Significant Windows failures with nightly

[ndmitchell]

Using Stack nightly with GHC 8.4.1 there seem to be quite a few Windows-only failures cropping up. In particular:

• Network having issues: https://ci.appveyor.com/project/ndmitchell/hoogle/build/1.0.318:

    Building library for network-2.6.3.4..
    [1 of 9] Compiling Network.Socket.ByteString.Internal ( Network\Socket\ByteString\Internal.hs, .stack-work\dist\893fbf7f\build\Network\Socket\ByteString\Internal.o )
    [2 of 9] Compiling Network.Socket.Types ( .stack-work\dist\893fbf7f\build\Network\Socket\Types.hs, .stack-work\dist\893fbf7f\build\Network\Socket\Types.o )
    
    C:\Users\appveyor\AppData\Local\Temp\1\stack1268\network-2.6.3.4\Network\Socket\Types.hsc:764:16: error:
        parse error on input `CALLCONV'
        |
    764 | foreign import CALLCONV unsafe "ntohs" ntohs :: Word16 -> Word16
        |                ^^^^^^^^

• Yaml having issues: https://ci.appveyor.com/project/hanjoosten/ampersand/build/build.1337-development (originally reported at Appveyor fails because of weeder ndmitchell/weeder#36 by @hanjoosten):

yaml-0.8.28: build
--  While building custom Setup.hs for package yaml-0.8.28 using:
      C:\sr\setup-exe-cache\i386-windows\Cabal-simple_Z6RU0evB_2.2.0.0_ghc-8.4.1.exe --builddir=.stack-work\dist\893fbf7f build --ghc-options " -ddump-hi -ddump-to-file"
    Process exited with code: ExitFailure (-1073741819)
    Logs have been written to: C:\stack\.stack-work\logs\yaml-0.8.28.log
    Configuring yaml-0.8.28...
    Preprocessing library for yaml-0.8.28..
    Building library for yaml-0.8.28..
    [ 1 of 10] Compiling Text.Libyaml     ( Text\Libyaml.hs, .stack-work\dist\893fbf7f\build\Text\Libyaml.o )
    [ 2 of 10] Compiling Data.Yaml.Parser ( Data\Yaml\Parser.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Parser.o )
    [ 3 of 10] Compiling Data.Yaml.Internal ( Data\Yaml\Internal.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Internal.o )
    [ 4 of 10] Compiling Data.Yaml.Include ( Data\Yaml\Include.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Include.o )
    [ 5 of 10] Compiling Data.Yaml.Builder ( Data\Yaml\Builder.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Builder.o )
    [ 6 of 10] Compiling Data.Yaml.Pretty ( Data\Yaml\Pretty.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Pretty.o )
    [ 7 of 10] Compiling Data.Yaml        ( Data\Yaml.hs, .stack-work\dist\893fbf7f\build\Data\Yaml.o )
    [ 8 of 10] Compiling Data.Yaml.TH     ( Data\Yaml\TH.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\TH.o )
    [ 9 of 10] Compiling Data.Yaml.Config ( Data\Yaml\Config.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Config.o )
    [10 of 10] Compiling Data.Yaml.Aeson  ( Data\Yaml\Aeson.hs, .stack-work\dist\893fbf7f\build\Data\Yaml\Aeson.o )

[RyanGlScott]

That network failure is simply haskell/network#313.

I have no idea yet what is going on with yaml.

[quasicomputational]

-1073741819 = unsigned 32-bit 0xC0000005, if that means anything to anyone.

[ndmitchell]

Same error code out of filepath: https://ci.appveyor.com/project/ndmitchell/filepath/build/1.0.72

[ndmitchell]

For info, this is with 32bit Windows, which might be the issue.

[DanBurton]

I am not able to reproduce the network and yaml build issues on my windows machine. However, I think I have some sort of linux compat layer installed (it's been a few years since I've used this machine for Haskell dev, so I don't recall exactly what it was) which may explain my differing results.

[edit] my windows machine is 64 bit

[RyanGlScott]

I am also unable to reproduce either the network or yaml issues using 32-bit GHC 8.4.1.

[ndmitchell]

Testing on appveyor, I get failures when using 32bit Stack+arch, and success on 64bit, holding everything else constant. I'm seeing it on filepath, yaml and my neil executable. The message " Process exited with code: ExitFailure (-1073741819)" is constant.

I'm ignoring network because it seems to be a bug that is understood and being tracked down separately.

[ndmitchell]

The network thing has been moved to commercialhaskell/stack#3944, which shows that stack is providing different environments in different circumstances. It is entirely possible this bug is a different manifestation of that, so suggesting waiting til that is resolved to investigate further.

[Mistuke]

The segfaults (0xC0000005) are a GHC bug, https://ghc.haskell.org/trac/ghc/ticket/15154 I'll have a fix for 8.6. I believe 8.4 and 8.2 suffer from this issue.

[jgm]

I'm seeing the same thing as @ndmitchell for pandoc - failures on 32-bit Windows with stack and ghc 8.4.3, with ExitFailure (-1073741819). I guess I'll have to revert to ghc 8.0.2 until the fix is out.

[arrowd]

The problem is now in lts-12.7 which is very sad.

[ndmitchell]

@arrowd See https://www.fpcomplete.com/blog/2018/06/sed-a-debugging-story which gives a detailed explanation of the problem and solution

[arrowd]

@arrowd See https://www.fpcomplete.com/blog/2018/06/sed-a-debugging-story which gives a detailed explanation of the problem and solution

This indeed worked for me. Thanks for sharing the link.

[mihaimaruseac]

Is this still an issue?

[ndmitchell]

I haven't seen it in a while, so closing.