mmcloughlin / globe

Globe wireframe visualizations in Golang
https://pkg.go.dev/github.com/mmcloughlin/globe
ISC License
1.59k stars 51 forks source link

TestGraticuleCenterOn fails on arm64 #9

Open mmcloughlin opened 2 years ago

mmcloughlin commented 2 years ago

On Apple M1:

> GOARCH=arm64 go test -v
=== RUN   TestGraticule
--- PASS: TestGraticule (0.10s)
=== RUN   TestGraticuleCenterOn
    globe_test.go:50:
            Error Trace:    globe_test.go:50
                                        globe_test.go:63
            Error:          Not equal:
                            expected: "04c8271e8d48ea580a6d340d9be7a261"
                            actual  : "80e7f15997d3597d8cc451876e0bc7ae"

                            Diff:
                            --- Expected
                            +++ Actual
                            @@ -1 +1 @@
                            -04c8271e8d48ea580a6d340d9be7a261
                            +80e7f15997d3597d8cc451876e0bc7ae
            Test:           TestGraticuleCenterOn
--- FAIL: TestGraticuleCenterOn (0.09s)
=== RUN   TestDrawDots
--- PASS: TestDrawDots (0.03s)
=== RUN   TestDrawLand
--- PASS: TestDrawLand (0.06s)
=== RUN   TestDrawCountries
--- PASS: TestDrawCountries (0.08s)
=== RUN   TestLine
--- PASS: TestLine (0.02s)
=== RUN   TestRect
--- PASS: TestRect (0.02s)
=== RUN   TestCartestian
--- PASS: TestCartestian (0.00s)
=== RUN   TestHaversine
--- PASS: TestHaversine (0.00s)
=== RUN   TestIntermediate
--- PASS: TestIntermediate (0.00s)
=== RUN   TestDestination
--- PASS: TestDestination (0.00s)
FAIL
exit status 1
FAIL    github.com/mmcloughlin/globe    0.597s

The test still passes under amd64. Unclear to me right now why the output should be different. On visual inspection, the images look right on both architectures, but there are tiny pixel value differences.