lpsmith / bytestring-builder Goto Github PK

View Code? Open in Web Editor NEW

6.0 6.0 4.0 70 KB

The new bytestring builder, packaged outside of GHC

License: Other

Haskell 95.59% C 3.25% Shell 1.16%

bytestring-builder's People

Contributors

Stargazers

Watchers

Forkers

borsboom snoyberg hvr phadej

bytestring-builder's Issues

Ambiguous module name if bytestring 0.10.2.* installed

I believe this was introduced with cc5b229. Prior to that change, you couldn't install bytestring-builder with bytestring 0.10.2.*, so this would never come up.

I imagine this case is rare, but I have seen it in the wild, and now that blaze-builder-0.4 uses bytestring Builders I think rare cases are going to come up more often.

Reproduction

bytestring-builder-test.cabal:

name:                bytestring-builder-test
version:             0.1.0.0
build-type:          Simple
cabal-version:       >=1.10
executable bytestring-builder-test
  main-is:             main.hs
  build-depends:       base
                     , bytestring == 0.10.2.*
                     , bytestring-builder
  default-language:    Haskell2010

main.hs:

import Data.ByteString.Builder.Extra
main = return ()

Build output:

main.hs:1:8:
    Ambiguous module name `Data.ByteString.Builder.Extra':
      it was found in multiple packages:
      bytestring-0.10.2.0 bytestring-builder-0.10.4.1.1

Compilation failure with ghc 7.4.2

Not sure if this still matters (old ghc after all), but on ghc 7.4.2 building fails with:

src/Data/ByteString/Builder/Prim.hs:553:13:
    Illegal bang-pattern (use -XBangPatterns):
    ! br'

Bizarrely, if we compile with --builddir=foo compilation succeeds! I've opened a ghc ticket about this at https://ghc.haskell.org/trac/ghc/ticket/12655 , but if you care about old ghc versions you might want to work-around this.

build failures with `ghc-7.0.1`

Current master fails to compile with ghc-7.0.1 with the error included at the bottom.

Looks like it's a bit nasty to fix, but it'd be nice if bytestring-builder would compile with that version of ghc.

Building bytestring-builder-0.10.8.1.0...
Preprocessing library bytestring-builder-0.10.8.1.0...
[ 9 of 13] Compiling Data.ByteString.Builder.Internal ( src/Data/ByteString/Builder/Internal.hs, dist/build/Data/ByteString/Builder/Internal.o )
ghc: panic! (the 'impossible' happened)
  (GHC version 7.0.1 for x86_64-unknown-linux):
        dsLet: unlifted
    AbsBinds
    [b{tv a2jk} [sk]]
    []
    [ip'{v a22y} [lid] <= [b{tv a2jk} [sk]] ip'{v a2jj} [lid]]
      ip'{v a22y} [lid]
        :: forall b{tv a2jk} [sk]. base:GHC.Ptr.Ptr{tc 33A} b{tv a2jk} [sk]
      [LclId]
      { {831:15-46}
        {831:15-46}
        !((ip'{v a2jj} [lid] :: base:GHC.Ptr.Ptr{tc 33A} b{tv a2jk} [sk]))
          = {831:22-46}
            {831:25-33}
            (base:GHC.Ptr.plusPtr{v rdH} [gid]) @ base:GHC.Word.Word8{tc 32U}
            @ b{tv a2jk} [sk]
              ip{v a22s} [lid] outRemaining{v a22v} [lid] }
    EvBinds{{}}
    base:GHC.Base.${v 019} [gid]
      @ (bytestring-builder-0.10.8.1.0:Data.ByteString.Builder.Internal.BuildSignal{tc r1Yb}
           a{tv a2iU} [sk])
      @ (ghc-prim:GHC.Types.IO{tc 32I}
           (bytestring-builder-0.10.8.1.0:Data.ByteString.Builder.Internal.BuildSignal{tc r1Yb}
              a{tv a2iU} [sk]))
      (base:GHC.Base.return{v 01T} [gid[ClassOp]]
         @ <nt>ghc-prim:GHC.Types.IO{tc 32I}
         $dMonad{v a2jl} [lid]
         @ (bytestring-builder-0.10.8.1.0:Data.ByteString.Builder.Internal.BuildSignal{tc r1Yb}
              a{tv a2iU} [sk]))
      (bytestring-builder-0.10.8.1.0:Data.ByteString.Builder.Internal.bufferFull{v r1YP} [lid]
         @ a{tv a2iU} [sk]
         (ghc-prim:GHC.Types.I#{(w) v 6d} [gid[DataCon]] 1)
         ope{v a22u} [lid]
         (go{v a2iW} [lid]
            (ip'{v a22y} [lid] @ base:GHC.Word.Word8{tc 32U})))

Please report this as a GHC bug:  http://www.haskell.org/ghc/reportabug

Build failure due to "unrecognized option `--disable-benchmarks'" with v0.10.4.0 on GHC-7.0.4

See https://travis-ci.org/sjakobi/path/jobs/138399070#L589.

The next build on GHC-7.0.4 succeeded because cabal chose v0.10.8.1.0 instead.

Could this be resolved by deprecating v0.10.4.0 and possibly other known-bad versions on hackage?

broken word-primitives on big endian platforms

The bug that affected bytestring sadly affects this package as well, see
haskell/bytestring@4f52849 for more details

`bytestring-builder` failures with GHC 7.2 and GHC 7.10

See https://ghc.haskell.org/~hvr/buildreports/bytestring-builder.html

src/Data/ByteString/Short/Internal.hs:139:10:
  No instance for (GHC.Generics.Generic ShortByteString)
  arising from a use of ‘Control.DeepSeq.$gdmrnf’
  In the expression: Control.DeepSeq.$gdmrnf
  In an equation for ‘rnf’: rnf = Control.DeepSeq.$gdmrnf
  In the instance declaration for ‘NFData ShortByteString’

Nothing built with ghc 7.6.3

If I try to install this package with ghc 7.6.3 on my Mac, it builds nothing. Only if I manually edit the bytestring_has_builder flag in the .cabal file to False does anything actually happen. This is preventing me from being able to resolve haskell/aeson#177 and its associated fix, haskell/aeson#178 — please help.