Uh oh!
There was an error while loading. Please reload this page.
add task-local reference counted smart pointers - #6241
Conversation
brson
commented
May 5, 2013
Awesome. This is an important topic and I want to think about it a bit before r+ing, maybe get the ok from @pcwalton who has opinions on this subject. We're going to need something like this in the runtime so once it does not depend on the local heap we may end up moving it to I've been creating a lot of these container types with 🌞 |
thestinger
commented
May 5, 2013
@nikomatsakis and I came to the conclusion that the restriction of only containing An An Therefore we can have two constructors (one for |
thestinger
commented
May 5, 2013
Once #6251 lands I can remove the use of |
brson
commented
May 6, 2013
This looks fine to me, but I'd like someone else to sign off on it and FWIW this code could avoid using the init intrinsics by allocating with |
thestinger
commented
May 6, 2013
@brson: I think using ~ is required anyway if I'm going to allow |
To provide a reference counted pointer type with deterministic destruction once managed boxes are switched over to a garbage collector. Unlike managed boxes, these can be moved instead of just copied/cloned which is helpful for avoiding reference counts.
To provide a reference counted pointer type with deterministic destruction once managed boxes are switched over to a garbage collector. Unlike managed boxes, these can be moved instead of just copied/cloned which is helpful for avoiding reference counts. Needs #5601 to be fixed in order for safety to be provided without the current ugly workaround of making the pointers contain `Option<@()>` and `Option<@mut ()>` (which are just set to `None`). @brson: r?
update Applicability of `rc_buffer` lint from `MachineApplicable` to `Unspecified` `Unspecified` This changes `rc_buffer` from MachineApplicable to Unspecified ``` changelog: change [`rc_buffer`] to Unspecified. ``` fixesrust-lang#6241 --- changelog: change [`rc_buffer`] to Unspecified.
6257: Don't suggest extracting out 1-tuple enum variants r=matklad a=repnop Fixesrust-lang#6241. Co-authored-by: Wesley Norris <repnop@outlook.com>
To provide a reference counted pointer type with deterministic
destruction once managed boxes are switched over to a garbage
collector. Unlike managed boxes, these can be moved instead of just
copied/cloned which is helpful for avoiding reference counts.
Needs #5601 to be fixed in order for safety to be provided without the current ugly workaround of making the pointers contain
Option<@()>andOption<@mut ()>(which are just set toNone).@brson: r?