Small mistake: I added only one screenshot showing all the labels.

Subpixeling issue ?
If so, it shouldn't be complicated to fix.

I think if we remove the floating part of the x,y,z in transform and also in the position it should work.
By that I mean, fixing CCLabel.cpp # onDraw.

@minggo, can we fix this for 3.0?
If not, what about v3.1 ?

I tried some stuff to ensure there is nothing drawn on sub pixels, but with no luck.

It is not easy to understand what is happening in the Label class.
The CCLabel code is quite complicated as it there is code for LabelTTF, LabelBMFont and Label all in one place. E.g. a Label does nervier call its onDraw or updateQuads methods.

Why is Label using the _textSprite to render? I thought Label is like a LabelBMFont, but with the difference that the texture is generated at runtime.

I think Label should be cleaned up by moving stuff in derived classes. It is very hard to understand if half of the code is never used depending on how the label was created.

The path to the ttf file was wrong. After fixing that onDraw is called. There is a warning about different behaviour in the header of CCLabel. I think this is not enough and the fallback when a rtf is missing is quite dangerous.

agreed.
the Label code needs to be cleaned up

I found the real source of the unsharp labels in my test program.

1. You have to set the projection of the director to Director::Projection::_2D
2. Do not set any design resolution or content scale factor.

Would be interesting to know why 2D projection is required. Thats ok for out current project, but also in 3D you want sharp labels

If a Label has CENTER alignment, the characters are sometimes drawn on sub pixels even with 2D projection and no content scale factor.

I send a small gitHub pull request, fixing that.

Is there a workaround for this defect, other than <ins>not</ins> setting the design size or avoiding the Label components?

Redmine issue system is closed, we are using github issue system instead.

This issue was moved to https://github.com/cocos2d/cocos2d-x/issues/8377