, I'll probably add dependency on bytes to the patchset, too
<dsheets> well, i think your naming suggestion is a good idea but it does not match what the cooperative libs use :-/
<dsheets> lwt calls bigstrings bytes
<dsheets> and async calls bigstrings bigstrings
<dsheets> and they all call bytes strings
<dsheets> but I think the transition makes sense and i would support sodium setting a trend
<dsheets> to the shining utopia of immutable strings and mutable byte arrays on the horizon
<whitequark> (lwt/async) ouch, this is confusing.
<whitequark> ok, I'll rename it later today.
<dsheets> really, the new built-in type should be called octets... what if ocaml is used on a machine with 6-bit bytes?!?!
<dsheets> whitequark, i'd rather not have sodium 0.2 depend on ocaml >=4.02 but the type names can be aligned
<whitequark> dsheets: there will be no dependency, ocamlfind installs a shim.
<hannes> dsheets: does 'octet' help there? octet implies sth with 8 for me (yeah, I learned Latin in school)..
<whitequark> octet = 8 bit, byte = minimally addressable quantity.
<dsheets> hannes, it's just an ietf joke :-P
<whitequark> no software made within last ~30 years will work with non-octet bytes anyway
<whitequark> I tried to port LLVM to codegen for PDP-7 (12-bit bytes)
<whitequark> I gave up rather quickly
<whitequark> don't ask why :D
<whitequark> ok, no, let's say "no so-called portable software". there are still DSPs made with 16-bit bytes. but you have to customize practically any code for them.
<whitequark> </pedantic>
avsm has joined #mirage
djs55 has joined #mirage
thomasga has quit [Quit: Leaving.]
thomasga has joined #mirage
rektide_ is now known as rektide
avsm has quit [Quit: Leaving.]
avsm has joined #mirage
tlockney_away has quit [Ping timeout: 276 seconds]
<avsm> the ubuntu trusty filesystem is pretty hard to use now
<avsm> dns resolv.conf is different, bridging seems broken..
tlockney_away has joined #mirage
tlockney_away is now known as tlockney
<dsheets> ugh... is there a magic daemon we have to poke over dbus or sth?
<avsm> dsheets: not sure. the apt is also cripplingly slow on the arm atm, checking for dma in the mmc driver...
<avsm> might need to switch to sata by default for now
<dsheets> almost like cards should use filesystems that write to RAM and flush to flash later?
<dsheets> still need a solution for process regions
<avsm> we could call it irminage
<dsheets> d'irminage
<dsheets> so are we marking all camlp4 packages with 4.02 conditional camlp4 deps?
<avsm> just camlp4
<avsm> there's a <4.02 dummy package
<avsm> and a 4.02 trunk package
<avsm> so no versions required on other deps
<avsm> just a depends: ["camlp4"]
<cebka> dsheets: actually there are vfs buffers for that purposes unless you are using direct io
<avsm> cebka: the issue here is that there's no spare ram for a buffer cache on the cubie, so it's going straight to disk
<cebka> no ram, no performance
<dsheets> avsm, i thought it had 512M? what is eating so much mem?
<cebka> avsm: I was rather annoyed about direct IO requirement for aio in linux, so you basically need some FS layer above DIO+AIO block device
<avsm> a gig, but there's some limitations on the memory map so dom0 is just at 128mb
<avsm> cebka: yeah, it's mainly for block io, not filesystems (like on blktap in xen)
<cebka> avsm: and maybe a different IO elevator would help to order requests more wisely...
<avsm> cebka: i believe the best elevator on ssd is still the dumb one
<cebka> I'd still suggest noop with fifo_batch disabled
<cebka> oh, deadline not noop
<cebka> but usually the default elevator is cfq
tlockney is now known as tlockney_away
tlockney_away is now known as tlockney
tlockney is now known as tlockney_away
tlockney_away is now known as tlockney
thomasga has quit [Quit: Leaving.]
thomasga has joined #mirage
thomasga has quit [Quit: Leaving.]
<dsheets> ocamlfind 1.4.0 doesn't install camlp4 META for 4.02.0 and camlp4 doesn't install camlp4 META for 4.02.0...
thomasga has joined #mirage
rgrinberg has joined #mirage
thomasga has quit [Quit: Leaving.]
djs55 has quit [Quit: Leaving.]
avsm has quit [Quit: Leaving.]
avsm has joined #mirage
<avsm> dsheets: ah ok, i guess camlp4 should do it
thomasga has joined #mirage
<dsheets> avsm, that and the camomile camlp4 dep block utop under 4.02.0 i think
<avsm> AltGr: any luck with the opam trunk?
<AltGr> it's getting in better shape
<AltGr> base packages are back, the error that happened on lwt should now be properly reported
<avsm> awesome!
djs55 has joined #mirage
<avsm> dsheets: This FUSE extension allows character devices to be |
<avsm> | implemented in userspace. |
<avsm> | based on CUSE, answer Y or M.
<avsm> | |
<avsm> | If you want to develop or use a userspace character device |
yastero has quit [Ping timeout: 252 seconds]
<dsheets> avsm, hm? that from the kernel config? i don't have any CUSE stuff yet
<avsm> ya
<avsm> just never saw it before
yastero has joined #mirage
AltGr has left #mirage [#mirage]
<dsheets> the uid/gid agent handling system works now and is fairly nice except the exemption for exceptions which i've mailed leo about
<avsm> nice! separate process?
<dsheets> two, a handler and a pool of assets
<avsm> dsheets: sounds like a useful library on its own
<dsheets> avsm, potentially... it definitely has utility outside of just lofs for other FUSE applications... i just wish exceptions and extensible types could be marshaled and then it would be gorgeously type-safe
<avsm> i wonder how much memory is actually left over after running make inside vagrant inside virtualbox inside docker inside fusion
<dsheets> i don't understand why they don't just use fusion directly?
* dsheets starts dreaming about lxc handlers
djs55 has quit [Quit: Leaving.]
<avsm> cebka: yes,although need to get bike repaired in the morning so wont know when i'll be in the lab until that's done :(
dsheets has quit [Ping timeout: 240 seconds]
<cebka> avsm: got it, thanks
thomasga has quit [Quit: Leaving.]
rgrinberg has quit [Quit: Leaving.]
rgrinberg has joined #mirage
avsm has quit [Quit: Leaving.]
djs55 has joined #mirage
djs55 has quit [Quit: Leaving.]
djs55 has joined #mirage
circus has quit [Ping timeout: 276 seconds]
circus has joined #mirage
djs55 has quit [Quit: Leaving.]
avsm has joined #mirage
thomasga has joined #mirage
dsheets has joined #mirage
rgrinberg has quit [Quit: Leaving.]
dsheets has quit [Quit: Leaving]
rgrinberg has joined #mirage
avsm has quit [Quit: Leaving.]
NoNNaN has quit [Remote host closed the connection]
NoNNaN has joined #mirage
thomasga has quit [Quit: Leaving.]