May 2017 - ghc-tickets

[GHC] #12606: Linux (ELF) Support for "ghc -static -shared"
by GHC 19 Jul '18

19 Jul '18

#12606: Linux (ELF) Support for "ghc -static -shared" ----------------------------------------+--------------------------------- Reporter: tmobile | Owner: Type: feature request | Status: new Priority: normal | Milestone: ⊥ Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Linux Architecture: Unknown/Multiple | Type of failure: None/Unknown Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: ----------------------------------------+--------------------------------- On some platforms it is possible to create a stand-alone shared library out of some object `Foo.o` including foreign exports with something like: `ghc -static -shared Foo.o ...` The resulting library will have `Foo`'s dependencies including `base` as well as the RTS statically linked in. On BSDs this is possible because generating position independent code is the default there, and on Windows this is possible because the dynamic loader uses some magic that doesn't require position independent code. Unfortunately on Linux this is not possible because `libHSbase`, `libHSrts`, etc. are not built with `-fPIC` unless one specifically asks for it when building GHC. As far as I know this is the only way to get `-static -shared` to work on this platform. There are use cases for compiling Haskell code into a single shared object that does not have additional runtime dependencies. Suppose one wishes to write a plugin for a program in Haskell. It would be nice to provide the plugin as a single shared object, rather than requiring GHC's libraries to be present at runtime as well. An easy solution might be adding a switch to the configure script or `build.mk` to make it easier to build GHC+libs with `-fPIC` on Linux. Ideally it would be possible to build and install both PIC and non-PIC libraries and have GHC choose the correct one at link time based on the flags and target platform, since PIC code carries a small space/performance overhead and isn't necessary on Linux in most cases. I don't know my way around GHC(outside of the GHC API) or the build system that well yet, but if adding extra build system flags and GHC link-time flag handling code to make this work is an acceptable solution I'm happy to spend the time to get this working. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12606> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 1

[GHC] #13194: Concurrent modification of package.cache is not safe
by GHC 18 Jul '18

18 Jul '18

#13194: Concurrent modification of package.cache is not safe -------------------------------------+------------------------------------- Reporter: arybczak | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: ghc-pkg | Version: 8.0.1 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: -------------------------------------+------------------------------------- There are a couple of different issues here. 1. On Linux, issuing `ghc-pkg register` for multiple packages in parallel might result in lost updates to package database because of how `registerPackage` function works - it reads existing package databases, picks the one to modify, then checks that package info for the package to register is fine and replaces package database with what was read in the beginning + new package info. Therefore, if updates interleave, it might happen that process1 reads the database, then process2 updates it while process1 still has the old version and uses it for its update later, so update made by process2 is lost. 2. On Windows, update to package database might fail - the issue is that GHC attempts to update it using rename trick, which fails whenever any other process has file to be replaced open for reading. Combine that with the fact that GHC reads package database when compiling packages and you get problems in both Stack (https://github.com/commercialhaskell/stack/issues/2617) and Cabal (https://github.com/haskell/cabal/issues/4005) BTW, rename trick (used for atomic database updates) not only doesn't work on Windows, it's also not atomic e.g. on NFS (https://stackoverflow.com/questions/41362016/rename-atomicity-and-nfs) The solution to both problems is to use OS specific features to lock database file (in shared mode when reading and in exclusive mode when writing). This can be done on Windows using LockFileEx. Unfortunately for POSIX things are a bit more complicated. There are two ways to lock a file on Linux: 1. Using fcntl(F_SET_LK) (POSIX API) 2. Using flock (BSD API) However, fcntl locks have a serious limitation: The record locks described above are associated with the process (unlike the open file description locks described below). This has some unfortunate consequences: * If a process closes any file descriptor referring to a file, then all of the process's locks on that file are released, regardless of the file descriptor(s) on which the locks were obtained. This is bad: it means that a process can lose its locks on a file such as /etc/passwd or /etc/mtab when for some reason a library function decides to open, read, and close the same file. * The threads in a process share locks. In other words, a multithreaded program can't use record locking to ensure that threads don't simultaneously access the same region of a file. Whereas flock is not guaranteed to work with NFS, according to https://en.wikipedia.org/wiki/File_locking#Problems: Whether and how flock locks work on network filesystems, such as NFS, is implementation dependent. On BSD systems, flock calls on a file descriptor open to a file on an NFS-mounted partition are successful no-ops. On Linux prior to 2.6.12, flock calls on NFS files would act only locally. Kernel 2.6.12 and above implement flock calls on NFS files using POSIX byte-range locks. These locks will be visible to other NFS clients that implement fcntl-style POSIX locks, but invisible to those that do not.[4] Assuming that the solution would be to go with locking the database, we would need to: 1. In `registerPackage`, lock all read databases in shared mode except for the database that will later be modified, which has to be locked in exclusive mode. The handle also would need to be kept open and passed to `changeDB` later and used for rewriting the database with updated version in `GHC.PackageDb.writePackageDb` instead of `writeFileAtomic` (which is not actually unconditionally atomic, as demonstrated above). 2. `GHC.PackageDb.decodeFromFile` would lock a file in appropriate mode and return the handle to open file if appropriate. 3. Add support for locking a file. This should be fairly easy to do in GHC.IO.Handle.FD by extending function `openFile'` with appropriate parameters and then adding wrapper function `openLockedFile` or something. We can add both blocking and non-blocking locking to make ghc-pkg show information about waiting for locked package database if appropriate. Alternatively we could add a function similar to the following: `hLock :: Handle -> LockMode -> Bool{-block-} -> IO Bool`, but that requires extracting file descriptor from Handle, which as far as I see is problematic. Is going with locking an acceptable solution here? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13194> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 15

[GHC] #13184: :show bindings broken under -fexternal-interpreter
by GHC 17 Jul '18

17 Jul '18

#13184: :show bindings broken under -fexternal-interpreter -------------------------------------+------------------------------------- Reporter: niteria | Owner: Type: bug | Status: new Priority: low | Milestone: Component: GHCi | Version: 8.0.1 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: -------------------------------------+------------------------------------- This is straight from GHC HEAD: {{{ $ ./inplace/bin/ghc-stage2 --interactive -fexternal-interpreter GHCi, version 8.1.20170123: http://www.haskell.org/ghc/ :? for help Loaded GHCi configuration from /home/bnitka/.ghc/ghci.conf package flags have changed, resetting and loading new packages... Loaded GHCi configuration from /home/bnitka/.ghci Prelude> let a = 1 Prelude> :show bindings a :: Num p => p = *** Exception: ghc-stage2: this operation requires -fno-external- interpreter Prelude> }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13184> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #8581: Add support for explicitly-bidirectional pattern synonyms
by GHC 12 Jul '18

12 Jul '18

#8581: Add support for explicitly-bidirectional pattern synonyms ------------------------------------+------------------------------------- Reporter: cactus | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: 5144 | Blocking: Related Tickets: | ------------------------------------+------------------------------------- Some patterns cannot, by themselves, be turned over into an expression, so they have to be defined as unidirectional. Maybe the most trivial example would be {{{ pattern P -> _ }}} Sometimes, however, it would be desirable to give an explicit way of turning these pattern synonyms into expressions. The PatternSynonyms wiki page has this example: {{{ import qualified Data.Sequence as Seq pattern Empty -> (Seq.viewl -> Seq.EmptyL) pattern x :< xs -> (Seq.viewl -> x Seq.:< xs) pattern xs :> x -> (Seq.viewr -> xs Seq.:> x) }}} It would make a ton of sense to be able to use this cons/snoc notation as "constructors" for `Seq`s. The proposed syntax for this {{{ pattern x :< xs -> (Seq.viewl -> x Seq.:< xs) where x :< xs = x Seq.<| xs }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8581> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 52

[GHC] #12050: Allow haddock comments on non-record types
by GHC 10 Jul '18

10 Jul '18

#12050: Allow haddock comments on non-record types -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Parser) | 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: -------------------------------------+------------------------------------- There are a variety of cases I have found myself wanting to place haddock documentation on fields of non-record data types rather often in the past. The syntax for this might look like, {{{#!hs data Foo = Foo -- ^ A foo Int -- ^ how big is the foo? String -- ^ the name of the foo }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12050> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #8822: Allow -- ^ Haddock syntax on record constructors
by GHC 10 Jul '18

10 Jul '18

#8822: Allow -- ^ Haddock syntax on record constructors -------------------------------------+------------------------------------- Reporter: Fuuzetsu | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler (Parser) | Version: 7.9 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | -------------------------------------+------------------------------------- Original ticket is on Haddock Trac [http://trac.haskell.org/haddock/ticket/144 here] which I'm opening here as it requires lexer/parser changes. The idea is that given something like {{{ -- | A tree with labeled internal nodes data Tree = Empty -- ^ Leaf | Branch -- ^ Internal node { label :: a -- ^ Node label , subtrees :: [Tree] -- ^ List of subtrees } }}} we'd like it to parse as the user would expect instead of failing. We're currently forced to do something like {{{ -- | A tree with labeled internal nodes data Tree = Empty -- ^ Leaf | -- | Internal node Branch { label :: a -- ^ Node label , subtrees :: [Tree] -- ^ List of subtrees } }}} which is quite ugly. I can't think of a reason why the first form wouldn't be allowed. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8822> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #11070: Type-level arithmetic of sized-types has weaker inference power than in 7.8
by GHC 10 Jul '18

10 Jul '18

#11070: Type-level arithmetic of sized-types has weaker inference power than in 7.8 -------------------------------------+------------------------------------- Reporter: cactus | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Given the following definitions (just a copy of the relevant bits from sized-types 0.3.5.1, to keep this example self-contained): {{{ -- Copied from sized-types 0.3.5.1 data X0 = X0 data N1 = N1 data X0_ a = X0_ Integer data X1_ a = X1_ Integer type X1 = X1_ X0 type X2 = X0_ (X1_ X0) class Size ix instance Size X0 instance Size a => Size (X1_ a) instance Size a => Size (X0_ a) type family APP0 a type instance APP0 X0 = X0 type instance APP0 N1 = X0_ N1 type instance APP0 (X1_ a) = X0_ (X1_ a) type instance APP0 (X0_ a) = X0_ (X0_ a) type family APP1 a type instance APP1 X0 = X1_ X0 type instance APP1 N1 = N1 type instance APP1 (X1_ a) = X1_ (X1_ a) type instance APP1 (X0_ a) = X1_ (X0_ a) type family SUCC a type instance SUCC X0 = X1_ X0 type instance SUCC N1 = X0 type instance SUCC (X1_ a) = APP0 (SUCC a) type instance SUCC (X0_ a) = APP1 a type family ADD a b type instance ADD a X0 = a type instance ADD X0 a = a type instance ADD X0 N1 = N1 type instance ADD N1 N1 = APP0 N1 type instance ADD N1 (X1_ b) = APP0 b type instance ADD N1 (X0_ b) = APP1 (ADD N1 b) type instance ADD (X1_ a) N1 = APP0 a type instance ADD (X0_ a) N1 = APP1 (ADD a N1) type instance ADD (X1_ a) (X1_ b) = APP0 (SUCC (ADD a b)) type instance ADD (X1_ a) (X0_ b) = APP1 (ADD a b) type instance ADD (X0_ a) (X1_ b) = APP1 (ADD a b) type instance ADD (X0_ a) (X0_ b) = APP0 (ADD a b) type family NOT a type instance NOT X0 = N1 type instance NOT N1 = X0 type instance NOT (X1_ a) = APP0 (NOT a) type instance NOT (X0_ a) = APP1 (NOT a) type SUB a b = ADD a (SUCC (NOT b)) }}} The following module typechecks with GHC 7.8.3: {{{ data B w = B (&*) :: (Size n, Size n', Size (ADD n' n), n ~ SUB (ADD n' n) n', n' ~ SUB (ADD n' n) n) => [(a, B n)] -> [(b, B n')] -> [((a, b), B (ADD n' n))] mks &* args = undefined foo :: [((), B X1)] foo = [((), B)] bar :: [(((), ()), B X2)] bar = [((), B)] &* foo }}} However, it fails with GHC 7.10.2, with {{{ /tmp/GHCBug.hs:70:7: Couldn't match type ‘ADD X1 n0’ with ‘X0_ (X1_ X0)’ The type variable ‘n0’ is ambiguous Expected type: [(((), ()), B X2)] Actual type: [(((), ()), B (ADD X1 n0))] In the expression: [((), B)] &* foo In an equation for ‘bar’: bar = [((), B)] &* foo /tmp/GHCBug.hs:70:17: Occurs check: cannot construct the infinite type: n0 ~ ADD (ADD X1 n0) N1 The type variable ‘n0’ is ambiguous Expected type: SUB (ADD X1 n0) X1 Actual type: n0 In the expression: [((), B)] &* foo In an equation for ‘bar’: bar = [((), B)] &* foo Failed, modules loaded: none. }}} The workaround/solution is to change the definition of `bar`: {{{ bar :: [(((), ()), B X2)] bar = [((), B :: B X1)] &* foo }}} This second version typechecks with GHC 7.10.2. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11070> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #13109: CUSK improvements
by GHC 10 Jul '18

10 Jul '18

#13109: CUSK improvements -------------------------------------+------------------------------------- Reporter: simonpj | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 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: -------------------------------------+------------------------------------- Inspired by looking at `RnTypes.bindLHsTyVarBndr`, and driven by #11592, Richard and I decided to make some improvements to the treatment of CUSKs. Here's a scrappy Skype dump as a memory-jogger; it is a memory-jogger, not a full description. Richard has notes too {{{ Richard Eisenberg: data F (x :: B a) data G (x :: C b) You have written a complete user-suppled kind signature, but the following variable is undetermined: a0 :: A Perhaps add a kind signature. Inferred kinds of user-written variables: b :: B a0 x :: C b SPJ claim: CUSKs should be handled INDIVIDUALLY and in isoloation, just like term-level type signatures SPJ claim: get rid of the "after the ::" side condition for data types (last bullet of HsDecls Note about CUSKs All of this should mean (SPJ claims) no need for dep_vars returns by bindHsQTyVar, bindLHsTyVarBnrds etc ToDo: clarify comments etc in bindLHStyVarBndr Related to D2914 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13109> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #11282: Injectivity annotation fails in the presence of higher-rank use of constraint family
by GHC 09 Jul '18

09 Jul '18

#11282: Injectivity annotation fails in the presence of higher-rank use of constraint family -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.11 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 {{{ {-# LANGUAGE TypeFamilies, RankNTypes #-} module Bug where import GHC.Exts type family F a = r | r -> a type family G a :: Constraint meth :: (G a => F a) -> () meth = undefined }}} I get {{{ Bug.hs:10:9: error: • Could not deduce: F a ~ F a0 from the context: G a0 bound by the type signature for: meth :: G a0 => F a0 at Bug.hs:10:9-26 NB: ‘F’ is a type function, and may not be injective The type variable ‘a0’ is ambiguous • In the ambiguity check for ‘meth’ To defer the ambiguity check to use sites, enable AllowAmbiguousTypes In the type signature: meth :: (G a => F a) -> () }}} The presence of `G a` is critical for exhibiting this bug. If I replace `G a` with `Show a` the problem disappears. This one was actually encountered in an attempt to do Useful Work, not just idle typecheckery. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11282> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #12928: Too easy to trigger CUSK condition using TH
by GHC 09 Jul '18

09 Jul '18

#12928: Too easy to trigger CUSK condition using TH -------------------------------------+------------------------------------- Reporter: int-index | Owner: Type: feature | Status: new request | Priority: high | Milestone: Component: Compiler | Version: 8.0.1 (Type checker) | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- While trying to make `singletons` class promotion work with kind- polymorphic type variables, I've encountered a problem regarding CUSKs. Consider this class: {{{ $(promoteOnly [d| class Cls a where fff :: Proxy a -> () |]) }}} The generated TH (slightly simplified by hand) looks like this: {{{ type FffSym1 (t_alvm :: Proxy a1627472612) = Fff t_alvm data FffSym0 (l_alvn :: TyFun (Proxy a1627472612) ()) type instance Apply FffSym0 l_alvn = FffSym1 l_alvn class PCls a_alve where type Fff (arg_alvl :: Proxy a_alve) :: () }}} However, it fails to compile with the following error: {{{ cusk.hs:25:1: error: You have written a *complete user-suppled kind signature*, but the following variable is undetermined: k0 :: * Perhaps add a kind signature. Inferred kinds of user-written variables: a1627472612 :: k0 l_alvn :: TyFun (Proxy a1627472612) () }}} As suggested by the error message, we need to specify the kind in its entirety (write a proper CUSK). So to sidestep the issue, we can write {{{ data FffSym0 (l_alvn :: TyFun (Proxy (a1627472612 :: k)) ()) }}} and it'll work. However, this means that the TH code generator should perform some guesswork to modify kind annotations. It sounds like a minefield — too much can go wrong, so it'd be great to avoid doing so. Looking at the module `TcHsType` in the GHC sources, I see that the error is reported by the `kcHsTyVarBndrs` function: {{{ -- If there are any meta-tvs left, the user has -- lied about having a CUSK. Error. ; let (meta_tvs, good_tvs) = partition isMetaTyVar qkvs ; when (not (null meta_tvs)) $ report_non_cusk_tvs (qkvs ++ tc_tvs) }}} However, I don't see why we can't generalize over `meta_tvs` instead of reporting a failure. I asked on IRC #ghc and ezyang suggested that it should be sound. Below I attach a self-contained example that demonstrates the issue and has no dependencies except `base`. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12928> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 14