Test JXcore on Android Nougat #754
Comments
Preliminary results are not very good: on Nougat the test/simple tests return 65 failures compared to 15 failures on Marshmallow (SpiderMonkey). |
Full results for SpiderMonkey coming soon. |
|
All results using SM on x86 emulator. Tests in 'test/simple' that fail on Marshmallow but not on Nougat:
Tests in 'test/simple' that fail on Nougat causing the test script to crash:
Tests in 'test/simple' that fail on Nougat causing the test script to hang:
Tests in 'test/simple' that fail on Nougat but not on Marshmallow:
|
Tests in 'test/simple' that fail on Nougat but not on Marshmallow, detailed info:
|
The first goal is to figure out why we are getting the new failures and then we can decide what we want to fix. There is at least a week of work here. |
Nougat ARMv8 using SM.
Tests that hang:
Tests in 'test/simple' that fail on Nougat but not on Marshmallow:
|
1/2 the tests were failing because of a problem with the test runner javascript file spawning a child instance so we fixed that by using a bash script instead. 10 of the remaining 30 failing test were failing because of a pipe for locking that isn't working and we are investigating that now. So once we deal with these 10 then we have 4 or 5 tests that are failing on Nougat that aren't failing on Marshmallow. |
12 tests that always failed on Marshmallow are now not failing due to work by Enrico. He thinks by today he should have 0 tests failing on Marshmallow and then will clean up Nougat. There seem to be two types of problems. In some cases the test just isn't appropriate for Android and in other cases there are small fixes needed in the tests themselves. We will also try to rerun the tests on iOS 9 to see how the fixes apply there and we can see where we are with iOS. We now can see more clearly when a test is failing and why. |
Four new issues open for tests failing on Nougat: |
Down to 15 bugs. 11 fail for the same reason in LibUV. The bugs have been classified above and we'll decide what to fix. The LibUV issue has to do with pipes.Another involves a symbolic link where the issue seems to be the test, not the code. In another test we fork a child and it doesn't seem to work. And the final one there is a parent and a child and the parent doesn't receive a standard output event because the child seems to exit before the event is received by the parent. So the last two are similar, we think. The last one that is still being investigated (don't have a bug for this one yet). |
The investigation is completed, resulting in the following issues: thaliproject/jxcore#52 |
PR submitted thaliproject/jxcore#57 |
Does it work there?
The text was updated successfully, but these errors were encountered: