[Bug] Sector bugs caused by precision.

[plainheart]

Version

5.3.0

Link to Minimal Reproduction

test/pie.html

Steps to Reproduce

1. Open the test case pie.html
2. Unselect all the legends
3. Watch if the final rendered chart is a circle

Current Behavior

Got a tiny sector that looks like a line

Expected Behavior

Got a circle

Environment

- OS: Any
- Browser: Any
- Framework: No

Any additional comments?

Arguments of this broken sector:

• startAngle: -1.5707963267948966
• endAngle: 4.712388980384691

The calculated arc is 6.283185307179587, which is slightly bigger than 2π... so that it can't be rendered as a circle.

[plainheart]

@pissang How can we perfactly solve such an issue?

[pissang]

@plainheart Didn't notice this when reviewing the code. mod is not precision safe. We need to check if the angle is larger than PI2(is a circle) before doing the mod. That's what normalizeArcAngles do.

[pissang]

The issue fixed in ecomfe/zrender#871 seems also related to precision. The startAngle is 0.6981317007977319 and endAngle is 0.6981317007977319. Perhaps we can add an epsilon when checking if startAngle is same with endAngle

[pissang]

Also we need more visual regression tests for these cases. Seems it's not reported in current tests.