Stabilize CI?

[mramato]

Random Vector3DTileContent test failure seem to be the primary cause of CI problems these days: https://travis-ci.org/AnalyticalGraphicsInc/cesium/builds/602749111

These tests are a mess and do way to much work to really be considered unit tests, the larger problem is the architecture that requires loading a full tileset just to test Vector3DTileContent.

For now, I'm just trying to stabilize CI and I think moving some code into beforeEach instead of beforeAll may do it. Opening this so I can force it a bunch of times and see how it does.

I'm tempted to just ignore Vector3DTileContent specs completely if this doesn't fix it, since having constant random CI failures is really a drag on Cesium development.

[cesium-concierge]

Thanks for the pull request @mramato!

• ✔️ Signed CLA found.
• ❔ CHANGES.md was not updated.
  • If this change updates the public API in any way, please add a bullet point to CHANGES.md.

Reviewers, don't forget to make sure that:

• Cesium Viewer works.
• Works in 2D/CV.
• Works (or fails gracefully) in IE11.

[mramato]

Nope, Vector3DTileContent still failed.

@lilleyse any objections to just xit-ing that entire file until we have a chance to due a full review of it?

[lilleyse]

@mramato fine with me until we can figure out why.

I ran just Vector3DTileContent tests on CI and it didn't time out. Either I got lucky or it fails only when the full suite is run.

[14:17:50] Starting 'test'...
0.433 secs: Scene/Vector3DTileContent renders polygons
0.309 secs: Scene/Vector3DTileContent renders points
0.169 secs: Scene/Vector3DTileContent renders on 3D Tiles
0.147 secs: Scene/Vector3DTileContent renders polylines
0.1 secs: Scene/Vector3DTileContent renders points with a batch table
0.093 secs: Scene/Vector3DTileContent renders polygons with a batch table
0.084 secs: Scene/Vector3DTileContent renders batched points with a batch table
0.078 secs: Scene/Vector3DTileContent renders polylines with a batch table
0.077 secs: Scene/Vector3DTileContent renders batched points
0.071 secs: Scene/Vector3DTileContent renders on 3D Tiles and globe
HeadlessChrome 78.0.3904 (Linux 0.0.0): Executed 28 of 10864 (skipped 10836) SUCCESS (9.263 secs / 2.267 secs)
TOTAL: 28 SUCCESS

[mramato]

I ran just Vector3DTileContent tests on CI and it didn't time out. Either I got lucky or it fails only when the full suite is run.

It fails randomly, sometimes it works, sometimes it doesn't. It is much more likely to fail on travis (possibly because of a slow machine, I don't know). But the problem is that we really need CI to only fail when there is a legitimate error or else it just slows us all down.

[mramato]

This continues to be a thorn in our side, I'm going to xdescribe the spec and bump CI a bunch of times.

[lilleyse]

Yup, fine with me. We can look into it later.