This was not caught on main since I wasn't the tests with -sil-verify-all
enabled... but it seems that it was caught by the verifier on 5.9. I have since
enabled sil-verify-all by default on both of those tests
(noimplicitcopy_consuming_parameter.swift and
noimplicitcopy_borrowing_parameter.swift).
rdar://110364874
DI marks all of of the previously initialized properties and Raw SIL
lowering emits `destroy_addr` before calling init accessor for such
properties to destroy previously set value.
- Adds a missing check to `collectClassSelfUses` to find assign_or_init instructions;
- RawSIL lowering should start emitting access around synthesized member references.
- Record all properties listed in `accesses` as loads;
- Record all properties listed in `initialized` as init-or-assign;
- Detect situations when double-init could happen i.e. if one of
the properties listed in `initializes` attribute is explicitly
initialized before init accessor call.
This instruction is similar to AssignByWrapperInst, but instead of having
a destination operand, the initialization is fully factored into the init
function operand. Like AssignByWrapper, AssignOrInit has partial application
operands of both the initializer and the setter, and DI will lower the
instruction to a call based on whether the assignment is initialization or
a setter call.
Just the $*T -> $*@moveOnly T variant for addresses. Unlike the object version
this acts like a cast rather than something that provides semantics from the
frontend to the optimizer.
The reason why I am using a different instruction for addresses and objects here
is that the object checker doesnt have to deal with things like initialization.
drop_deinit ultimately only affects the semantics of its
destroy_value. Avoid generating releases for destroys in which the
deinit has been dropped. Instead, individually release the members.
The value `self` is mutable (i.e., var-bound) in
a `consuming` method. Since you're allowed to
reinitialize a var after consuming, that means
you were also naturally allowed to reinitialize
self after `discard self`. But that capability was
not intended; after you discard self you shouldn't
be reinitializing it, as that's probably a mistake.
This change makes reinitialization of `self`
reachable from a `discard self` statement an error.
rdar://106098163
As part of SE-390, you're required to write either:
- `consume self`
- pass self as a `consuming` parameter to a function
- `discard self`
before the function ends in a context that contains a
`discard self` somewhere. This prevents people from accidentally
invoking the deinit due to implicit destruction of `self` before
exiting the function.
rdar://106099027
Pattern matching as currently implemented is consuming, but that's not
necessarily what we want to be the default behavior when borrowing pattern
matching is implemented. When a binding of noncopyable type is pattern-matched,
require it to be annotated with the `consume` operator explicitly. That way,
when we introduce borrowing pattern matching later, we have the option to make
`switch x` do the right thing without subtly changing the behavior of existing
code. rdar://110073984
* Add @_used and @_section attributes for global variables and top-level functions
This adds:
- @_used attribute that flags as a global variable or a top-level function as
"do not dead-strip" via llvm.used, roughly the equivalent of
__attribute__((used)) in C/C++.
- @_section("...") attribute that places a global variable or a top-level
function into a section with that name, roughly the equivalent of
__attribute__((section("..."))) in C/C++.
Before the previous commit, we didn't see load [take] very often since it occurs
mostly on temporaries where we treat the copy_addr as the relevant take. Now
that we are checking temporaries though, we need to support this behavior.
this also fixes a bug where sometimes we simply emit
'consumed here' twice and other times we'd said 'other
consume here' for the same "consumed more than once"
message. so I went through and changed all of the 2nd
consumes into "consumed again".
rdar://109281444
- refer to a "consuming use" as simply a "consume", to reserve "use" for non-consuming uses.
- refer to "non-consuming uses" as just a "use".
- don't call it a "user defined deinit" and instead a "deinitializer" to match Sema
- be specific about what binding a closure is capturing that is preventing consumption.
rdar://109281444
- replaces "move-only" terminology with "noncopyable"
- replaces compiler jargon like "guaranteed parameters"
and "lvalue" with corresponding language-level notions
- simplifies diagnostics about closures.
and probably more.
rdar://109281444
Change SILGen to emit the `debug_value` instruction using the original inout
parameter address, instead of the `mark_must_check` inserted for move-only
parameters, because code in the MoveOnlyAddressChecker did not expect to
find the debug_value anywhere but on the original address. Update move-only
diagnostics so that they pick up the declaration name for a memory location
from any debug_value instruction if there are more than one. rdar://109740281
After a value is consumed, we emit a `debug_value undef` to indicate that the
variable value is no longer valid to the debugger. However, once a value is
reassigned, it becomes valid again, so emit a `debug_value %original_address` to
reassociate the variable with the valid memory location. rdar://109218404
Previously, we would emit a "compiler doesn't understand error" since we would
detect the escape and fail. That is the correct behavior here if the
partial_apply is not already identified as escaping by an earlier pass. But in
the case where we see that the partial_apply's callee was marked with
semantics::NO_MOVEONLY_DIAGNOSTICS, then we:
1. Suppress the "compiler doesn't understand error" for this specific
mark_must_check.
2. Suppress function wide the "copy of noncopyable type" error. Since we stopped
processing the mark_must_check that was passed to the partial_apply, we may
have left copies of noncopyable types on that mark_must_check value. This is
ok since the user will get the error, will recompile, and if any further show
up after they fix the inout escaping issue, they will be emitted
appropriately.
The previous commits in this series of changes reverted the allocbox to stack
change. Even with that though, we were treating the forming of the partial apply
and the destroys of the partial apply as the liveness requiring uses. This is
not the correct semantics for the non-escaping let closures. Instead, we want to
treat the passing off of the partial apply to another function or the invocation
of the partial apply as the liveness requiring uses. This makes sense since when
we capture a noncopyable type in the closure, we capture them as
inout_aliasable, that is as a pointer, so we are not going to actually use the
value until we call the partial_apply.
Originally this was a method to determine if we had emitted a diagnostic when
running our gather visitor. It was pretty footgunny since one could easily
forget to set it. Instead of doing that, we now maintain a counter in the
diagnostic emitter that counts how many diagnostics we have emitted and use that
to determine if during the walk if we emitted any additional diagnostics.
