* 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
Add support for cross import modules by ingesting swiftoverlay files for
the cross import into CAS file system.
The long-term better fix will be just passing the cross import
information from scanner to swift-frontend so frontend doesn't need to
read overlay files again to figure out the cross import module.
rdar://123839248
Renamed "getUsesMoveableValueDebugInfo" to "usesMoveableValueDebugInfo".
Clarifies the predicate from "does the receiver have the
usesMoveableValueDebugInfo field set?" to "does the receiver use moveable
value debug info?".
We preserve the current semantics that we have today by requiring that either all SILResultInfo are transferring or none are transferring. This also let me swap to @sil_transferring representation.
I did both of these things to fix SIL issues around transferring.
It also ensures that we now properly emit
Add an experimental option to tell dependency scanner to report clang
cc1 args should be used to construct clang importer in all constructed
swift-frontend tasks.
In certain cases (e.g. using arm64e interface to build arm64 target),
the target needs to be updated when building swiftinterface. Push the
target overwrite as early as possible to swiftinterface parsing by
providing a preferred target to relevant functions. In such cases, the
wrong target is never observed by other functions to avoid errors like
the sub-invocation was partially setup for the wrong target.
When caching build is enabled, teach dependency scanner to report
command-lines with `-direct-clang-cc1-module-build` so the later
compilation can instantiate clang importer with cc1 args directly. This
avoids running clang driver code, which might involve file system
lookups, which are the file deps that are not captured and might result
in different compilation mode.
rdar://119275464
Otherwise they may have module dependencies of their own which will not be detected by the scanner and included in the list of explicit inputs for compilation.
Our standard conception of suppressible features assumes we should
always suppress the feature if the compiler doesn't support it.
This presumes that there's no harm in suppressing the feature, and
that's a fine assumption for features that are just adding information
or suppressing new diagnostics. Features that are semantically
relevant, maybe even ABI-breaking, are not a good fit for this,
and so instead of reprinting the decl with the feature suppressed,
we just have to hide the decl entirely. The missing middle here
is that it's sometimes useful to be able to adopt a type change
to an existing declaration, and we'd like older compilers to be
able to use the older version of the declaration. Making a type
change this way is, of course, only really acceptable for
@_alwaysEmitIntoClient declarations; but those represent quite a
few declarations that we'd like to be able to refine the types of.
Rather than trying to come up with heuristics based on
@_alwaysEmitIntoClient or other sources of information, this design
just requires the declaration to opt in with a new attribute,
@_allowFeatureSuppress. When a declaration opts in to suppression
for a conditionally-suppressible feature, the printer uses the
suppression serially-print-with-downgraded-options approach;
otherwise it uses the print-only-if-feature-is-available approach.
we only check if the loaded module is built from a package interface. This is
not enough as a binary module could just contain exportable decls if built with
experimental-skip-non-exportable-decls, essentially resulting in content equivalent
to interface content. This might be made a default behavior so this PR requires
a module to opt in to allow non-resilient access by a participating client in the
same package.
Since it affects module format, SWIFTMODULE_VERSION_MINOR is updated.
rdar://123651270
There are scenarios where different compilers are distributed with
compatible serialization format versions and the same tag. Distinguish
swiftmodules in such a case by assigning them to different distribution
channels. A compiler expecting a specific channel will only read
swiftmodules from the same channel. The channels should be defined by
downstream code as it is by definition vendor specific.
For development, a no-channel compiler loads or defining the env var
SWIFT_IGNORE_SWIFTMODULE_REVISION skips this new check.
rdar://123731777
When scanning finds a dependency in the same package, do not load
public/private swiftinterface since they do not have the package level
decl to compile the current module. Always prefer package module (if
enabled), or use binary module, unless it is building a public/private
swiftinterface file in which case the interface file is preferred.
This also does some clean up to sync up the code path between implicit
and explicit module finding path.
rdar://122356964
Recently, unexpected binary module sharing between compilers that had
inconsistent views of the `PlatformKind` enumeration caused me to spend several
days debugging undefined behavior in the compiler. We should validate that
enumeration values decoded during deserialization are valid, and fail fast when
they are not to make debugging this kind of issue less time consuming. This
change just validates the `PlatformKind` value decoded for an `AvailableAttr`,
but more of deserialization could be updated to do similar validation.
Resolves rdar://123770273
For years, optimizer engineers have been hitting a common bug caused by passes
assuming all SILValues have a parent function only to be surprised by SILUndef.
Generally we see SILUndef not that often so we see this come up later in
testing. This patch eliminates that problem by making SILUndef uniqued at the
function level instead of the module level. This ensures that it makes sense for
SILUndef to have a parent function, eliminating this possibility since we can
define an API to get its parent function.
rdar://123484595
Fix a use-after-free bug in package-name extraction code, where the
BumpPtrAllcoator is destroyed before a StringRef that uses the allocated
name is used.
