Hadrian fails to run the testsuite.

This is on windows with a tree based on the commit below.

commit a5373c1f (master)
Author: Simon Peyton Jones <simonpj@microsoft.com>
Date:   Wed Jan 16 16:34:24 2019 +0000

    Fix bogus worker for newtypes

Output as follows.

$ hadrian/build.bat -c -j8 --flavour=quick test
Up to date
Up to date
hadrian: Encountered missing dependencies:
libiserv ==8.7

shakeArgsWith         0.001s    0%
Function shake        0.010s    0%
Database read         0.516s   13%  ===
Database compression  0.074s    1%
With database         0.018s    0%
Running rules         3.280s   84%  =========================
Total                 3.900s  100%
Error when running Shake build system:
  at src/Main.hs:58:30-42:
* Depends on: test
  at src/Rules/Test.hs:90:5-17:
* Depends on: _build/stage1/lib/bin/ghc-iserv.exe
  at src/Development/Shake/Internal/Rules/Oracle.hs:157:43-68:
* Depends on: OracleQ (ContextDataKey (Context {stage = Stage1, package = Package {pkgType = Program, pkgName = "iserv", pkgPath = "utils/iserv"}, way = v}))
  at src/Hadrian/Haskell/Cabal/Parse.hs:202:5-36:
* Depends on: _build/stage1/utils/iserv/setup-config
* Raised the exception:
ExitFailure 1

Regular build with hadrian works.

Trac metadata

Trac field	Value
Version	8.6.3
Type	Bug
TypeOfFailure	OtherFailure
Priority	normal
Resolution	Unresolved
Component	Compiler
Test case
Differential revisions
BlockedBy
Related
Blocking
CC
Operating system	Unknown/Multiple
Architecture

changed weight to 5

added Tbug Trac import labels

Trac metadata

Trac field	Value
Component	Compiler → Build System (Hadrian)
Operating system	Unknown/Multiple → Windows

Verbose output: https://www.irccloud.com/pastebin/rWn7WMsx/

Andrey, any idea of what's going on here?

Alp, thanks for tagging me!

As far as I understand, iserv is not supported on Windows, so when running the testsuite on Windows, we shouldn't depend on it and skip all associated tests.

I guess, this line may be relevant:

https://gitlab.haskell.org/ghc/ghc/blob/aad05fb3b36b93b919622f8a6dc032109d040d16/hadrian/src/Settings/Default.hs\#L137

Andreas: perhaps, you could try making iserv conditional, as for example, we do with timeout:

https://gitlab.haskell.org/ghc/ghc/blob/aad05fb3b36b93b919622f8a6dc032109d040d16/hadrian/src/Settings/Default.hs\#L141

Does this help?

Great, thanks for replying Andrey. Note that this stage<N>Packages & testsuitePackages split has been bothering me a fair bit recently for WIP patches that I have, and it should all probably be revisited, but one thing at a time. :-)

added Windows label

@snowleopard iserv was supported on Windows and should still be. If it's not working this is a regression. The original implementation has Windows support on day 1. https://github.com/ghc/ghc/blob/master/utils/iserv/src/Main.hs#L29 and other supporting code.

@angerman do you know anything about this?

Yes. Iserv does work on windows. I use it every day!

I believe this camre with the iserv -> iserv+libiserv split and hadrian just fails to build the iserv library.

If iserv does work on Windows, then perhaps it should simply be added to the list of packages built by default and the list of testsuite packages too. I think it was behind "not Windows" flag for as long as I can remember.

FWIW, my somewhat recent series of patches related to Windows or the testsuite should get us closer to being able to run the testsuite. Perhaps we can even run it. At one point I had a branch (with most of those changes) where I successfully ran the testsuite against a GHC built by Hadrian on Windows, resulting in huuuundreds of unexpected failures, but still. I'll try and see where we stand on this next week, when I come back to work.

added Pnormal label