There were missing extractions from the Builtin (and other) modules.
This was actually caused by two issues:
* we did not visit all required modules, as for example the `Builtin`
module does not appear as being imported by anybody (together with
another mysterious `__Objc` module)
* moreover the `Builtin` module works internally by only creating
declarations on demand, and does not provide a list of its top level
declarations.
The first problem was solved by moving module collection to the actual
visiting. This may mean we extract less modules, as we only extract the
modules we actually use something from (recursively). This change can
be reverted if we feel we need it.
The second one was solved by explicitly listing the builtin symbols
encountered during a normal extraction. This does mean this list needs
to be kept up to date.
Add module context to all declarations. As keeping the `std::variant`s
required slightly more work for this, and we intended to remove them
any way, this is done in this change.
For background, we put those in as we were not extracting all modules at
the time, so we were missing some DB ids that we put back in by adding
the "defining" traps not only in the trap file related to the defining
module but also to the one where the declaration was used. Since then
we fixed module extraction, so this should not be needed any more.
We have found out there can be separate declarations (`VarDecl` or
`AccessorDecl`) which are effectively the same (with equal mangled name)
but come from different clang modules. This is the case for example
for glibc constants like `L_SET` that appear in both `SwiftGlibc` and
`CDispatch`.
In this patch, we simply avoid full deduplication in that case by
appending the module name to the trap key for non-swift modules.
A more solid solution should be found in the future.
This should cover `-merge-modules` mode.
Dumping of the configuration to the target files was moved to a
separate pair of header/source files, as now it is also done in
`SwiftOutputRewrite.cpp`.
Previously we were not extracting any `swiftmodule` file that was not
a system or a built-in one. This was done to avoid re-extracting
`swiftmodule` files that were built previously in the same build, but it
turned out to be too eager, as there are legitimate cases where a
non-system, non-built-in precompiled swift module can be used. An
example of that is the `PackageDescription` module used in Swift
Package Manager manifest files (`Package.swift`).
We now relax the test and trigger module extraction on all loaded
modules that do not have source files (we trigger source file extraction
for those). The catch, is that we also create empty trap files for
current output `swiftmodule` files (including possible alias locations
set up by XCode).
This means that if a following extractor run loads a previously built
`swiftmodule` file, although it will trigger module extraction, this
will however be skipped as it will find its target file already present
(this is done via the `TargetFile` semantics).
Deduplication of `ConcreteVarDecl` is triggered only if its
`DeclContext` is not local. This avoids a mangled name conflict.
Also added more thourough tests for `ConcreteVarDecl` and `ParamDecl`.
Firstly, this change reworks how inter-process races are resolved.
Moreover some responsability reorganization has led to merging
`TrapArena` and `TrapOutput` again into a `TrapDomain` class.
A `TargetFile` class is introduced, that is successfully created
only for the first process that starts processing a given trap output
file. From then on `TargetFile` simply wraps around `<<` stream
operations, dumping them to a temporary file. When `TargetFile::commit`
is called, the temporary file is moved on to the actual target trap
file.
Processes that lose the race can now just ignore the unneeded
extraction and go on, while previously all processes would carry out
all extractions overwriting each other at the end.
Some of the file system logic contained in `SwiftExtractor.cpp` has been
moved to this class, and two TODOs are solved:
* introducing a better inter process file collision avoidance strategy
* better error handling for trap output operations: if unable to write
to the trap file (or carry out other basic file operations), we just
abort.
The changes to `ExprVisitor` and `StmtVisitor` are due to wanting to
hide the raw `TrapDomain::createLabel` from them, and bring more
funcionality under the generic caching/dispatching mechanism.
Now `TypeRepr` is a final class in the AST, which is more or less just
a type with a location in code.
As the frontend does not provide a direct way to get a type from a
type representation, this information must be provided when fetching
the label of a type repr.
This meant:
* removing the type repr field from `EnumIsCaseExpr`: this is a virtual
AST node introduced in place of some kinds of `IsEpxr`. The type
repr is still available from the `ConditionalCheckedCastExpr` wrapped
by this virtual node, and we will rebuild the original `IsExpr` with
the IPA layer.
* some logic to get the type of keypath roots has been added to
`KeyPathExpr`. This was done to keep the `TypeRepr` to `Type` relation
total in the DB, but goes against the design of a dumb extractor. The
logic could be moved to QL in the future
* in the control flow library, `TypeRepr` children are now ignored. As
far as I can tell, there is no runtime evaluation going on in
`TypeRepr`s, so it does not make much sense to have control flow
through them.
