November 2015 - ghc-tickets

[GHC] #10853: Refine addTopDecls
by GHC 28 Mar '16

28 Mar '16

#10853: Refine addTopDecls -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.12.1 Component: Template | Version: 7.10.2 Haskell | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- This suggested refactoring/documentation task is spurred on by #10486. Readers do not have to read that ticket to understand this task. Currently, `addTopDecls` is restricted in what it can add. In an email exchange with Geoff Mainland (the original implementor of `addTopDecls`), it was unclear exactly why this restriction is in place. So, I propose lifting the restriction. We do have to be careful about ordering. As it is currently implemented, here is what happens: {{{ -- region A, including everything up to but not including a top-level splice $( {- region B, the contents of a top-level splice -} ) -- region C, everything after the splice }}} In addition to those regions, we also have {{{ -- region D, consisting of all declarations added with `addTopDecls`. }}} By reading the code (`TcRnDriver.tc_rn_src_decls`), it seems that the expected behavior should be this: 1. Rename region A. Anything defined in other regions is not in scope. 2. Rename region D, with declarations from region A in scope. 3. Type-check regions A and D together. 4. When running TH code in region B, all declarations from A and D are in scope and can be accessed via `reify` and friends. 5. Combine regions B and C together, and recur to step 1 (where the combined B and C is now a new region A). This all means that A and D are mutually recursive w.r.t. the type-checker but not the renamer. This is a bit peculiar, but not terrible. For example, this means that there could be a function whose definition is in A and type signature in D. Proper mutual recursion would be hard (impossible?) without better renaming support. In any case, I don't see a reason to restrict what's allowed in `addTopDecls`. So, at a minimum, I propose: * Remove restrictions in `addTopDecls` * Document `addTopDecls` generally * Document the above behavior, detailing the interaction between these regions Perhaps better, we could do: * Remove restrictions in `addTopDecls` * Rejigger the implementation of `addTopDecls` to lump region D in with B and C. This looks quite straightforward, and it makes for a simpler story about mutual recursion. This is a potentially-breaking change, if existing code has, say, a function defined in A with its type signature in D. * Document the new, simpler behavior Regardless of which proposal we go with, some testing is in order, to make sure that this is all correct. I've done no testing in formulating this ticket, just reading code. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10853> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #10338: GHC Forgets Constraints
by GHC 26 Mar '16

26 Mar '16

#10338: GHC Forgets Constraints -------------------------------------+------------------------------------- Reporter: crockeea | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.3 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- This is possibly fixed by #10195, but I don't have a convenient means of testing it. At any rate, this testcase is considerably simpler than the one in #10195. {{{#!hs {-# LANGUAGE TypeFamilies, FlexibleContexts, GADTs, MultiParamTypeClasses #-} type family F r class (Functor t) => T t r where fromScalar :: r -> t r data Foo t r where Foo :: t (F r) -> Foo t r Scalar :: r -> Foo t r toF :: r -> F r toF = undefined convert :: (T t (F r)) => Foo t r -> Foo t r convert (Scalar c) = let fromScalar' = fromScalar in Foo $ fromScalar' $ toF c }}} This code compiles with GHC 7.8.4. When I add a generic instance for `T` (which requires `FlexibleInstances`): `instance (Functor t, Num r) => T t r` GHC complains: {{{#!hs Could not deduce (Num (F r)) arising from a use of ‘fromScalar’ from the context (T t (F r)) bound by the type signature for convert :: (T t (F r)) => Foo t r -> Foo t r at Main.hs:(17,12)-(18,23) In the expression: fromScalar In an equation for ‘fromScalar'’: fromScalar' = fromScalar In the expression: let fromScalar' = fromScalar in Foo $ fromScalar' $ toF c }}} Of course the problem can be fixed by adding a type signature to `fromScalar` and adding `ScopedTypeVariables`: {{{#!hs convert :: forall t r . (T t (F r)) => Foo t r -> Foo t r convert (Scalar c) = let fromScalar' = fromScalar :: F r -> t (F r) in Foo $ fromScalar' $ toF c }}} Like #10195, this is triggered when a generic instance is in scope. The (main) problem of course is that GHC tries to match the generic instance instead of using the `T (F r)` constraint supplied by `convert`. A secondary issue is that I think the example should have the same behavior pre- and post-instance, i.e either both should compile or both should not compile. I'm not sure if a monomorphism restriction is actually being triggered here or if that's just a red herring. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10338> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #11145: Template Haskell quotes of data instance GADTs is totally broken
by GHC 24 Mar '16

24 Mar '16

#11145: Template Haskell quotes of data instance GADTs is totally broken -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Template | Version: 7.10.2 Haskell | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- When I say {{{ data family Fuggle x y [d| data instance Fuggle Int (Maybe (a,b)) where MkFuggle :: Fuggle Int (Maybe Bool) |] }}} I get (with `-ddump-splices`) {{{ [d| data instance Fuggle Int (Maybe (a_ajz6, b_ajz7)) where MkFuggle_ajz5 :: Fuggle Int (Maybe Bool) |] ======> data instance Fuggle Int (Maybe (a_ajAf, b_ajAg)) = (b_ajAg ~ Maybe Bool, a_ajAf ~ Int) => MkFuggle_ajAe }}} But that's not at all what I said! Admittedly, what I said was garbage, but it should fail outright. Is this addressed by Phab:D1465? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11145> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #9405: Ticky results should be displayed even during interrupts (Ctrl-c)
by GHC 24 Mar '16

24 Mar '16

#9405: Ticky results should be displayed even during interrupts (Ctrl-c) -------------------------------------+------------------------------------- Reporter: thoughtpolice | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.10.1 Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- After talking with Simon & Simon today, we discussed ticky profiling, and it dawned on us that ticky probably isn't reported when the program is interrupted via the user. But it should be! The motivating case is when you're dealing with an infinite loop and you want to find the binding that has a high entry count (i.e. it's the looping body). But the only way to move forward is to interrupt it, which doesn't spit out the reports. Fixing this would make ticky much more useful in such cases. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9405> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #10320: -ddump-to-file should create empty dump files when there was nothing to dump
by GHC 24 Mar '16

24 Mar '16

#10320: -ddump-to-file should create empty dump files when there was nothing to dump -------------------------------------+------------------------------------- Reporter: rwbarton | Owner: Type: bug | Status: new Priority: low | Milestone: Component: Compiler | Version: 7.10.1 Keywords: newcomer | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- For example, I ran with -ddump-rule-firings -ddump-to-file when there were no rule firings. GHC should have created an empty foo.dump-rule-firings file. (Instead it left the existing foo.dump-rule-firings from a previous run intact, which could have been misleading if I didn't already know that the most recent run had no rule firings.) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10320> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 19

[GHC] #10162: Add unicode syntax for banana brackets
by GHC 24 Mar '16

24 Mar '16

#10162: Add unicode syntax for banana brackets -------------------------------------+------------------------------------- Reporter: zardoz | Owner: Type: feature | Status: new request | Milestone: Priority: low | Version: 7.8.4 Component: Compiler | Operating System: Unknown/Multiple Keywords: | Type of failure: None/Unknown Architecture: | Blocked By: Unknown/Multiple | Related Tickets: Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- While there is a “pretty” unicode alternative available for “-<“ within arrow notation (⤙), banana brackets can only be written as “(|” and “|)”, which is a bit of an eyesore when they are used in the middle of the fancy arrow glyphs. Maybe it would be convenient to add the following unicode alternatives, which to the best of my knowledge are mapped by default within the XCompose definitions shipped with XOrg. ⦆ right white parenthesis (U+2986) ⦅ left white parenthesis (U+2985) Alternatively, these would also correspond well with the plain ASCII syntax: ⦇ z notation left image bracket (U+2987) ⦈ z notation right image bracket (U+2988) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10162> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 9

[GHC] #10075: Muddling Constraint and * means that Core is inconsistent
by GHC 17 Mar '16

17 Mar '16

#10075: Muddling Constraint and * means that Core is inconsistent -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.4 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- I can say this: {{{ type family F (a :: k) type instance F (a :: Constraint) = Int type instance F (a :: *) = Bool }}} This compiles to the following Core: {{{ F :: forall (k :: BOX). k -> * axF1 :: [a :: Constraint].F Constraint a ~ Int axF2 :: [a :: *]. F * a ~ Bool }}} I can then build the following: {{{ sym axF1[Any Constraint] ; axF2[Any *] :: Int ~ Bool }}} The transitivity is sound because, in Core, `Constraint` and `*` are indistinguishable. I don't think there's a way to actually cause a bug in a running program due to this problem, but we should probably patch it up. Possible resolution: Have `coreView` change `Constraint` to `*`. Possible resolution once we have `* :: *`: Make `Constraint` and `*` representationally equal. In other words, it would be as if `Constraint` were a newtype around `*`. I haven't thought out the ramifications of this decision, but there's something nice about it. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10075> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #10303: Make it easier to print stack traces when debugging GHC itself
by GHC 17 Mar '16

17 Mar '16

#10303: Make it easier to print stack traces when debugging GHC itself -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: task | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- As suggested in [http://mail.haskell.org/pipermail/ghc- devs/2015-February/008271.html the mailing list], GHC should have better support for printing its own stack trace when profiling is enabled. This ticket is to track progress toward this goal. There are a few moving parts: * Just having `pprTrace` use `traceStack` is a bad idea, because a profiled compiler will spew lots and lots and lots of output. Instead, I propose a new `pprTraceStack :: String -> SDoc -> a -> a`. * Currently, `ASSERT` will print a stack trace, but it's sometimes the wrong one. `ASSERT` throws an error; if this error is caught and re- thrown, the stack trace is at the re-throw. Instead, it should be at the origin. * Is it worth adding a new mode in `build.mk` that enables profiling, `DEBUG`, and `-fprof-auto`, just to get good stack traces? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10303> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #9803: Poor error message for unbound variable in pattern synonym
by GHC 17 Mar '16

17 Mar '16

#9803: Poor error message for unbound variable in pattern synonym -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.9 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Easy (less than 1 | Type of failure: hour) | None/Unknown Blocked By: | Test Case: Related Tickets: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- When I say {{{ pattern OneElt = [x] }}} I get {{{ Right-hand side of bidirectional pattern synonym cannot be used as an expression [x] }}} I have two complaints with this error message: 1. It's very unclear what's actually wrong with the pattern -- `[x]` looks like a perfectly fine expression, until I realize that `x` is unbound. 2. The herald of the error is more than 80 characters (if you count the indentation) and wraps on my 80-character terminal. GHC tends to avoid this elsewhere, and so the wrapping is suboptimal. I would easily say that issue (1) is more important than issue (2). Perhaps this would be better: {{{ Variable `x' is unbound in the bidirectional pattern synonym `OneElt': [x] }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9803> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #8132: Warning for Typeable instances misplaced
by GHC 16 Mar '16

16 Mar '16

#8132: Warning for Typeable instances misplaced -------------------------+------------------------------------------------- Reporter: | Owner: scottgw | Status: new Type: bug | Milestone: Priority: | Version: 7.7 normal | Operating System: Unknown/Multiple Component: | Type of failure: Incorrect warning at Compiler | compile-time Keywords: | Test Case: Architecture: | Blocking: Unknown/Multiple | Difficulty: | Unknown | Blocked By: | Related Tickets: | -------------------------+------------------------------------------------- There is currently a warning for hand-rolled instances of Typeable. However, this leads to some pretty surprising error messages: because the instance is ignored any code that relies on that instance in the same module just fails, which appears pretty weird to the user. {{{ #!haskell import Data.Typeable data K = K instance Typeable K where typeRep _ = undefined ex :: TypeRep ex = typeRep (undefined :: Proxy K) }}} Gives as an error: {{{ No instance for (Typeable * K) arising from a use of ‛typeRep’ In the expression: typeRep (undefined :: Proxy K) In an equation for ‛ex’: ex = typeRep (undefined :: Proxy K) }}} Which is of course surprising because the user just defined the instance. After commenting out the code that uses the instance, one gets: {{{ TypEx.hs:1:1: Warning: Typeable instances can only be derived; ignoring the following instance: instance Typeable * K -- Defined at TypEx.hs:5:10 }}} This information should clearly be somehow presented before or with the previous error if the user is to have any way to figure out what is going on. I suppose this could be fixed by: 1. Turning the Typeable instance warning into an error 2. Checking the missing instance in the existing error for Typeable and suggest that the user should look for hand written instances as a source of the problem. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8132> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 17