This came up with SourceKit fuzz testing on the stdlib. I don't have a
reduced testcase right now.
Progress on <rdar://problem/19554069>.
Swift SVN r30882
- Produce more specific diagnostics relating to different kinds of invalid
- add a testcase, nfc
- Reimplement FailureDiagnosis::diagnoseGeneralMemberFailure in terms of
Not including r30787 means that we still generate bogus diagnostics like:
[1, 2, 3].doesntExist(0) // expected-error {{type 'Int2048' does not conform to protocol 'IntegerLiteralConvertible'}}
But it is an existing and separable problem from the issues addressed here.
Swift SVN r30819
r30787 causes our tests to time out; the other commits depend on r30787.
Revert "revert part of my previous patch."
Revert "Produce more specific diagnostics relating to different kinds of invalid"
Revert "add a testcase, nfc"
Revert "- Reimplement FailureDiagnosis::diagnoseGeneralMemberFailure in terms of"
Revert "Fix places in the constraint solver where it would give up once a single "
Swift SVN r30805
performMemberLookup, eliminating a ton of duplicated logic, but keeping the
same general behavior.
- Now that r30787 landed, we can have diagnoseGeneralMemberFailure inform
clients when a member lookup fails due to referencing a candidate decl of
ErrorType (i.e, it is already invalid somehow). When this happens, there is
no reason to diagnose a problem, because the original issue has been diagnosed
and anything we produce now is just garbage.
The second point cleans up a bunch of bogus diagnostics in the testsuite, which are
*actually* due to upstream error that are already diagnosed.
Swift SVN r30789
constraint failed, leaving a bunch of other solvable constraints laying
around in the system as inactive.
This is a problem for diagnostics emission, because it turns around and
reaches into the constraint system for some inactive constraint, assuming
that anything left could not be solved. The constraint system attempted to
solve this by taking the first failure and putting it into the failedConstraint
with the intention of driving diagnostics, but just because it happened to fail
first in constraint-solver-worklist-order doesn't mean it is the most pertinent
one to diagnose.
Swift SVN r30787
"unavoidable failure" path, along with Failure::DoesNotHaveNonMutatingMember and
just doing some basic disambiguation in CSDiags.
This provides some benefits:
- Allows us to plug in much more specific diagnostics for the existing "only has
mutating members" diagnostic, including producing notes for why the base expr
isn't mutable (see e.g. test/Sema/immutability.swift diffs).
- Corrects issues where we'd drop full decl name info for selector references.
- Wordsmiths diagnostics to not complain about "values of type Foo.Type" instead
complaining about "type Foo"
- Where before we would diagnose all failures with "has no member named", we now
distinguish between when there is no member, and when you can't use it. When you
can't use it, you get a vauge "cannot use it" diagnostic, but...
- This provides an infrastructure for diagnosing other kinds of problems (e.g.
trying to use a private member or a static member from an instance).
- Improves a number of cases where failed type member constraints would produce uglier
diagnostics than a different constraint failure would.
- Resolves a number of rdars, e.g. (and probably others):
<rdar://problem/20294245> QoI: Error message mentions value rather than key for subscript
Swift SVN r30715
Some types are born to mutating methods, and some types have mutating methods thrust upon them. Protocol extensions put classes into the latter category. A 'mutating' protocol extension method requires a mutable 'self' base even when applied to a class. Fix the synthesized getter for lazy properties so that it's only marked 'mutating' on value types, since otherwise it'd now be impossible to access lazy properties on immutable class references.
Fixes part of rdar://problem/21578832, allowing 'mutating' methods from protocol extensions to apply to mutable class references. Properties with mutating setters are still improperly allowed on immutable class reference bases though.
Swift SVN r30557
Requiring a variadic parameter to come at the end of the parameter
list is an old restriction that makes no sense nowadays, and which we
had all thought we had already lifted. It made variadic parameters
unusable with trailing closures or defaulted arguments, and made our
new print() design unimplementable.
Remove this restriction, replacing it with a less onerous and slightly
less silly restriction that we not have more than one variadic
parameter in a given parameter clause. Fixes rdar://problem/20127197.
Swift SVN r30542
If a function declaration possessed default parameters, and was invoked with a single argument expression that was modeled as a type variable, the compiler would often crash during type application. This was due to the fact that during simplification, we would bind the type variable to the full tuple type of the parameter list. Later on, during constraint application, we would then look to whatever expression created the type variable for information on its default arguments - even if no such thing was possible. (E.g., we would examine, say, an IfExpr expecting to find information on its default arguments.) In these cases, we should have instead been binding the argument type variable to the first element of the parameter tuple.
Swift SVN r30486
with no returns *must* be (), add a defaulting constraint
so that it will be inferred as () in the absence of
other possibilities.
The chief benefit here is that it allows better QoI when
the user simply hasn't yet written the return statement.
Doing this does regress a corner case where an attempt
to recover from an uncalled function leads to the
type-checker inferring a result for a closure that
doesn't make any sense at all.
Swift SVN r30476
which allows solving of a constraint system to succeed without emitting
errors in the face of ambiguous solutions. This is important for CSDiag
because it is in the business of trying to solve subexpressions of a global
expression - and it wants to know the difference between a subexpression
that is inherently impossible to solve, vs one that is simply ambiguous
because its context has been removed.
Use this in CSDiag's typeCheckChildIndependently() to provide it an
extra flag that enables this behavior. This is currently unused, so NFC
with this patch.
Swift SVN r30402
This is a travesty that makes it hard to reason about tuple vs. scalar
types. It's only use in our own code involved dealing with the
difference in element types between Set and Dictionary in
gyb-generated code. Ewww. Fixes rdar://problem/17963034.
Swift SVN r30074
The isDependentType() query is woefully misunderstood. Some places
seem to want it to mean "a generic type parameter of dependent member
type", which corresponds to what is effectively a type parameter in
the language, while others want it to mean "contains a type parameter
anywhere in the type". Tease out these two meanings in
isTypeParameter() and hasTypeParameter(), respectively, and sort out
the callers.
Swift SVN r29945
value member constraints do.
This fixes:
<rdar://problem/21662365> QoI: diagnostic for for-each over an optional sequence isn't great
before we'd produce:
t.swift:3:10: error: '[Int]?' does not have a member named 'Generator'
for x in array {
^
now we produce:
t.swift:3:10: error: value of optional type '[Int]?' not unwrapped; did you mean to use '!' or '?'?
for x in array {
^
Swift SVN r29902
It looks like we were checking in the wrong place, as a result we didn't
catch stuff like
class G<T : AnyObject> {}
_ = G<P>()
This would crash later in IRGen.
Make the conformsToProtocol() check do the right thing, and remove some
other miscellaneous diagnostics in the process. Also, make the
"type 'T' does not conform to protocol 'P'" diagnostic a bit more
detailed.
Unfortunately in a few instances we lose a more descriptive diagnostic to
a general 'cannot invoke 'foo' with argument list of type 'T'' error. The
argument matching diagnostics need to be addressed anyway though.
Fixes <rdar://problem/20311619>.
Swift SVN r29737
We don't know what's inside the existential type container, so we cannot
refer to uncurried instance members of the type without running into
theoretical difficulties.
Swift SVN r29378
For a concrete type A and protocol type P, A <c P now implies
A.Type <c P.Type, not just A.Type < P.Type. This in turn means
that A.Type.Type <c P.Type.Type. To make the coercion work,
recursively peel off metatype layers when collecting conformances
and in a similar situation in IRGen.
Swift SVN r29377
We were using getRValueType() and getRValueObjectType() when
setting up the constraint, but getRValueType() when simplifying
it. This led to a crash if the type was a single-argument tuple
with a named argument.
Swift SVN r29376
X.Protocol is an instance of Y.Type only if X conforms to Y. Since X
is a protocol, this is only true if X contains Y and Y is
self-conforming.
Note that this updates some tests that actually contained invalid code.
Fixes <rdar://problem/20915927>.
Swift SVN r29349
If 'x.init' appears as a member reference other than 'self.init' or 'super.init' within an initializer, treat it as a regular static member lookup for 'init' members. This allows a more explicit syntax for dynamic initializations; 'self.someMetatype()' looks too much like it's invoking a method. It also allows for partial applications of initializers using 'someMetatype.init' (though this needs some SILGen fixes, coming up next). While we're in the neighborhood, do some other correctness and QoI fixes:
- Only lookup initializers as members of metatypes, not instances, and add a fixit (instead of crashing) to insert '.dynamicType' if the initializer is found on an instance.
- Make it so that constructing a class-constrained archetype type correctly requires a 'required' or protocol initializer.
- Warn on unused initializer results. This seems to me like just the right thing to do, but is also a small guard against the fact that 'self.init' is now valid in a static method, but produces a newly-constructed value instead of delegating initialization (and evaluating to void).
Swift SVN r29344
If P is a protocol, calling static methods or constructors
via values of type P.Protocol makes no sense, so let's prohibit
this.
Fixes <rdar://problem/21176676>.
Swift SVN r29338
Rename existentialConformsToSelf() to existentialTypeSupported(). This
predicate is the "protocol has no Self or associated type requirements"
check, which is a looser condition than self-conformance. This was being
tested to see if the user could refer to the protocol via an existential
type.
The new existentialConformsToSelf() now checks for protocol being @objc,
and for the absence of static methods. This is used as part of the
argument type matching logic in matchType() to determine if the
existential can be bound to a generic type parameter.
The latter condition is stricter, for two reasons:
1) We allow binding existentials to multiple type parameters all sharing
the same generic type parameter T, so we don't want the user to be
able to see any static methods on T.
2) There is an IRGen limitation whereby only existentials without witness
tables can be passed in this manner.
Using the above, the representsNonTrivialGenericParameter() function
has been renamed to canBindGenericParamToExistential(). It now allows
an existential type to be bound to a generic type parameter only under
the following circumstances:
A) If the generic type parameter has no conformances, the match is allowed.
B) If the generic type parameter has at least one conformance, then all
of the conformances on the generic type parameter must be
existentialConformsToSelf() (condition 1 above), and all conformances
on the existential must be @objc (condition 2 above).
Fixes <rdar://problem/18378390> and <rdar://problem/18683843>, and lays
the groundwork for fixing a few other related issues.
Swift SVN r29337
variable has the must-be-materializable bit set if the old one does.
When assigning a fixed type to a type variable that must be
materializable, transfer the bit to any type variables within the fixed
type, as appropriate.
Add Options field to SavedTypeVariableBinding to save/restore type
variable options during solution.
<rdar://problem/21026806> Propagate MustBeMaterializable bit among type variables appropriately
Swift SVN r28883
Add a new option, TVO_MustBeMaterializable, to
TypeVariableType::Implementation, and set it for type variables
resulting from opening a generic type. This solution isn't complete (we
don't yet copy the non-materializable bit on unification of type
variables, and it's possible to bind a must-be-materializable type
variable to a type with type variables that later get bound to
non-materializable types) but it addresses all reported crashes for this
issue.
<rdar://problem/20807269> Crash in non-materializable type
Swift SVN r28792
This just deletes some code out of matchCallArguments to make sure that CSApply
and CSSimplify do the same argument matching. This code presumably
dated to a period of time when the stuff after it was just for error recovery,
but now we do argument reordering matching stuff after this point.
Swift SVN r28670
Modules occupy a weird space in the AST now: they can be treated like
types (Swift.Int), which is captured by ModuleType. They can be
treated like values for disambiguation (Swift.print), which is
captured by ModuleExpr. And we jump through hoops in various places to
store "either a module or a decl".
Start cleaning this up by transforming Module into ModuleDecl, a
TypeDecl that's implicitly created to describe a module. Subsequent
changes will start folding away the special cases (ModuleExpr ->
DeclRefExpr, name lookup results stop having a separate Module case,
etc.).
Note that the Module -> ModuleDecl typedef is there to limit the
changes needed. Much of this patch is actually dealing with the fact
that Module used to have Ctx and Name public members that now need to
be accessed via getASTContext() and getName(), respectively.
Swift SVN r28284
checking checked cast via bridging. It prevented bridging upcasts using
'as!' from typechecking; we should emit an 'as!'->'as' warning instead.
Also, use ExplicitConversion constraint instead of Conversion when
determining whether a checked cast can be carried out unconditionally.
This matches the constraint used after applying the 'as!'->'as' fixit.
(Also, fix the error that was responsible for breaking
the expr/cast/bridged.swift test.)
<rdar://problem/19813772>
Swift SVN r28034
checking checked cast via bridging. It prevented bridging upcasts using
'as!' from typechecking; we should emit an 'as!'->'as' warning instead.
Also, use ExplicitConversion constraint instead of Conversion when
determining whether a checked cast can be carried out unconditionally.
This matches the constraint used after applying the 'as!'->'as' fixit.
<rdar://problem/19813772>
Swift SVN r28028
