Log message intended for compiler developers No file/line/column stuff

Log messages intended for end users. No file/line/column stuff.

Diagnostics from the compiler. This constructor is very powerful as it allows the construction of a MessageClass with a completely arbitrary permutation of Severity and DiagnosticReason. As such, users are encouraged to use the mkMCDiagnostic smart constructor instead. Use this constructor directly only if you need to construct and manipulate diagnostic messages directly, for example inside Error. In all the other circumstances, especially when emitting compiler diagnostics, use the smart constructor.

The Maybe DiagnosticCode field carries a code (if available) for this diagnostic. If you are creating a message not tied to any error-message type, then use Nothing. In the long run, this really should always have a DiagnosticCode. See Note [Diagnostic codes].

Ignore this message, for example in case of suppression of warnings users don't want to see. See Note [Suppressing Messages]

Born as a warning.

Warning was enabled with the flag.

Born as an error.

An "unknown" hint. This type constructor allows arbitrary -- hints to be embedded. The typical use case would be GHC plugins -- willing to emit hints alongside their custom diagnostics.

Suggests adding a particular language extension. GHC will do its best trying to guess when the user is using the syntax of a particular language extension without having the relevant extension enabled.

Example: If the user uses the keyword "mdo" (and we are in a monadic block), but the relevant extension is not enabled, GHC will emit a 'SuggestExtension RecursiveDo'.

Test case(s): parsershould_failT12429, parsershould_failT8501c, parsershould_failT18251e, ... (and many more)

Suggests possible corrections of a misspelled pragma. Its argument represents all applicable suggestions.

Example:

Test case(s): parsershould_compileT21589

Suggests that a monadic code block is probably missing a "do" keyword.

Example: main = putStrLn "hello" putStrLn "world"

Test case(s): parsershould_failT8501a, parsershould_failreadFail007, parsershould_failInfixAppPatErr, parsershould_failT984

Suggests that a "let" expression is needed in a "do" block.

Test cases: None (that explicitly test this particular hint is emitted).

Suggests to add an ".hsig" signature file to the Cabal manifest.

Triggered by: DriverUnexpectedSignature, if Cabal is being used.

Example: See comment of DriverUnexpectedSignature.

Test case(s): driver/T12955

Suggests to explicitly list the instantiations for the signatures in the GHC invocation command.

Triggered by: DriverUnexpectedSignature, if Cabal is not being used.

Example: See comment of DriverUnexpectedSignature.

Test case(s): driver/T12955

Suggests to use spaces instead of tabs.

Triggered by: PsWarnTab.

Examples: None Test Case(s): None

Suggests adding a whitespace after the given symbol.

Examples: None Test Case(s): parsershould_compileT18834a.hs

Suggests adding a whitespace around the given operator symbol, as it might be repurposed as special syntax by a future language extension. The second parameter is how such operator occurred, if in a prefix, suffix or tight infix position.

Triggered by: PsWarnOperatorWhitespace.

Example: h a b = a+b -- not OK, no spaces around +.

Test Case(s): parsershould_compileT18834b.hs

Suggests wrapping an expression in parentheses

Examples: None Test Case(s): None

Suggests to increase the -fmax-pmcheck-models limit for the pattern match checker.

Triggered by: DsMaxPmCheckModelsReached

Test case(s): pmcheckshould_compileTooManyDeltas pmcheckshould_compileTooManyDeltas pmcheckshould_compileT11822

Suggests adding a type signature, typically to resolve ambiguity or help GHC inferring types.

Suggests to explicitly discard the result of a monadic action by binding the result to the '_' wilcard.

Example: main = do _ <- getCurrentTime

Suggests adding an identifier to the export list of a signature.

Suggests increasing the limit for the number of iterations in the simplifier.

Suggests to explicitly import Type from the Kind module, because using "*" to mean Type relies on the StarIsType extension, which will become deprecated in the future.

Triggered by: PsWarnStarIsType Example: None Test case(s): wcompat-warnings/WCompatWarningsOn.hs

Suggests placing the qualified keyword after the module name.

Triggered by: PsWarnImportPreQualified Example: None Test case(s): module/mod184.hs

Suggests using TemplateHaskell quotation syntax.

Triggered by: PsErrEmptyDoubleQuotes only if TemplateHaskell is enabled. Example: None Test case(s): parsershould_failT13450TH.hs

Suggests alternative roles in case we found an illegal one.

Triggered by: PsErrIllegalRoleName Example: None Test case(s): rolesshould_failRoles7.hs

Suggests qualifying the * operator in modules where StarIsType is enabled.

Triggered by: PsWarnStarBinder Test case(s): warningsshould_compileStarBinder.hs

Suggests that a type signature should have form variable :: type in order to be accepted by GHC.

Triggered by: PsErrInvalidTypeSignature Test case(s): parsershould_failT3811

Suggests to move an orphan instance or to newtype-wrap it.

Triggered by: TcRnOrphanInstance Test cases(s): warningsshould_compileT9178 typecheckshould_compileT4912

Suggests to use a standalone deriving declaration when GHC can't derive a typeclass instance in a trivial way.

Triggered by: DerivBadErrConstructor Test cases(s): typecheckshould_failtcfail086

Suggests the user to fill in the wildcard constraint to disambiguate which constraint that is.

Example: deriving instance _ => Eq (Foo f a)

Triggered by: DerivBadErrConstructor Test cases(s): partial-sigsshould_failT13324_fail2

Suggests to use an identifier other than forall Triggered by: TcRnForallIdentifier

Suggests to use the appropriate Template Haskell tick: a single tick for a term-level NameSpace, or a double tick for a type-level NameSpace.

Triggered by: TcRnIncorrectNameSpace.

Suggests enabling -ddump-splices to help debug an issue when a Name is not in scope or is used in multiple different namespaces (e.g. both as a data constructor and a type constructor).

Concomitant with NoExactName or SameName errors, see e.g. "GHC.Rename.Env.lookupExactOcc_either". Test cases: T5971, T7241, T13937.

Suggests adding a tick to refer to something which has been promoted to the type level, e.g. a data constructor.

Test cases: T9778, T19984.

Something is split off from its corresponding declaration. For example, a datatype is given a role declaration in a different module.

Test cases: T495, T8485, T2713, T5533.

Suggest a similar name that the user might have meant, e.g. suggest traverse when the user has written travrese.

Test case: mod73.

Remind the user that the field selector has been suppressed because of -XNoFieldSelectors.

Test cases: NFSSuppressed, records-nofieldselectors.

Suggest importing from a module, removing a hiding clause, or explain to the user that we couldn't find a module with the given ModuleName.

Test cases: mod28, mod36, mod87, mod114, ...

Suggest importing a data constructor to bring it into scope Triggered by: TcRnTypeCannotBeMarshaled

Test cases: ccfail004

Suggest using pattern matching syntax for a non-bidirectional pattern synonym

Test cases: patsynshould_failrecord-exquant typecheckshould_failT3176

Suggest tips for making a definition visible for the purpose of writing a SPECIALISE pragma for it in a different module.

Test cases: none

An unknown binding (i.e. too complicated to turn into a Name)

Suggest to enable the input extension. This is the hint that GHC emits if this is not a "known" fix, i.e. this is GHC giving its best guess on what extension might be necessary to make a certain program compile. For example, GHC might suggests to enable BlockArguments when the user simply formatted incorrectly the input program, so GHC here is trying to be as helpful as possible. If the input SDoc is not empty, it will contain some extra information about the why the extension is required, but it's totally irrelevant/redundant for IDEs and other tools.

Suggest to enable the input extensions. The list is to be intended as disjunctive i.e. the user is suggested to enable any of the extensions listed. If the input SDoc is not empty, it will contain some extra information about the why the extensions are required, but it's totally irrelevant/redundant for IDEs and other tools.

Suggest to enable the input extensions. The list is to be intended as conjunctive i.e. the user is suggested to enable all the extensions listed. If the input SDoc is not empty, it will contain some extra information about the why the extensions are required, but it's totally irrelevant/redundant for IDEs and other tools.

Suggest to enable the input extension in order to fix a certain problem. This is the suggestion that GHC emits when is more-or-less clear "what's going on". For example, if both DeriveAnyClass and GeneralizedNewtypeDeriving are turned on, the right thing to do is to enabled DerivingStrategies, so in contrast to SuggestSingleExtension GHC will be a bit more "imperative" (i.e. "Use X Y Z in order to ... "). If the input SDoc is not empty, it will contain some extra information about the why the extensions are required, but it's totally irrelevant/redundant for IDEs and other tools.