Implicitly imported decls may end up in inlinable code and break the
module API. This have been known to lead to deserialization crash and
could in theory break the generated swiftinterfaces files. Let's
explicitly check for such a case, keeping it to a warning until Swift 6
where we can make it an error.
rdar://95816286
Remove the allowUnavailable parameter to lookupConformance(), and instead
explicitly check the result for hasUnavailableConformance() in the places
where we used to pass 'false'.
Also, narrow down this check in those places to the Sendable protocol
only, fixing a regression with Hashable conformance synthesis.
Fixes rdar://problem/94460143.
See #59144 for more on why this is a bad idea.
Patch out the synthesized file unit accessor to only clear the source cache, then patch up all the places that were assuming they could iterate over the module's file list and see synthesized files.
rdar://94164512
Synthesized file units were designed for autodiff to emit synthesized declarations, and also to sidestep the design implications of doing so late in the compiler pipeline.
A call to materialize synthesized file units was added to the GetImplicitSendable request. This introduced a source of iterator invalidation into forEachFileToTypeCheck in whole-module builds. Any call to insert a new file into the module has the potential to cause the underlying SmallVector to reallocate.
This patch provides a narrow workaround that stops using iterators altogether in forEachFileToTypeCheck. However, this bug reveals a severe architectural flaw in the concept of a synthesized file unit. Iterating over the files in a module is an extremely common operation, and there now are myriad ways we could wind up calling a function that mutates the module's list of files in the process. This also means the number and kind of files being visited by compiler analyses is dependent upon whether a request that inserts these files has or has not been called.
This suggests the call to ModuleDecl::addFile in FileUnit::getOrCreateSynthesizedFile is deleterious and should be removed. Doing so will come as part of a larger refactoring.
rdar://94043340
When determining whether a superclass conforms to a particular protocol,
skip unavailable conformances. This way, we don't minimize away a
constraint that might only apply to subclasses of the specified
superclass.
Fixes rdar://91853658.
fec7a0b79b skipped all non-visible
`ValueDecls` but missed `ExtensionDecls`, which have the same issue.
Make sure to skip these too.
Resolves rdar://91279771.
This was benign with `Sendable`, but is not benign for the `Encodable`
and `Decodable` synthesis for distributed actors, which results in a
crash in TBD generation.
Fixes rdar://92008955.
Consider both frameworks and free floating modules imported from more
private path as SPI. This will make the compiler raise errors on public
imports of more private modules.
rdar://91904154
`SemaAnnotator` always attempts to retrieve the location of the decl. This
requires generating the USR, which needs to resolve the type. But that's
invalid in the presence of `@_implementationOnly`.
This most commonly comes up during index while building as even though
it skip internal decls, by that point it's too late (`SemaAnnotator` has
already tried retrieving the location). Other uses of
`SourceEntityWalker` should be unaffected as they only run over the
current decls in the current module.
For now skip walking internal top level decls from serialized modules in
`SemaAnnotator`. Consider expanding this to most, if not all,
`ASTWalker` clients in the future - it's unlikely `SemaAnnotator` is the
only one with this problem.
Resolves rdar://91279771.
When emitting a symbol graph file for a module that import modules via
`@_exported import`, emits those modules' symbols as well.
SR-15753
rdar://89547374
To help maintain source-compatibility, the presence of an `await` in
top-level code to kick the top-level code over to being a concurrent
context.
This, of course, means that in the test cases that exist today, they
will go back to behaving identically to how they did before I added all
of this because they don't have any awaits in the top-level. I'll be
adding new tests to verify the differences in behavior between swift 5,
swift 6, with and without async top level enabled in the next commit.
Introduce the `@preconcurrency` attribute name for `@_predatesConcurrency`,
which has been the favored name in the pitch thread so far. Retain the
old name for now to help smooth migration.
Rather than tacking the "add `@_predatesConcurrecy` to import"
diagnostic on to the prior diagnostic as a note, make it its own
remark. Then, ensure that we only emit this remark once per source
file per imported module, so we're not overwhelming the user.
* rename the CrossModuleSerializationSetup pass to simply CrossModuleOptimization
* remove the CMO specific serializer pass. Instead run the CrossModuleSerializationSetup pass directly before the standard serializer pass.
* correctly handle shared functions (e.g. specializations)
* refactoring
When looking for a Swift module on disk, we were scanning all module search paths if they contain the module we are searching for. In a setup where each module is contained in its own framework search path, this scaled quadratically with the number of modules being imported. E.g. a setup with 100 modules being imported form 100 module search paths could cause on the order of 10,000 checks of `FileSystem::exists`. While these checks are fairly fast (~10µs), they add up to ~100ms.
To improve this, perform a first scan of all module search paths and list the files they contain. From this, create a lookup map that maps filenames to the search paths they can be found in. E.g. for
```
searchPath1/
Module1.framework
searchPath2/
Module1.framework
Module2.swiftmodule
```
we create the following lookup table
```
Module1.framework -> [searchPath1, searchPath2]
Module2.swiftmodule -> [searchPath2]
```
We noticed some Swift clients rely on the serialized search paths in the module to
find dependencies and droping these paths altogether can lead to build failures like
rdar://85840921.
This change teaches the serialization to obfuscate the search paths and the deserialization
to recover them. This allows clients to keep accessing these paths without exposing
them when shipping the module to other users.
We've recently added the -experimental-hermetic-seal-at-link compiler flag,
which turns on aggressive dead-stripping optimizations and assumes that library
code can be optimized against client code because all users of the library
code/types are present at link/LTO time. This means that any module that's
built with -experimental-hermetic-seal-at-link requires all clients of this
module to also use -experimental-hermetic-seal-at-link. This PR enforces that
by storing a bit in the serialized module, and checking the bit when importing
modules.
