[lydell]

Expect.all currently stops at the first failing expectation. This PR uses the Multiple reason from #126 (added for Expect.oneOf) to report all failing expectations in one go.

Unfortunately, the flattening from commit 59f6117 assumes that Multiple is only used for Expect.oneOf, despite being named generically. I modified the existing test case to use a mix of oneOf and all:

Expect.oneOf
    [ 1 |> Expect.equal 2
    , Expect.all
        [ 2 |> Expect.equal 3
        , 3 |> Expect.equal 4
        ]
    , 4 |> Expect.equal 5
    ]

The failure from the above is:

{ description = "Expect.oneOf: none of the expectations passed."
, given = Nothing
, reason =
    Multiple
        [ { description = "Expect.equal", given = Nothing, reason = Equality "2" "1" }
        , { description = "Expect.equal", given = Nothing, reason = Equality "3" "2" }
        , { description = "Expect.equal", given = Nothing, reason = Equality "4" "3" }
        , { description = "Expect.equal", given = Nothing, reason = Equality "5" "4" }
        ]
}

• The "Expect.all: some of the expectations failed." description is missing.
• The failure makes it look like fixing any of the sub-failures will make the test pass, but that’s not true: The two middle ones need to both be fixed together, since they are in an Expect.all.