If I have test/Test.hs with
module Main where
import Test.HUnit
test1 :: Test
test1 = TestCase $ assertEqual "Should be one" 1 5
test2 :: Test
test2 = TestCase $ assertEqual "Shold both be zero" 0 0
main :: IO Counts
main = runTestTT $ TestList [test1, test2, test1]
and a .cabal with
test-suite my-test
type: exitcode-stdio-1.0
hs-source-dirs: test
main-is: Test.hs
build-depends: base >= 4.8.1.0 && <4.9,
HUnit >= 1.3
default-language: Haskell2010
and I run cabal test --show-details='always' then I get
Test suite my-test: RUNNING...
### Failure in: 0
test/Test.hs:6
Should be one
expected: 1
but got: 5
### Failure in: 2
test/Test.hs:6
Should be one
expected: 1
but got: 5
Cases: 3 Tried: 3 Errors: 0 Failures: 2
Test suite my-test: PASS
Why does my test suite pass when I've had failures? Likewise, if I cabal sdist I get no warning that my tests have failed.
According to the Cabal users' guide,
Test suites using the
exitcode-stdio-1.0interface are executables that indicate test failure with a non-zero exit code when run; they may provide human-readable log information through the standard output and error channels.
You've defined
main :: IO Counts
main = runTestTT $ TestList [test1, test2, test1]
This runs tests, prints out test information, and then always exits successfully. If you want Cabal to know the test has failed, you need to capture the Counts, check for errors and failures, and exit with a non-zero status if you find such.
import System.Exit
main :: IO ()
main = do
results <- runTestTT $ TestList [test1, test2, test1]
if (errors results + failures results == 0)
then
exitWith ExitSuccess
else
exitWith (ExitFailure 1)
The test-framework package provides convenient defaultMain functions that do this sort of thing; you may want to consider that approach.
You should note that the exitcode-stdio-1.0 interface is considered semi-deprecated; the Cabal maintainers recommend switching to their rather more Haskellian detailed-0.9 interface.
Collected from the Internet
Please contact [email protected] to delete if infringement.
Comments