This one’s screwily specific--if you have:
* a private generic type,
* with a nested generic type in it,
* and the nested type conforms to a protocol with an associated type,
* and that associated type’s witness is a generic type,
* and some of the witness type’s generic parameters are generic parameters of the nested type,
demangling would fail. The problem is that the substitution machinery in the runtime would consider there to be three, not two, generic context depths involved. Depth 1, which should correspond to the nested type, would instead have no generic parameters. The fix is to skip over depths with zero generic parameters.
Fixes <rdar://problem/47773183>.
1) Enable tests that use `import Dispatch` on Linux. Add substitution
`%import-libdispatch` that needs to be used for all cross-platform
tests (i.e., tests that are intended to be run on other platforms
than Darwin) that do `import Dispatch` or enable thread sanitizer.
2) Make sure as many existing Dispatch and TSan tests as possible run on
Linux. Mark tests that would require substantial work with
`UNSUPPORTED: OS=linux-gnu`.
3) Add integration-style Swift test that shows that TSan finds a simple
race when using `Dispatch.async` incorrectly. A more complete test
suite for TSan's libdispatch support lives on the LLVM/compiler-rt
side.
rdar://problem/49177535
When we find a protocol conformance descriptor for a given type, make sure
we adjust to the conforming type of that descriptor (following the superclass
chain as needed) before instantiating the witness table.
Fixes rdar://problem/49741838.
Recent Swift uses 2 as the is-Swift bit when running on newer versions, and 1 on older versions. Since it's difficult or impossible to know what we'll be running on at build time, make the selection at runtime.
When emitting metadata for a Swift-defined @objc protocol that has
provided a specific Objective-C name (e.g., via @objc(renamed)),
mangle such protocols using their Objective-C names so they can be
found at runtime.
Only do this for metadata, because doing it anywhere else would cause
an ABI break. Fixes rdar://problem/47877748.
When calling a witness table accessor, IRGen was forcing the
conforming type to have complete metadata, even though only abstract
metadata is required for that query. This could cause cyclic metadata
dependencies when checking conditional conformances.
Fixes SR-5958.
This fixes the last couple of instances of the interpreter not being
passed to the python the scripts which makes them fail on OSes where
shebangs are not honoured (i.e. Windows)
When we encounter the demangle tree for a bound generic class type, look
through the "Type" node of the child tree before checking whether we
have an Objective-C class name. If we do have an Objective-C class name,
there is no way to preserve the generic arguments, so we ignore them and
return the (non-generic) class type.
Fixes rdar://problem/47028102.
We were creating a local Demangler instance, demangling a type name
using it, and then returning one of the resulting nodes to the caller.
Fixes rdar://problem/46817009.
Anonymous context descriptors were being treated as non-generic by
IRGen, which lead to problems for (file)private types within generic
types. Emit generic parameters and requirements for anonymous contexts
as well.
The runtime was mostly prepared for this, and the ABI already
accounted for it, so the runtime change is minor---it only affected
building a demangle tree from metadata.
Fixes rdar://problem/46853806.
Metadata uniquing might encounter witness tables that were distinctly
generated but come from identical descriptors. Handle this case in metadata
uniquing be looking into the protocol conformance descriptors themselves.
When a (file)private entity occurs inside a generic context, we still need
information about the genericity of the enclosing context to demangle
to metadata. Emit complete context descriptors for parents of anonymous
contexts.
Fixes rdar://problem/46109026.
The “old” mangling that is used for runtime names of @objc protocols
uses a simpler substitution scheme, so you can’t simply take a mangled
name from the new mangling and fix up the ends.
Fixes rdar://problem/45685649.
When checking conformance requirements against an @objc protocol, also
check for an @objc existential using protocol_conformsToProtocol().
Fixes rdar://problem/45685649.
We were strangely excluding protocols from being symbolically referenced
in the any-generic-type production, which meant that we could not resolve
(e.g.) associated type references to private protocols at runtime. Allow
protocol symbolic references in this position, and cope with it in the
demangler.
Fixes the rest of rdar://problem/44977236.
Rather than mapping all of the builtin floating-point and vector
types down to the type metadata symbols for power-of-two integers,
map down to the type metadata symbols provided by the runtime.
This allows us to round-trip type metadata <-> mangled name for all of
the builtin types that have such symbols. When we don’t have a runtime
symbol, we will fail to link if that builtin type gets referenced.
Expose symbols for metadata for the various builtin floating point types
and vector types. This is used by the demangler to handle builtin names.
This is a narrow fix for rdar://problem/45569984 (where we couldn’t
demangle a builtin vector type). A more extensive fix will require us
to add a general runtime facility for creating opaque type metadata
with specific size/alignment/stride/uniquing name.
(file)private entities are always available in the current file, so use
symbolic references to refer to them within mangled names (always). This
also eliminates problems stemming from our inability to demangle names
referring to private entities, because we (intentionally) don't allow
lookup for them.
Should fix rdar://problem/44977236.
@autoclosure is currently represented in mangled names, but the
demangle-to-metadata path did not handle it. Add basic support for it
so we can round-trip.
When searching the superclasses at runtime, e.g., to find a suitable
protocol conformance record, also consider the superclasses of CF
types, which were recorded in the metadata but otherwise unused.
(file)private entities are always available in the current file, so use
symbolic references to refer to them within mangled names (always). This
also eliminates problems stemming from our inability to demangle names
referring to private entities, because we (intentionally) don't allow
lookup for them.
Should fix rdar://problem/44977236.
When forming metadata for a nested generic type, gather all of the
generic arguments from the parent type “as written”, so that we can directly
map generic parameters to those generic arguments when they occur within
requirements. This allows us to demangle nested types within extensions
that have same-type constraints on generic parameters into type metadata.
Fixes rdar://problem/37170296.
Clang-importer-synthesized declarations get an extra tag character included in their mangling, which was not being preserved in type context descriptors. This caused runtime lookup for these synthesized types to fail. Fix this by adding the tag information to type context descriptors and teaching the runtime to match it up when fetching metadata by mangled name. Fixes rdar://problem/40878715.
Support demangling for types nested within some simple extension contexts.
Still does not support nested types within constrained extensions that
involve same-type constraints among generic parameters, nor
deeply-nested types in extensions. However, it fixes
rdar://problem/40071688.
This test broke when we switched to the swift-4.2-branch of Clang/LLVM.
It seems to be a problem with LLDB's CreateTargetWithFileAndArch() API
(rdar://problem/39960149). In the meantime, since this script does not
really need to specify the architecture, we can work around the problem
by using CreateTarget(). That's simpler anyway.
rdar://problem/38181372
Stress tests are, by definition, stressful. They intentionally burn a
lot of resources by using randomness to hopefully surface state machine
bugs. Additionally, many stress tests are multi-threaded these days and
they may attempt to use all of the available CPUs to better uncover
bugs. In isolation, this is not a problem, but the test suite as a whole
assumes that individual tests are single threaded and therefore running
multiple stress tests at once can quickly spiral out of control.
This change formalizes stress tests and then treats them like long
tests, i.e. tested via 'check-swift-all' and otherwise opt-in.
Finally, with this change, the CI build bots might need to change if
they are still only testing 'validation' instead of all of the tests.
I see three options:
1) Run all of the tests. -- There are very few long tests left these
days, and the additional costs seems small relative to the cost of
the whole validation test suite before this change.
2) Continue checking 'validation', now sans stress tests.
3) Check 'validation', *then* the stress tests. If the former doesn't
pass, then there is no point in the latter, and by running the stress
tests separately, they stand a better chance of uncovering bugs and
not overwhelming build bot resources.
