You mean if the caller wanted nothing to happen until it starts polling the future?
Zoxc
vadimcn: Yeah
vadimcn
Well, then you either allocate, or your Future is gonna have a lifetime...
even if coroutine started out running you'd have the same problem, wouldn't you?
Zoxc
You'd have more problems then. Coroutines which start out running only help out futures which don't start suspended
WindowsBunny joined the channel
WindowsBunny has quit
jseyfried
nrc: ping -- I'd like to discuss some ideas re goals / design space for macros 2.0 hygiene
(at some point)
ubsandroid joined the channel
brson has quit
montanonic joined the channel
niconii joined the channel
petso has quit
petso joined the channel
montanonic has quit
ubsandroid_ joined the channel
kimundi has quit
ubsandroid has quit
kimundi joined the channel
WindowsBunny joined the channel
ubsandroid_ has quit
ubsandroid joined the channel
Zoxc
Totally should use `not move` as a keyword for non-movable generators
woboats joined the channel
jseyfried has quit
jseyfried joined the channel
nagisa has quit
nagisa joined the channel
petso has quit
petso joined the channel
petso has quit
woboats has quit
nagisa has quit
petso joined the channel
petso has quit
petso joined the channel
woboats joined the channel
srwalker101 joined the channel
srwalker101 has quit
srwalker101 joined the channel
srwalker101 has quit
srwalker101 joined the channel
nagisa joined the channel
jseyfried
woboats: ping -- I have an idea re our earlier conversation about hygiene and `#[defines]`
srwalker101 has quit
petso has quit
srwalker101 joined the channel
nagisa
woboats: in response to your post, consider that people who are unsatisfied do complain, but those who aren’t don’t
nobody really needs to use modules either
nagisa has quit
nagisa joined the channel
petso joined the channel
srwalker101 has quit
woboats: for mod, its pretty orthogonal to `use` and any `mod` could be simply replaced by `use` from crate root :)
and if module is not `use`d, it does not exist
want to export modules from crate root? `pub use mod1::mod2::mod3` exports `mod1::mod2::mod3`, but not, say, `mod1::mod2::mod4`
petso has quit
woboats has quit
est31 joined the channel
kimundi has quit
petso joined the channel
blank_name1 has quit
nagisa has quit
petso has quit
petso joined the channel
petso has quit
petso joined the channel
petso has quit
petso joined the channel
petso has quit
Regexident joined the channel
Ericson2314 has quit
Regexident has quit
petso joined the channel
Regexident joined the channel
blank_name1 joined the channel
Aaronepower has quit
arielby joined the channel
kimundi joined the channel
petso has quit
ubsandroid_ joined the channel
ubsandroid has quit
blank_name1 has quit
petso joined the channel
niconii has quit
petso has quit
petso joined the channel
petso has quit
petso joined the channel
ubsandroid_ has quit
ubsandroid joined the channel
petso has quit
Regexident has quit
ubsandroid_ joined the channel
ubsandroid has quit
ubsandroid_ has quit
ubsandroid joined the channel
arielby has quit
arielby joined the channel
arielby has quit
arielby joined the channel
arielby has quit
nmatsakis
aturon: this `pub(restricted)` thing is so annoying
aturon
nmatsakis: yes :(
nmatsakis: i personally feel pretty ok about restricting to crate, super and self
eddyb
and I like the perversion with braces
aturon
eddyb: ugh please no :P
eddyb
ok it is a bit ugly
aturon
i promise you it will be widely seen as a shark-jumping moment
nmatsakis: it might be worth spending some time at lang team meeting today to see if we can get to the bottom of it
nmatsakis
aturon: I feel .. bah. I guess I'm ok w/ crate/super/self. I know in practice it'll be tolerable, it just annoys me that we are awfully close to a general model, but we wind up with these sort of artificial restrictions that force you into a shallower module hierarchy
at that level I am tempted to just make `crate` a new kind of decl, similar to `pub`, though it doesn't solve the `pub(self)` idea
that reminds me, I need to get back to trying to write up the virtual struct stuff we were talking about, which has already escaped my head. I'm glad I took some detailed notes.
aturon
nmatsakis: yeah, i hear you re: artificial restrictions. though if we allow `super::super` etc, at least in principle it's fully expressive
we could consider the lookahead idea instead...
nmatsakis
yeah, I guess super::super does retain the principle
just seems ... super annoying
somehow once it gets past one ../ I am lost
aturon
yeah it'd be super ugly :)
nmatsakis has a limited stack depth in his head or something
nmatsakis
it's annoying that we have to pick something and will be kind of stuck with it becuase of macros
if we adopt that visibility proposal, I guess
simulacrum
I also can't comprehend beyond about one level of super
It's why I love rust's preference for from-root paths instead of NodeJS' relative paths
aturon
FWIW petrochenkov did some analysis on common visibility and IIRC crate + super covered the vast majority
nmatsakis: thoughts on lookahead hackery instead?
nmatsakis
I'm sure it will be the vast majority, yes
it's the princple of the thing =)
re: lookahead, well, I think it is ok for a real parser, but I am concerned about doing it in the macro parser,
I'm trying to decide why
I guess because it seems to imply a grammar
(tbh current state of macro parser is kin dof confusing anyway in this area)
I really really wish we had not tied fragment specifies to rust grammar
hmm
aturon
i mean, it'll be a burden to other tooling as well
nmatsakis
I mean I guess you could just imagine it being defined as greedy
yeah, it will
I guess I am ok w/ crate/super/self -- the main thing I was weighing yesteday is whether, if we went that road,
I would rather have `crate fn foo()`
aturon
cc pnkfelix
nmatsakis
which would then also allow `crate impl Foo { ... }` for inherent impls local to a crate
aturon
nmatsakis: yeah, i was just thinking about that
nmatsakis
which I dearly want :)
brson joined the channel
seems a shame to lose pub(self) except that this is a horrible embarassing hack anyway that I'd probably never be able to recommend with a straight-face
eddyb
huh
nmatsakis: isn't pub(self) the private default?
ubsandroid_ joined the channel
nmatsakis
yes, the idea was that we could hack privacy into things like traits/ enums which (imo) should not have had defaults, but do