Witness tables for conformances that require runtime instantiation
should not be public, because it is an error to directly reference
such a symbol from outside the module.
Use a different mangling for witness table patterns and give them
non-public linkage.
Most of the work of this patch is just propagating metadata states
throughout the system, especially local-type-data caching and
metadata-path resolution. It took a few design revisions to get both
DynamicMetadataRequest and MetadataResponse to a shape that felt
right and seemed to make everything easier.
The design is laid out pretty clearly (I hope) in the comments on
DynamicMetadataRequest and MetadataResponse, so I'm not going to
belabor it again here. Instead, I'll list out the work that's still
outstanding:
- I'm sure there are places we're asking for complete metadata where
we could be asking for something weaker.
- I need to actually test the runtime behavior to verify that it's
breaking the cycles it's supposed to, instead of just not regressing
anything else.
- I need to add something to the runtime to actually force all the
generic arguments of a generic type to be complete before reporting
completion. I think we can get away with this for now because all
existing types construct themselves completely on the first request,
but there might be a race condition there if another asks for the
type argument, gets an abstract metadata, and constructs a type with
it without ever needing it to be completed.
- Non-generic resilient types need to be switched over to an IRGen
pattern that supports initialization suspension.
- We should probably space out the MetadataStates so that there's some
space between Abstract and Complete.
- The runtime just calmly sits there, never making progress and
permanently blocking any waiting threads, if you actually form an
unresolvable metadata dependency cycle. It is possible to set up such
a thing in a way that Sema can't diagnose, and we should detect it at
runtime. I've set up some infrastructure so that it should be
straightforward to diagnose this, but I haven't actually implemented
the diagnostic yet.
- It's not clear to me that swift_checkMetadataState is really cheap
enough that it doesn't make sense to use a cache for type-fulfilled
metadata in associated type access functions. Fortunately this is not
ABI-affecting, so we can evaluate it anytime.
- Type layout really seems like a lot of code now that we sometimes
need to call swift_checkMetadataState for generic arguments. Maybe
we can have the runtime do this by marking low bits or something, so
that a TypeLayoutRef is actually either (1) a TypeLayout, (2) a known
layout-complete metadata, or (3) a metadata of unknown state. We could
do that later with a flag, but we'll need to at least future-proof by
allowing the runtime functions to return a MetadataDependency.
Abstract type/heap metadata access goes into MetadataRequest.
Metadata access starting from a heap object goes into GenHeap.
Accessing various components of class metadata goes into GenClass
or MetadataLayout.
This includes global generic and non-generic global access
functions, protocol associated type access functions,
swift_getGenericMetadata, and generic type completion functions.
The main part of this change is that the functions now need to take
a MetadataRequest and return a MetadataResponse, which is capable
of expressing that the request can fail. The state of the returned
metadata is reported as an second, independent return value; this
allows the caller to easily check the possibility of failure without
having to mask it out from the returned metadata pointer, as well
as allowing it to be easily ignored.
Also, change metadata access functions to use swiftcc to ensure that
this return value is indeed returned in two separate registers.
Also, change protocol associated conformance access functions to use
swiftcc. This isn't really related, but for some reason it snuck in.
Since it's clearly the right thing to do, and since I really didn't
want to retroactively tease that back out from all the rest of the
test changes, I've left it in.
Also, change generic metadata access functions to either pass all
the generic arguments directly or pass them all indirectly. I don't
know how we ended up with the hybrid approach. I needed to change all
the code-generation and calls here anyway in order to pass the request
parameter, and I figured I might as well change the ABI to something
sensible.
This is simpler, because the native form of that last argument is: a
pointer to a buffer (*) of pointers (*) to witness tables, which is
modelled as a buffer of void *s. Thus, void ***.
The count of the number of witness tables was designed to be an
assertion/check that we've hooked up all the infrastructure
correctly. Everything is now implemented, and the assertion has never
triggered, so it can be removed, saving some work.
Fixes rdar://problem/38038928.
The conformance type's contextual type might not be canonical, such as in the case of substituting a `typealias` type for a generic placeholder, so ensure we get the canonical type for the conformance.
Factor out and reuse logic in the lowering of CondFailInst to emit
non-mergeable traps, everywhere we emit traps. This should address a
debugging quality issue with ambiguous ud2 instructions.
rdar://32772768
Emit nominal type access functions for imported types. These access
functions work with non-unique metadata references, so they perform
uniquing through the runtime on first access.
Fixes rdar://problem/36430234.
The dummy WitnessTableEntry for the protocol conformance descriptor
was used to ensure that the witness indices were offset by 1 (to
account for the protocol conformance descriptor), but it led to some
odd index adjustments. Instead, make the index adjustments explicit
when we're passing a WitnessIndex off to be loaded for a witness
table.
Extend witness tables with a pointer to the protocol conformance
descriptor from which the witness table was generated. This will allow
us to determine (for example) whether two witness tables were
generated from the same (or equivalent) conformances in the future, as
well as discover more information about the witness table itself.
Fixes rdar://problem/36287959.
Emit protocol conformance descriptors as separate symbols, rather than
inlining them within the section for protocol conformance records. We
want separate symbols for protocol conformances both because it is easier
to make them variable-length (as required for conditional
conformances) and because we want to reference them from witness
tables (both of which are coming up).
Only foreign classes and other imported types were making use of the
type metadata reference form in conformance records. Switch those over
to using nominal type descriptors, so we're using nominal type
descriptors for everything possible.
Only Objective-C-defined classes use a different representation now.
If the 'self' type is abstract, we still have to consider it as
a source of type metadata, because it might be a class-bound
generic parameter and the class might have generic parameters.
Fixes <rdar://problem/36093833>.
A witness table instantiation function can end up referring to
conditional requirements to, e.g., fill in base witness tables. Store
the conditional requirements first, and also bind the witness tables
for conditional requirements within the local scope so they can be
used directly. Fixes SR-6478.
When emitting a witness table accessor (e.g., an associated type metadata
or associated type conformance accessor) for a conditional conformance,
bind the conditional requirements so we can refer to them within that
accessor. Witness methods get this behavior already through the SelfWitnessTable parameter, but everything the witness table needs it.
Fixes most of SR-6478.
Rather than storing contextual types in the type witnesses and associated
conformances of NormalProtocolConformance, store only interface types.
@huonw did most of the work here, and @DougGregor patched things up to
complete the change.
This commit is mostly refactoring.
*) Introduce a new OptimizationMode enum and use that in SILOptions and IRGenOptions
*) Allow the optimization mode also be specified for specific SILFunctions. This is not used in this commit yet and thus still a NFC.
Also, fixes a minor bug: we didn’t run mandatory IRGen passes for functions with @_semantics("optimize.sil.never")
When calling an accessor, one has to pull the witness tables for each
conditional conformance requirement into a(n appropriately ordered) buffer that
is passed to the accessor. This is simple enough, if the appropriate
specialization of the relevant conformances are known, which the compiler didn't
track deep enough until now.
Conditional requirements shouldn't appear in witness_method's (IR) signature,
since they differ from conformance to conformance. PolymorphicConvention just
needs to consider these fulfilled by the self witness table, since they
can/should be pulled out of there.
This requires the witness table accessor function to gain two new parameters: a
pointer to an array of witness tables and their count. These are then passed down
to the instantiation function which reads them out of the array and writes them
into the newly-allocated witness table.
We use the count to assert that the number of conditional witness tables passed
in is what the protocol conformance expects, which is especially useful while
the feature is still experimental: it is a compiler/runtime bug if an incorrect
number is passed.
A concrete conformance may involve conditional conformances, which are witness
tables that we can access from the original conformance's one. We need to track
metadata and be able to follow it in a metadata path.
Pointers to the witness tables of any conditional conformances are placed into
private data, closest to the base pointer, i.e. wtable_ptr[-1] will be the
first (if any) conditional conformance witness table. These always need to be
provided by the context, and copied in when the witness table is instantiated,
making reserving space easy: increment the size of the private data section.
Finally, remove the parent type metadata argument from type
constructors.
Now that type constructors don't take a parent metadata pointer,
we can hit some asserts concerning type constructors that do not
have any parameters. This happens when you define a concrete type
in a fully-constrained extension of a generic type.
A more efficient ABI would use concrete type metadata for these
cases, but that would be a bigger change that we can do later, so
for now just relax these assertions.
This resolves a runtime crasher since a circular metadata case is
no longer circular. I renamed the crasher to reference the more
specific radar since the more general issue of circular metadata
is still unresolved.
This change makes it so that type metadata stores arguments
from each nesting depth. This means that parent metadata is
no longer needed to fulfill generic arguments, but we still
store a parent pointer in metadata and pass it to type
constructor functions at this point.
