stack "errors" when building package with benchmark that depends on itself #2153
Description
Originally discovered here. Steps to reproduce:
$ stack unpack vector-binary-instances
Unpacked vector-binary-instances-0.2.3.2 to /home/ryanglscott/Documents/Hacking/Haskell/vector-binary-instances-0.2.3.2/
$ cd vector-binary-instances-0.2.3.2/
$ stack init --resolver lts-5.16
Using cabal packages:
- vector-binary-instances.cabal
Selected resolver: lts-5.16
Initialising configuration using resolver: lts-5.16
Total number of user packages considered: 1
Writing configuration to file: stack.yaml
All done.
$ stack bench --only-dependencies
...
$ stack bench
vector-binary-instances-0.2.3.2: benchmarks
vector-binary-instances-0.2.3.2: configure (bench)
setup-Simple-Cabal-1.22.5.0-ghc-7.10.3: No benchmarks enabled. Did you
remember to configure with '--enable-benchmarks'?
Configuring vector-binary-instances-0.2.3.2...
vector-binary-instances-0.2.3.2: build (bench)
Preprocessing library vector-binary-instances-0.2.3.2...
[1 of 1] Compiling Data.Vector.Binary ( Data/Vector/Binary.hs, .stack-work/dist/x86_64-linux/Cabal-1.22.5.0/build/Data/Vector/Binary.o )
In-place registering vector-binary-instances-0.2.3.2...
Preprocessing benchmark 'benchmarks' for vector-binary-instances-0.2.3.2...
[1 of 1] Compiling Main ( benchmarks/Benchmarks.hs, .stack-work/dist/x86_64-linux/Cabal-1.22.5.0/build/benchmarks/benchmarks-tmp/Main.o )
Linking .stack-work/dist/x86_64-linux/Cabal-1.22.5.0/build/benchmarks/benchmarks ...
Completed 2 action(s).
-- While building package vector-binary-instances-0.2.3.2 using:
/home/ryanglscott/.stack/setup-exe-cache/x86_64-linux/setup-Simple-Cabal-1.22.5.0-ghc-7.10.3 --builddir=.stack-work/dist/x86_64-linux/Cabal-1.22.5.0 bench benchmarks
Process exited with code: ExitFailure 1
stack seems to be confused about vector-binary-instances's benchmarks, for the reason that (I assume) its benchmarks transitively depends on vector-binary-instances itself. It still appears to build correctly, but it exits with an error code and doesn't actually run the benchmarks.