explicitly written and disagree with context, and when context provides a
non-explicitly written type that disagrees with the body of the closure.
Swift SVN r30984
using it to improve closure diagnostics by inferring the types of otherwise
untyped closure paramdecls from this context information. This
resolves:
<rdar://problem/20371273> Type errors inside anonymous functions don't provide enough information
producing
error: binary operator '==' cannot be applied to operands of type 'Int' and 'UInt'
note: overloads for '==' exist with these partially matching parameter lists: (UInt, UInt), (Int, Int)
and:
<rdar://problem/20978044> QoI: Poor diagnostic when using an incorrect tuple element in a closure
producing:
error: value of tuple type '(Int, Int)' has no member '2'
and probably a lot more. We're still limited from getting things like "foo.map {...}" because
we're not doing type subsitutions from the base into the protocol extension member.
Swift SVN r30971
as a proper error, and change it to not be incorrect. Multi-statement
closures *only* need a return type if they cannot be inferred.
This fixes:
<rdar://problem/22086634> "multi-statement closures require an explicit return type" should be an error not a note
Swift SVN r30937
down to call argument lists that have more than one operand (heavily leveraging
"computeTupleShuffle"). This resolves a great number of QoI radars, including
things like:
<rdar://problem/19981782> QoI: poor diagnostic for call to memcmp with UInt length parameter
where we used to produce:
error: cannot invoke 'memcmp' with an argument list of type '([UInt8], [UInt8], UInt)'
return memcmp(left, right, UInt(left.count)) == 0
^
note: expected an argument list of type '(UnsafePointer<Void>, UnsafePointer<Void>, Int)'
but now we produce:
error: cannot convert value of type 'UInt' to expected argument type 'Int'
return memcmp(left, right, UInt(left.count)) == 0
^~~~~~~~~~~~~~~~
which is more "to the point"
Swift SVN r30930
argument. For now we start with some of the most simple cases: single argument
calls. This dramatically improves the QoI for error messages in argument lists,
typically turning a error+note combo into a single specific error message.
Some minor improvements coming (and also generalizing this to n-ary calls), but it
is nice that all the infrastructure is starting to come together...
Swift SVN r30905
which we have a contextual type that was the failure reason. These are a bit
longer but also more explicit than the previous diagnostics.
Swift SVN r30669
version of the new CTP_ReturnStmt conversion, used to generate return-specific
diagnostics. Now that we have a general solution, we can just use that.
This improves diagnostics in returns for accessors, since they were apparently
not getting the bit set.
Swift SVN r30665
conversion failures, making a bunch of diagnostics more specific and useful.
UnavoidableFailures can be very helpful, but they can also be the first constraint
failure that the system happened to come across... which is not always the most
meaningful one. CSDiag's expr processing machinery has a generally better way of
narrowing down which ones make the most sense.
Swift SVN r30647
type check the subexpressions of a callexpr more consistently,
always checking the arguments independently (not just if one argument
is inout). This routes around issues handling tuples, and brings more
consistency to the experience. Factor this logic out and use it for
operators and subscripts as well.
Swift SVN r30583
independently (not just if one argument is inout). This routes around issues handling tuples,
and brings more consistency to the experience. Factor this logic out and use it for operators
and subscripts as well.
This improves a small collection of diagnostics, including the infamous:
// Infer incompatible type.
- func6(fn: {a,b->Float in 4.0 }) // expected-error {{cannot convert return expression of type 'Double' to expected return type 'Float'}}
+ func6(fn: {a,b->Float in 4.0 }) // expected-error {{cannot invoke 'func6' with an argument list of type '(fn: (_, _) -> Float)'}}
+ // expected-note @-1 {{expected an argument list of type '(fn: (Int, Int) -> Int)'}}
Swift SVN r30570
diagnose problems inside of them instead of punting on them completely.
This leads to substantially better error messages in many cases, fixing:
<rdar://problem/19870975> Incorrect diagnostic for failed member lookups within closures passed as arguments ("(_) -> _")
<rdar://problem/21883806> Bogus "'_' can only appear in a pattern or on the left side of an assignment" is back
<rdar://problem/20712541> QoI: Int/UInt mismatch produces useless error inside a block
and possibly others. We are not yet capitalizing on available type information we do
have about closure exprs, so there are some cases where we produce
"error: type of expression is ambiguous without more context"
when this isn't strictly true, but this is still a huge step forward.
Swift SVN r30547
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
fixing:
<rdar://problem/20789423> Unclear diagnostic for multi-statement closure with no return type
<rdar://problem/21829141> BOGUS: unexpected trailing closure
<rdar://problem/21784170> Incongruous `unexpected trailing closure` error in `init` function which is cast and called without trailing closure.
Swift SVN r30443
return statements, or a return statement with no operand.
Also, fix a special-case diagnostic about converting a return
expression to (1) only apply to converting the actual return
expression, not an arbitrary sub-expression, and (2) use the
actual operand and return types, not the drilled-down types
that caused the failure.
Swift SVN r30420
This teaches overload constraint diagnosis to look at the resolved anchor
expression that fails (instead of assuming that it is the expr itself) and
walks up the AST to find the applyexpr in question. This allows us to give
much more specific diagnostics for overload resolution failures, and to give
much more specific location information.
Where before my recent patches we used to produce:
t.swift:2:3: error: cannot invoke 'assert' with an argument list of type '(Bool, String)'
assert(a != nil, "ASSERT COMPILATION ERROR")
^
t.swift:2:9: note: expected an argument list of type '(@autoclosure () -> Bool, @autoclosure () -> String, file: StaticString, line: UWord)'
assert(a != nil, "ASSERT COMPILATION ERROR")
^
with this and the other recent patches, we now produce:
t.swift:2:12: error: cannot invoke '!=' with an argument list of type '(Int, nil)'
assert(a != nil, "ASSERT COMPILATION ERROR")
~~^~~~~~
Swift SVN r29792
path associated with them, and to dig the expression the constraint refers to out
of the locator. Also teach simplifyLocator how to simplify closureexpr results out.
This eliminates a class of completely bogus diagnostics where the types reported
don't make any sense, resolving a class of radars like 19821875, where we now
produce excellent diagnostics.
That said, we still pick constraints to report that are unfortunate in some cases,
such as the example in expr/closure/closures.swift.
Swift SVN r29757
- Fix TypeCheckExpr.cpp to be more careful when propagating sugar from an
argument to the result of the function. We don't want to propagate parens,
because they show up in diagnostics later.
- Restructure FailureDiagnosis::diagnoseFailure() to strictly process the tree
in depth first order. Before it would only do this if contextual typing was
unavailable, leading to unpredictable inconsistencies between diagnostics.
- Always perform diagnoseContextualConversionError early, as part of the thing
that calls the visitor, instead of in each visit method. This may change in
the future, but is a simplification for now.
- Make the operator processing code handle the "candidate is an exact match"
case by emitting a diagnostic indicating that the result type of the operator
must not match expectations, instead of emitting the silly things like
"binary operator '&' cannot be applied to two Int operands" which is obviously
false.
These changes lead to minor improvements across the testsuite, and should make the
diagnostics more predictable for more complex real-world ones, but I haven't gone
through the radars yet.
Major missing pieces:
- CallExpr isn't using the same logic that the operators are.
- When you have a near match (only one argument mismatches) we should specifically
complain about that argument, instead of spewing an entire argument list.
- The noescape function attr diagnostic is being emitted twice now.
Swift SVN r29733
win from this other than simplification. Some minor wins are that we handle varargs
better and don't get extraneous ()'s in types in some cases.
Swift SVN r29729
- If the closure being rewritten was nested inside of another
closure, we would rewrite the nested closure into a new closure
having cs.DC as its parent, rather than the DeclContext of the
outer closure as required.
- If the closure being rewritten had nested closures, the parent
DeclContext of the nested closures was set to the old closure,
not the new one.
Fix both problems by having coerceClosureExprToVoid() modify the
closure in place instead of creating a new closure.
Fixes <rdar://problem/20931915>.
Swift SVN r29563
We no do not require "self." for closures capturing self in static/class methods.
While we do actually capture the metatype more than we should (rdar://21030087),
this doesn't matter to the developer, since this capture cannot cause a cycle
in the reference graph that they should have to reason about.
Swift SVN r28804
This sets the location of the implicit closure decls (like $0) to being the location
of the { in a ClosureExpr, instead of the location of the first use. The capture tracker
uses source location information of the decl and the DeclRefExpr to determine if the
referenced value was captured too early, which is what is causing this incorrect error.
Swift SVN r28802
<rdar://problem/15975935> warning that you can use 'let' not 'var'
<rdar://problem/18876585> Compiler should warn me if I set a parameter as 'var' but never modify it
<rdar://problem/17224539> QoI: warn about unused variables
This uses a simple pass in MiscDiagnostics that walks the body of an
AbstractFunctionDecl. This means that it doesn't warn about unused properties (etc),
but it captures a vast majority of the cases.
It also does not warn about unused parameters (as a policy decision) because it is too noisy,
there are a variety of other refinements that could be done as well, thoughts welcome.
Swift SVN r28412
If you want to make the parameter and argument label the same in
places where you don't get the argument label for free (i.e., the
first parameter of a function or a parameter of a subscript),
double-up the identifier:
func translate(dx dx: Int, dy: Int) { }
Make this a warning with Fix-Its to ease migration. Part of
rdar://problem/17218256.
Swift SVN r27715
The rule changes are as follows:
* All functions (introduced with the 'func' keyword) have argument
labels for arguments beyond the first, by default. Methods are no
longer special in this regard.
* The presence of a default argument no longer implies an argument
label.
The actual changes to the parser and printer are fairly simple; the
rest of the noise is updating the standard library, overlays, tests,
etc.
With the standard library, this change is intended to be API neutral:
I've added/removed #'s and _'s as appropriate to keep the user
interface the same. If we want to separately consider using argument
labels for more free functions now that the defaults in the language
have shifted, we can tackle that separately.
Fixes rdar://problem/17218256.
Swift SVN r27704
This post-hoc diagnostic replaces r24915, r25045, and r25054 by doing a
very basic check for representation incompatibility between two types.
More cases can be added as necessary.
rdar://problem/19600325, again.
Swift SVN r25117
John pointed out that messing with the type checker's notion of "subtype"
is a bad idea. Instead, we should just have a separate check for ABI
compatibility...and eventually (rdar://problem/19517003) just insert the
appropriate thunks rather than forcing the user to perform the conversion.
I'm leaving all the tests as they are because I'm adding a post-type-checking
diagnostic in the next commit, and that should pass all the same tests.
Part of rdar://problem/19600325
Swift SVN r25116
This re-applies r24987, reverted in r24990, with a fix for a spuriously-
introduced error: don't use a favored constraint in a disjunction to avoid
applying a fix. (Why not? Because favoring bubbles up, i.e. the
/disjunction/ becomes favored even if the particular branch is eventually
rejected.) This doesn't seem to affect the outcome, though: the other
branch of the disjunction doesn't seem to be tried anyway.
Finishes rdar://problem/19600325
Swift SVN r25054
And even if we don't suggest wrapping in a closure (say, because there's
already a closure involved), emit a more relevant diagnostic anyway.
(Wordsmithing welcome.)
Wrapping a function value in a closure essentially explicitly inserts a
conversion thunk that we should eventually be able to implicitly insert;
that's rdar://problem/19517003.
Part of rdar://problem/19600325
Swift SVN r24987
Previously, trailing closures would try to match the first parameter
of (possibly optional) function type that didn't seem to have an
argument already, but in practice this broke when there were
parameters with default arguments before the function parameter.
The new rule is far simpler: a trailing closure matches the last
parameter. Fixes rdar://problem/17965209.
Swift SVN r24898
- Closures that are comprised of only a single return statement are now considered to be "single expression" closures. (rdar://problem/17550847)
- Unannotated single expression closures with non-void return types can now be used in void contexts. (rdar://problem/17228969)
- Situations where a multi-statement closure's type could not be inferred because of the lack of a return-type annotation are now properly diagnosed. (rdar://problem/17212107)
I also encountered a number of crashers along the way, which should now be fixed.
Swift SVN r24817
a vararg subpattern of a TuplePattern should be a TypedPattern
Check for a vararg subpattern to be a typed pattern seemed to be missing in closure arguments parsing.
Swift SVN r24733
Most tests were using %swift or similar substitutions, which did not
include the target triple and SDK. The driver was defaulting to the
host OS. Thus, we could not run the tests when the standard library was
not built for OS X.
Swift SVN r24504
