This increases regionck performance greatly - type-checking on
librustc decreased from 9.1s to 8.1s. Because of Amdahl's law,
total performance is improved only by about 1.5% (LLVM wizards,
this is your opportunity to shine!).
before:
576.91user 4.26system 7:42.36elapsed 125%CPU (0avgtext+0avgdata 1142192maxresident)k
after:
566.50user 4.84system 7:36.84elapsed 125%CPU (0avgtext+0avgdata 1124304maxresident)k
I am somewhat worried really need to find out why we have this Red Queen's
Race going on here. Originally I suspected it may be a problem from RFC1214's
warnings, but it seems to be an effect from other changes.
However, the increase seems to be mostly in LLVM's time, so I guess
it's the LLVM wizards' problem.
Refactors the "desugaring" of closures to expose the types of the upvars. This is necessary to be faithful with how actual structs work. The reasoning of the particular desugaring that I chose is explained in a fairly detailed comment.
As a side-effect, recursive closure types are prohibited unless a trait object intermediary is used. This fixes#25954 and also eliminates concerns about unrepresentable closure types that have infinite size, I believe. I don't believe this can cause regressions because of #25954.
(As for motivation, besides #25954 etc, this work is also intended as refactoring in support of incremental compilation, since closures are one of the thornier cases encountered when attempting to split node-ids into item-ids and within-item-ids. The goal is to eliminate the "internal def-id" distinction in astdecoding. However, I have to do more work on trans to really make progress there.)
r? @nrc
This commit finalizes the work of the past commits by fully moving the fulfillment context into
the InferCtxt, cleaning up related context interfaces, removing the Typer and ClosureTyper
traits and cleaning up related intefaces
This first patch starts by moving around pieces of state related to
type checking. The goal is to slowly unify the type checking state
into a single typing context. This initial patch moves the
ParameterEnvironment into the InferCtxt and moves shared tables
from Inherited and ty::ctxt into their own struct Tables. This
is the foundational work to refactoring the type checker to
enable future evolution of the language and tooling.
particular to treat an AutoUnsize as as kind of "instantaneous" borrow
of the value being unsized. This prevents us from feeding uninitialized
data.
This caused a problem for the eager reborrow of comparison traits,
because that wound up introducing a "double AutoRef", which was not
being thoroughly checked before but turned out not to type check.
Fortunately, we can just remove that "eager reborrow" as it is no longer
needed now that `PartialEq` doesn't force both LHS and RHS to have the
same type (and even if we did have this problem, the better way would be
to lean on introducing a common supertype).
Disallow writing through mutable pointers stored in non-mut Box.
Fix#14270
The fix works by making `cmt::freely_aliasable` result more fine-grained.
Instead of encoding the aliasability (i.e. whether the cmt is uniquely writable or not) as an option, now pass back an enum indicating either: 1. freely-aliasable (thus not uniquely-writable), 2. non-aliasable (thus uniquely writable), or 3. unique but immutable (and thus not uniquely writable, according to proposal from issue #14270.)
This is all of course a giant hack that will hopefully go away with an eventually removal of special treatment of `Box<T>` (aka `ty_unique`) from the compiler.
