<copumpkin>
gchristensen , niksnut: we need full laziness cached in an external DB so that ofborg's re-evals never evaluate the same subexpression twice, ever :)
<gchristensen>
"The ultimate research goal of Enchilada is: 'Information cannot be destroyed'. This is as far reaching in that even (unwritten) Terms are immutable in Enchilada."
<makefu>
how ironic: The requested URL /enchilada.jnlp was not found on this server.
<gchristensen>
needs more blockchain (?)
<makefu>
i think etherium will do the trick
<makefu>
"how about javascript ... but for money!"
jtojnar has quit [Ping timeout: 248 seconds]
<taktoa>
copumpkin: I've actually thought about this before; full laziness would be too slow but a `builtins.cache :: ∀α. (Serializable α) ⇒ α → α` would probably be pretty useful
<taktoa>
err, maybe it would be `∀α β. (Serializable α, Serializable β) ⇒ (α → β) → (α → β)`
<taktoa>
I guess the ideal place to put that `cache` combinator would be in `callPackage`
<taktoa>
though I guess there are closures inside a derivation attribute set (as opposed to the serialized `.drv` file) that would need to be recovered somehow
michaelpj_ has joined #nixos-dev
jtojnar has joined #nixos-dev
michaelpj_ has quit [Read error: Connection reset by peer]
<gchristensen>
omg you're not a trusted user, you can't
<Sonarpulse>
so far, I haven't needed to be :)
<Sonarpulse>
always a good sign!
<gchristensen>
hmm might as well make eelco one too
<Sonarpulse>
haha yeah
<gchristensen>
ok you're a trusted user now
<dtz>
\o/
pie__ has joined #nixos-dev
<gchristensen>
Sonarpulse: ok that failure is all yours :P
<Sonarpulse>
gchristensen: hahah ok
<Sonarpulse>
thanks!
pie_ has quit [Ping timeout: 240 seconds]
FRidh2 has quit [Quit: Konversation terminated!]
<niksnut>
the closure of the minimal NixOS install created by nixos-generate-config has grown from 798864648 bytes in 17.09 to 1011163512 bytes in master
jtojnar has quit [Ping timeout: 240 seconds]
<gchristensen>
O.o
<niksnut>
this appears to be mostly because thin-provisioning-tools now depends on boost.dev, which also pulls in icu4c
<Dezgeg>
anything using the new boost pulls in boost.dev I think
<niksnut>
eh
<Dezgeg>
due to some __FILE__ references in boost headers
<niksnut>
is that hard to fix?
<niksnut>
otherwise we have to revert boost
<Dezgeg>
I don't know, just that it's fairly recent
<LnL>
feel like we should use disallowedRequisites more somehow
<niksnut>
I've thought about having the NixOS tests grep the closure for -dev paths and give a fatal error if they find one