I ran into this while fixing the parent commit when attempting to add the
interpreter test in this commit into the aforementioned parent commit.
rdar://107974302
Code can only locally interact with a mutable memory location within a
formal access, and is only responsible for maintaining its invariants
during that access, so the move-only address checker does not need to,
and should not, observe operations that occur outside of the access
marked with the `mark_must_check` instruction. And for immutable
memory locations, although there are no explicit formal accesses, that's
because every access must be read-only, so although individual
accesses are not delimited, they are all compatible as far as
move-only checking is concerned. So we can back out the changes to SILGen
to re-project a memory location from its origin on every access, a
change which breaks invariants assumed by other SIL passes.
This is used to teach the checker that the thing being checked is supposed to be
uninitialized at the mark_must_check point so that we don't put a destroy_addr
there.
The way this is implemented is that we always initially add
assignable_but_not_consumable but in DI once we discover that the assign we are
guarding is an init, we convert the assignable to its initable variant.
rdar://106525988
The reason why we are doing this is that:
1. For non-copyable types, switches are always at +1 for now.
2. non-copyable enums with deinits cannot be switched upon since that would
invalidate the deinit.
So deinits on non-copyable enums are just not useful at this point since you
cannot open the enum.
Once we make it so that you can bind a non-copyable enum at +0, we will
remove this check.
I added an experimental feature MoveOnlyEnumDeinits so tests that validate the
codegen/etc will still work.
rdar://101651138
* [IRGen] Make pointers to accessor functions in layout strings relative
rdar://106319336
Pointers embedded in static layout strings should always be relative, so layout strings can reside in read-only memory.
* Properly handle reference storage ownership
* Pass layout tag and metadata / type layout ppointers separately
* Layout string instantiation fully working
* Fix cases where hasLayoutString flag was not set when it should have
* Update include/swift/ABI/Metadata.h
I also added an interpreter test that validates that ref_element_addr works as
expected (I fixed that in an earlier commit, but did not add an interpreter
test).
rdar://106724277
the main things still left behind the experimental flag(s) are
- move-only classes (guarded by MoveOnlyClasses feature)
- noimplicitcopy
- the _borrow operator
We already did this for the situation without the begin_access. In truth, using
the terminator is a bit too wide, but it works for these sorts of arguments that
use assignable_but_not_consumable so for expediency (and since we are just
walking blocks), I just decided to do something quick.
rdar://106208343
We don't have any language or runtime support for noncopyable types as generic
or dynamic types yet, and existing reflection code almost certainly assumes it
can copy the values it's working with, and will trap or corrupt state if it does
so with noncopyable types. But a class can have noncopyable fields while the
type itself is copyable, and existing code assumes that it can use `Mirror` or
other reflection mechanisms to safely traverse the contents of an arbitrary
class.
Allow this sort of code to continue working, while still preparing for forward
compatibility with future runtimes that do support noncopyable generics, by
emitting the type references for fields using a function that probes the
address of a new symbol in the Swift runtime. The symbol will either be missing
or defined with an absolute address of zero in current or previous runtime
versions, but can be changed to a non-null address in the future.
The subexpression of a MaterializePackExpr (which is always a tuple value
currently) is emitted while preparing to emit a pack expansion expr, and its
elements are projected from within the dynamic pack loop. This means that a
materialized pack is only evaluated once, rather than being evaluated on
every iteration over the pack elements.
* [IRGen] Add layout strings for generic and resilient types
rdar://105837048
* Add some corner cases
* Add flag to enable generic instantiation and some fixes
* Fix resilient types
* Fix metadata accessor function pointers in combined layout strings
If there's a mismatch between the arguments we match and the arguments we actually have, we can end up indexing off the end of the argumentTypeNames vector. This can happen when an argument has a dependent generic type. Add a bounds check and print <unknown> when we're out of bounds to avoid crashing.
For correctness, we should match generic dependent types and add them to the arguments array, but we'll fix the crashes first.
rdar://104438524
