When Embedded Swift emits a symbol that was imported from another
module, ensure that the symbol is emitted as a weak definition. This
way, importing the same module (and using its symbol) into several
different modules doesn't cause duplicate-symbol errors at link time.
Rather, the linker will merge the different symbol definitions. This
makes Embedded Swift libraries work without resorting to
`-mergeable-symbols` or `-emit-empty-object-file`.
Removes a workaround previously added to handle types that were
overlooked during a serializability detection pass. Now that the
pass has been refactored with instruction visitor, we can replace
it with assert.
Resolves rdar://137711038
Currently, types from @_implementationOnly modules can be serialized into client modules if their
defining modules are SDK or system modules.
However, @_implementationOnly is intended to hide types from external clients, and may cause
the type’s metadata (e.g., field offsets) to be stripped. If such types are serialized and later
accessed by a client module, it can lead to linker errors due to the missing metadata.
This PR prevents all types imported with @_implementationOnly from being serialized.
Resolves rdar://144181455
Starting in Swift 6.0, `package` access level and `@_spiOnly` attribute have been increasingly used in import statements.
However, existing import filtering prevented serialization of package APIs that included such decls, leading to a
significant drop in overall serialization. This PR removes these restrictive filters, and allows decls from SDK or system
modules to be included in serialization.
rdar://130788606
In embedded mode CrossModuleOptimization must visit all vtable methods to make sure that no private/internal methods are serialized.
Fixes a compiler crash
rdar://143153941
When a keypath instruction was checked for serializability, its referenced function was
sometimes incorrectly deemed serializable when its referenced method had package or
public access level. This resulted in incorrectly serializing a function that dynamically
accesses a property on a generic type using key path. This PR fixes the issue by skipping
the access level check if the referenced function is determined to be un-serializable.
Resolves rdar://142950306
Currently, InstructionVisitor is only used during the serialization pass,
that comes after a pass that checks if instructions are serializable. This
causes inconsistencies as some types bypass the first pass but are processed
in the second, leading to assert fails.
This PR uses InstructionVisitor in the first pass to be exhaustive and to
be consistent with the coverage in the second pass. The downside is the visitor
is SILCloner, which introduces overhead of cloning and discarding insts per pass
-- a potential area for future refactoring.
Resolves rdar://130788545.
This PR ensures library-evolution is enabled for Package CMO; without it,
it previously fell back to regular CMO, which caused mismatching serialization
attributes if importing another module that had Package CMO enbaled, causing
an assert fail for loadable types.
Resolves rdar://135308288
This is needed so that client modules can de-virtualize witness method calls.
Fixes a false "cannot use a value of protocol type in embedded Swift" error.
rdar://133993657
Currently not all types are visited in canSerialize* calls, sometimes
resulting in an internal type getting @usableFromInline, which is
incorrect.
For example, for `let q = P() as? Q`, where Q is an internal class
inherting a public class P, Q is not visited in the canSerialize*
checks, thus resulting in `@usableFromInline class Q`; this is not
the intended behavior in the conservative mode used by PackageCMO
as it modifies AST.
To properly fix, instruction visitor needs to be refactored to do
both the "canSerialize" check (that visits all types) and serialize
or update visibility (modify AST in non-conservative modes).
This PR provides a short-term fix that prevents modifying AST, and
also ensures that the generated interfaces with PackageCMO flags
are not affected by the optimization or contain modified AST.
rdar://130292190
- Keep witness thunk linkage private for a package protocol member in SILGen.
- Optimize private/hidden functions during Package CMO; if they don't contain
references that have private/hidden symbols, serialize them and set the linkage
to shared. For unserialized witness thunks, set the linkage to package, so the
witness table itself can be serialized.
- Update witness table and vtable serialization.
Resolves rdar://129976582
Although I don't plan to bring over new assertions wholesale
into the current qualification branch, it's entirely possible
that various minor changes in main will use the new assertions;
having this basic support in the release branch will simplify that.
(This is why I'm adding the includes as a separate pass from
rewriting the individual assertions)
* Do not use [serialized_for_package] for witness method before Package CMO.
* Update v-table and witness-table and their entry serialization.
-- Allow non-serialized but visible entries in a serialized table so they
can be optimized by other SIL opt passes.
* Only serialize MethodInst if it has the right visibility.
Resolves rdar://129089105&129088935
inlining, generic/closure specialization, and devirtualization optimization passes.
SILFunction::canBeInlinedIntoCaller now exlicitly requires a caller's SerializedKind_t arg.
isAnySerialized() is added as a convenience function that checks if [serialized] or [serialized_for_pkg].
Resolves rdar://128704752
[serialized_for_package] if Package CMO is enabled. The latter kind
allows a function to be serialized even if it contains loadable types,
if Package CMO is enabled. Renamed IsSerialized_t as SerializedKind_t.
The tri-state serialization kind requires validating inlinability
depending on the serialization kinds of callee vs caller; e.g. if the
callee is [serialized_for_package], the caller must be _not_ [serialized].
Renamed `hasValidLinkageForFragileInline` as `canBeInlinedIntoCaller`
that takes in its caller's SerializedKind as an argument. Another argument
`assumeFragileCaller` is also added to ensure that the calle sites of
this function know the caller is serialized unless it's called for SIL
inlining optimization passes.
The [serialized_for_package] attribute is allowed for SIL function, global var,
v-table, and witness-table.
Resolves rdar://128406520
package-wide resilience domain if Package CMO is enabled.
The purpose of the attribute includes:
- Indicates that certain types such as loadable types are
allowed in serialized functions in resiliently built module
if the optimization is enabled, which are otherwise disallowed.
- Used during SIL deserialization to determine whether such
functions are allowed.
- Used to determine if a callee can be inlined into a caller
that's serialized without package-cmo, e.g. with an explicit
annotation like @inlinable, where the callee was serialized
due to package-cmo.
Resolves rdar://127870822
module when package serialization is enabled, return maximal resilience expansion
in SILFunction::getResilienceExpansion(). This allows aggregate types to be generated
as loadable SIL types which otherwise are address-only in a serialized function.
During type lowering, opaque flag setting is also skipped if package serialization
is enabled.
Resolves rdar://127400743
enabled. If two modules are in the same package and package cmo is enabled,
v-table or witness-table calls should not be generated at the use site in the
client module. Modified conformance serialization check to allow serializing
witness thunks.
Also reordered SIL functions bottom-up so the most nested referenced functions
can be serialized first. Allowed serializing a function if a shared definition
(e.g. function `print`). Added a check for resilient mode wrt struct instructions.
Added tests for SIL tables and resilient mode on/off.
rdar://124632670
* Add a new flag -experimental-package-cmo that requires -experimental-allow-non-resilient-access.
* Support serializing package decls for CMO in package if enabled.
* Only applies to default mode CMO.
* Unlike the existing CMO, package CMO can be built with -enable-library-evolution as package
modules are required to be built together in the same project.
* Create hasPublicOrPackageVisibility to opt in for package decls; needed for CMO, SILVerifier,
and other call sites that verify or determine codegen.
Resolves rdar://121976014
Decls with a package access level are currently set to public SIL
linkages. This limits the ability to have more fine-grained control
and optimize around resilience and serialization.
This PR introduces a separate SIL linkage and FormalLinkage for
package decls, pipes them down to IRGen, and updates linkage checks
at call sites to include package linkage.
Resolves rdar://121409846
access level for optimization: `public`. It requires an extra check for
the actual access level that was declared when determining serialization
since the behavior should be different.
This PR sets its effective access level to `package` as originally defined,
updates call sites to make appropriate acces level comparisons, and removes
`package` specific checks.
- Add a flag to the serialized module (IsEmbeddedSwiftModule)
- Check on import that the mode matches (don't allow importing non-embedded module in embedded mode and vice versa)
- Drop TBD support, it's not expected to work in embedded Swift for now
- Drop auto-linking backdeploy libraries, it's not expected to backdeploy embedded Swift for now
- Drop prespecializations, not expected to work in embedded Swift for now
- Use CMO to serialize everything when emitting an embedded Swift module
- Change SILLinker to deserialize/import everything when importing an embedded Swift module
- Add an IR test for importing modules
- Add a deserialization validation test
The relationship between the code in these two libraries was fundamentally circular, indicating that they should not have been split. With other changes that I'm making to remove circular dependencies from the CMake build graph I eventually uncovered that these two libraries were required to link each other circularly, but that had been hidden by other cycles in the build graph previously.
The check to avoid infinite recursion in case of call graph cycles didn't work correctly.
It didn't result in crashes, because the function also has an additional max-depth check, but it could lead to exponential complexity in some cases.
Unfortunately I don't have a test case for this fix.
For example:
```
public static var privateFunctionPointer: (Int) -> (Int) = { $0 }
```
Fixes a verifier crash and/or undefined symbol error
rdar://99493254
This is a follow-up of 1dfb3b1a2a.
We need to be more conservative about types as for functions, because types can also "produce" symbols, like direct field offsets, etc.
rdar://96953318
