February 2018 - ghc-tickets

[GHC] #13624: loadObj() does not respect alignment
by GHC 04 Feb '19

04 Feb '19

#13624: loadObj() does not respect alignment -------------------------------------+------------------------------------- Reporter: tmcdonell | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Runtime | Version: 8.0.1 System (Linker) | 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 perhaps known, but I'll write it down here in case somebody else runs into this problem as well. Since `loadObj()` just `mmap()`s the entire object file and decodes it ''in place'', it does not respect the alignment requirements specified in the section headers. This is problematic for instructions which require alignment, e.g. SSE, AVX. The attached `map.ll` program is `map (+1)` over an array of floating point numbers. In particular, the core loop is 8-way SIMD vectorised x 4-way unrolled, for 32-elements per loop iteration. A tail loop handles any remainder one-at-a-time. You can compile it using `llc -filetype=obj -mcpu=native map.ll`. For a CPU with AVX instructions (sandy bridge or later) you should get the following: {{{ $ objdump -d map.o Disassembly of section .text: 0000000000000000 <map>: 0: 49 89 f3 mov %rsi,%r11 3: 49 29 fb sub %rdi,%r11 6: 0f 8e f9 00 00 00 jle 105 <map+0x105> c: 49 83 fb 20 cmp $0x20,%r11 10: 0f 82 bd 00 00 00 jb d3 <map+0xd3> 16: 4d 89 da mov %r11,%r10 19: 49 83 e2 e0 and $0xffffffffffffffe0,%r10 1d: 4d 89 d9 mov %r11,%r9 20: 49 83 e1 e0 and $0xffffffffffffffe0,%r9 24: 0f 84 a9 00 00 00 je d3 <map+0xd3> 2a: 49 01 fa add %rdi,%r10 2d: 48 8d 44 ba 60 lea 0x60(%rdx,%rdi,4),%rax 32: 49 8d 7c b8 60 lea 0x60(%r8,%rdi,4),%rdi 37: c5 fc 28 05 00 00 00 vmovaps 0x0(%rip),%ymm0 # 3f <map+0x3f> 3e: 00 3f: 4c 89 c9 mov %r9,%rcx 42: 66 66 66 66 66 2e 0f data16 data16 data16 data16 nopw %cs:0x0(%rax,%rax,1) 49: 1f 84 00 00 00 00 00 50: c5 f8 10 4f a0 vmovups -0x60(%rdi),%xmm1 55: c5 f8 10 57 c0 vmovups -0x40(%rdi),%xmm2 5a: c5 f8 10 5f e0 vmovups -0x20(%rdi),%xmm3 5f: c5 f8 10 27 vmovups (%rdi),%xmm4 63: c4 e3 75 18 4f b0 01 vinsertf128 $0x1,-0x50(%rdi),%ymm1,%ymm1 6a: c4 e3 6d 18 57 d0 01 vinsertf128 $0x1,-0x30(%rdi),%ymm2,%ymm2 71: c4 e3 65 18 5f f0 01 vinsertf128 $0x1,-0x10(%rdi),%ymm3,%ymm3 78: c4 e3 5d 18 67 10 01 vinsertf128 $0x1,0x10(%rdi),%ymm4,%ymm4 7f: c5 f4 58 c8 vaddps %ymm0,%ymm1,%ymm1 83: c5 ec 58 d0 vaddps %ymm0,%ymm2,%ymm2 87: c5 e4 58 d8 vaddps %ymm0,%ymm3,%ymm3 8b: c5 dc 58 e0 vaddps %ymm0,%ymm4,%ymm4 8f: c4 e3 7d 19 48 b0 01 vextractf128 $0x1,%ymm1,-0x50(%rax) 96: c5 f8 11 48 a0 vmovups %xmm1,-0x60(%rax) 9b: c4 e3 7d 19 50 d0 01 vextractf128 $0x1,%ymm2,-0x30(%rax) a2: c5 f8 11 50 c0 vmovups %xmm2,-0x40(%rax) a7: c4 e3 7d 19 58 f0 01 vextractf128 $0x1,%ymm3,-0x10(%rax) ae: c5 f8 11 58 e0 vmovups %xmm3,-0x20(%rax) b3: c4 e3 7d 19 60 10 01 vextractf128 $0x1,%ymm4,0x10(%rax) ba: c5 f8 11 20 vmovups %xmm4,(%rax) be: 48 83 e8 80 sub $0xffffffffffffff80,%rax c2: 48 83 ef 80 sub $0xffffffffffffff80,%rdi c6: 48 83 c1 e0 add $0xffffffffffffffe0,%rcx ca: 75 84 jne 50 <map+0x50> cc: 4d 39 cb cmp %r9,%r11 cf: 75 05 jne d6 <map+0xd6> d1: eb 32 jmp 105 <map+0x105> d3: 49 89 fa mov %rdi,%r10 d6: 4c 29 d6 sub %r10,%rsi d9: 4a 8d 04 92 lea (%rdx,%r10,4),%rax dd: 4b 8d 0c 90 lea (%r8,%r10,4),%rcx e1: c5 fa 10 05 00 00 00 vmovss 0x0(%rip),%xmm0 # e9 <map+0xe9> e8: 00 e9: 0f 1f 80 00 00 00 00 nopl 0x0(%rax) f0: c5 fa 58 09 vaddss (%rcx),%xmm0,%xmm1 f4: c5 fa 11 08 vmovss %xmm1,(%rax) f8: 48 83 c0 04 add $0x4,%rax fc: 48 83 c1 04 add $0x4,%rcx 100: 48 ff ce dec %rsi 103: 75 eb jne f0 <map+0xf0> 105: c5 f8 77 vzeroupper 108: c3 req }}} The attached `test.c` will load the object file and try to execute it. The `#define N` on line 7 will change the size of the array. For fewer than 32 elements this works as expected (where the input array is [0..N-1]): {{{ $ ./build.sh + llc-4.0 -filetype=obj -mcpu=native map.ll + ghc --make -no-hs-main test.c $ ./a.out array size is 31 calling function... ok 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 16.0 17.0 18.0 19.0 20.0 21.0 22.0 23.0 24.0 25.0 26.0 27.0 28.0 29.0 30.0 31.0 }}} For 32 elements or larger (i.e. entering the core loop) the program will (almost certainly) segfault. {{{ $ lldb a.out (lldb) target create "a.out" Current executable set to 'a.out' (x86_64). (lldb) run Process 7294 launched: '<snip>/a.out' (x86_64) array size is 32 calling function... Process 7294 stopped * thread #1: tid = 0xc41676, 0x000000010019f207, queue = 'com.apple.main- thread', stop reason = EXC_BAD_ACCESS (code=EXC_I386_GPFLT) frame #0: 0x000000010019f207 -> 0x10019f207: vmovaps 0xe1(%rip), %ymm0 0x10019f20f: movq %r9, %rcx 0x10019f212: nopw %cs:(%rax,%rax) 0x10019f220: vmovups -0x60(%rdi), %xmm1 }}} The `VMOVAPS` instruction requires the source address to be 32-byte aligned. It is attempting to load 8 floats from one of the const sections (the ones for the +1), but since the section was not loaded at the required alignment, fails. I've tested this on x86_64 macOS (Mach-O) and ubuntu (ELF). I don't have any other systems to test on. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13624> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 21

[GHC] #10857: "ghci -XMonomorphismRestriction" doesn't turn on the monomorphism restriction
by GHC 04 Feb '19

04 Feb '19

#10857: "ghci -XMonomorphismRestriction" doesn't turn on the monomorphism restriction -------------------------------------+------------------------------------- Reporter: rwbarton | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: GHCi | Version: 7.10.1 Keywords: newcomer | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Revisions: | -------------------------------------+------------------------------------- Compare {{{ rwbarton@morphism:/tmp$ ghci-7.10.1 -XMonomorphismRestriction GHCi, version 7.10.1: http://www.haskell.org/ghc/ :? for help Prelude> let a = (+) Prelude> :t a a :: Num a => a -> a -> a }}} with {{{ rwbarton@morphism:/tmp$ ghci-7.10.1 GHCi, version 7.10.1: http://www.haskell.org/ghc/ :? for help Prelude> :set -XMonomorphismRestriction Prelude> let a = (+) Prelude> :t a a :: Integer -> Integer -> Integer }}} Confusing! This also occurred in 7.8. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10857> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 18

[GHC] #9693: Reloading GHCi with Template Haskell names can panic GHC
by GHC 03 Feb '19

03 Feb '19

#9693: Reloading GHCi with Template Haskell names can panic GHC -------------------------------------+------------------------------------- Reporter: maxs | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.3 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: GHC Blocked By: | rejects valid program Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- First, load the following program into GHCi. '''Fun.hs''' {{{#!hs module Fun where import Language.Haskell.TH stuff = do -- let x = mkName "X" x <- newName "X" sequence $ [dataD (return []) x [] [ normalC x [] ] []] }}} '''thbug.hs''' {{{#!hs {-# LANGUAGE TemplateHaskell #-} import Fun stuff }}} Then comment out the newName, un comment the mkName and reload GHCi. {{{ ghc: panic! (the 'impossible' happened) (GHC version 7.8.3 for x86_64-apple-darwin): kcLookupKind APromotionErr RecDataConPE Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug }}} It is important to note this is a valid program, if you then close GHCi and start it again with the modified file, it will load correctly. Reloading after attempting to compile the newName version will cause a panic. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9693> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #11084: Some type families don't reduce with :kind!
by GHC 03 Feb '19

03 Feb '19

#11084: Some type families don't reduce with :kind! -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 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: -------------------------------------+------------------------------------- Sadly, I'm unable to reproduce this without `singletons`. :( Observe: {{{ > ghci GHCi, version 7.10.2: http://www.haskell.org/ghc/ :? for help Prelude> import Data.Promotion.Prelude -- singletons-2.0.1 Prelude Data.Promotion.Prelude> :kind! Lookup 'True '[ '(True, True) ] Lookup 'True '[ '(True, True) ] :: Maybe Bool = Data.Singletons.Prelude.List.Case_1628035408 'True 'True 'True '[] (Data.Singletons.Apply (Data.Singletons.Apply (:==$) 'True) 'True) Prelude Data.Promotion.Prelude> :kind! Lookup 'True '[ '(True, True) ] Lookup 'True '[ '(True, True) ] :: Maybe Bool = JustSym1 'True }}} Note that the exact same command run twice yields different results. I tried peeking at `-ddump-tc-trace` but got nothing of use. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11084> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 2

[GHC] #12001: RFC: Add pattern synonyms to base
by GHC 02 Feb '19

02 Feb '19

#12001: RFC: Add pattern synonyms to base -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: | Version: 7.10.3 libraries/base | 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: -------------------------------------+------------------------------------- Since we have pattern synonyms it's worth considering if some belong in base: [https://hackage.haskell.org/package/lens-4.14/docs/Data-Complex-Lens.html Data.Complex.Lens] contains patterns that could be defined in base, here are some more suggestions: === Data.Array === {{{#!hs pattern ArrayIx :: Ix i => (i, i) -> [(i, e)] -> Array i e pattern ArrayIx low'high xs <- ((\arr -> (bounds arr, assocs arr)) -> (low'high, xs)) where ArrayIx low'high xs = array low'high xs }}} === Data.Bits === {{{#!hs pattern ZeroBits :: (Eq a, Bits a) => a pattern ZeroBits <- ((== zeroBits) -> True) where ZeroBits = zeroBits pattern BitSize :: Bits a => Int -> a pattern BitSize n <- (bitSizeMaybe -> Just n) pattern Signed :: Bits a => a pattern Signed <- (isSigned -> True) pattern Unsigned :: Bits a => a pattern Unsigned <- (isSigned -> False) pattern PopCount :: Bits a => Int -> a pattern PopCount n <- (popCount -> n) }}} === Data.Char === {{{#!hs pattern ControlChar :: Char pattern ControlChar <- (isControl -> True) pattern SpaceChar :: Char pattern SpaceChar <- (isSpace -> True) }}} === Data.Complex === {{{#!hs pattern Conjugate :: Num a => Complex a -> Complex a pattern Conjugate a <- (conjugate -> a) where Conjugate a = conjugate a pattern Polar :: RealFloat a => a -> a -> Complex a pattern Polar m theta <- (polar -> (m, theta)) where Polar m theta = mkPolar m theta -- See https://github.com/ekmett/lens/issues/653 pattern Real :: Num a => a -> Complex a pattern Real r <- r :+ _ where Real r = r :+ 0 pattern Imaginary :: Num a => a -> Complex a pattern Imaginary i <- _ :+ i where Imaginary i = 0 :+ i }}} === GHC.Float === {{{#!hs pattern NegativeZero :: RealFloat a => a pattern NegativeZero <- (isNegativeZero -> True) where NegativeZero = -0 pattern Denormalized :: RealFloat a => a pattern Denormalized <- (isDenormalized -> True) pattern NaN :: RealFloat a => a pattern NaN <- (isNaN -> True) where NaN = 0 / 0 -- How ever negative infinity is handled pattern Infinity :: RealFloat a => a pattern Infinity <- ((== 1 / 0) -> True) where Infinity = 1 / 0 }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12001> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #11004: hsc2hs does not handle single quotes properly
by GHC 01 Feb '19

01 Feb '19

#11004: hsc2hs does not handle single quotes properly -------------------------------------+------------------------------------- Reporter: sergv | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: hsc2hs | Version: 7.10.2 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: -------------------------------------+------------------------------------- The tool gets confused by quotes for promoted constructors (https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/promotion.h… #promotion-syntax). Here's an example: {{{#!hs {-# LANGUAGE DataKinds #-} {-# LANGUAGE GADTs #-} data Foo = Foo | Bar data Indexed ix where Indexed :: Indexed 'Foo main :: IO () main = print $ #size int }}} gets translated into {{{#!hs {-# LINE 1 "Test.hsc" #-} {-# LANGUAGE DataKinds #-} {-# LINE 2 "Test.hsc" #-} {-# LANGUAGE GADTs #-} data Foo = Foo | Bar data Indexed ix where Indexed :: Indexed 'Foo main :: IO () main = print $ #size int }}} Then GHC complains about `#size`: {{{ Test.hsc:10:16: parse error on input ‘#’ }}} Judging from the source of utils/hsc2hs/HSCParser.hs, the tool treats single quotes the same way as double quotes. I.e. it looks for balanced, possibly multiline, strings delimited by '. At least, it seems unnecessary to expect newlines in that string-like structure, because character literals cannot contain newlines, AFAIK. It's interesting to note that this problem looks like a bit of an edge case, since highlighting in the ticket seems to be confused as well. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11004> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #13834: Error cascade with type applications
by GHC 01 Feb '19

01 Feb '19

#13834: Error cascade with type applications -------------------------------------+------------------------------------- Reporter: mpickering | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: newcomer, | Operating System: Unknown/Multiple TypeApplications | Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Using type applications with an out of scope identifier causes an unfortunate error cascade. {{{ foo = notInScope @Bool }}} Leads to the right out of scope error but also an error about using type application when `notInScope` is not a polytype. I think the second error should be suppressed. {{{ t.hs:4:7: error: Variable not in scope: notInScope | 4 | foo = notInScope @Bool | ^^^^^^^^^^ t.hs:4:7: error: • Cannot apply expression of type ‘t1’ to a visible type argument ‘Bool’ notInScope • In the expression: notInScope @Bool In an equation for ‘foo’: foo = notInScope @Bool | 4 | foo = notInScope @Bool | ^^^^^^^^^^^^^^^^ }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13834> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #11222: Teach strictness analysis about `catch`-like operations
by GHC 31 Jan '19

31 Jan '19

#11222: Teach strictness analysis about `catch`-like operations -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: feature | Status: new request | Priority: normal | Milestone: 8.2.1 Component: Compiler | Version: 7.10.3 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Runtime Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Consider the `catch#` primop, {{{#!hs catch# :: (State# RealWorld -> (# State# RealWorld, a #) ) -- ^ thing to catch exceptions from -> (b -> State# RealWorld -> (# State# RealWorld, a #) ) -- ^ exception handler -> State# RealWorld -> (# State# RealWorld, a #) }}} Semantically, this operation will always evaluate its first argument. Ideally we would indicate this in the primop's strictness signature in `primops.txt.pp`. Sadly, we can't do this at the moment due to a subtle wrinkle (discovered in #10712): Consider, {{{#!hs let r = \st -> raiseIO# blah st in catch (\st -> ...(r st)..) handler st }}} If we give the first argument of catch a strict signature, we'll get a demand `C(S)` for `r`; that is, `r` is definitely called with one argument, which indeed it is. The trouble comes when we feed `C(S)` into `r`'s RHS as the demand of the body as this will lead us to conclude that the whole `let` will diverge; clearly this isn't right. As Simon noted in ticket:10712#comment:4, > There's something very special about catch: it turns divergence into non-divergence. In order to apply a proper strictness signature to `catch`-like operations we would need to teach the strictness analyzer about this property. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11222> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 12

[GHC] #12092: Wrong error message with TypeApplications
by GHC 31 Jan '19

31 Jan '19

#12092: Wrong error message with TypeApplications -------------------------------------+------------------------------------- Reporter: kwf | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 (Type checker) | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Incorrect Unknown/Multiple | warning at compile-time Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- When I write the following: {{{#!hs {-# language TypeApplications #-} module Bug where huh = spam @Int }}} I expect to get the error: {{{ • Variable not in scope: spam • Perhaps you meant ‘span’ (imported from Prelude) }}} But instead, I get the error: {{{ • Cannot apply expression of type ‘t0’ to a visible type argument ‘Int’ • In the expression: spam @Int In an equation for ‘huh’: huh = spam @Int }}} While it is //technically// true that I cannot apply an out-of-scope identifier to a visible type argument, this is probably the less useful error to receive. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12092> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #12758: Bring sanity to our performance testsuite
by GHC 28 Jan '19

28 Jan '19

#12758: Bring sanity to our performance testsuite -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: task | Status: new Priority: normal | Milestone: 8.2.1 Component: Test Suite | 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: -------------------------------------+------------------------------------- The GHC testsuite's performance tests tends to be a constant source of busy-work as it requires that contributors manually bump performance numbers and propagate these changes across platforms. Moreover, the testsuite is poor at noticing performance regressions due to false positive failures (due to spurious environmental differences) and false negatives (due to the rather generous acceptance windows that many tests have). Joachim and I briefly discussed this at Hac Phi and came up with the following proposal: * We rip expected performance numbers out of the `.T` files * -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12758> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 20