core: Fix text culling check

[britt-j]

The check relied on y_max being the same as height, which isn't the case since y_max can be negative
(close #16877)

[britt-j]

Not sure what value this should have

[adrian17]

LGTM, but wanna make sure kjarosh approves too :)

[kjarosh]

Thank you for your contribution!

There are several pitfalls that we can encounter in text-based tests.

It's important to make sure the test produces the same output as Flash Player. Unfortunately that's not the case for edittext_negative_bounds. If you open the SWF in FP, you can see that the displayed text differs a lot.

This is specifically caused by the fact that device fonts are used to render the text. It's the case despite having a font set, because useOutlines is false. It has to be set to true in order for embedded fonts to work. When you set it to true, note that it will not work properly because currently, the embedded font has 0 ascent and descent set.

In general, I recommend using fonts that are vastly different from the default ones. I personally use my own crafted fonts and commit them (which you can also just embed to the SWF), see for instance: https://github.com/ruffle-rs/ruffle/blob/e9926f9a2e4f518a4095d9c1643440031033f14f/tests/tests/swfs/visual/fonts/leading_define_font/TestFont.ttf. I recommend doing that because of several reasons:

1. a test font does not have complex shapes,
2. it's easy to spot that a wrong font is used,
3. it's possible to modify properties of the test font and provide custom values.

Regarding the output, it's best if the output comes from the Flash Player, and not Ruffle. It's enough to screenshot FP and replace the output.expected.png with it. In that case, you'll need to set tolerance and outliers. I usually set outliers = 0 and tolerance = 128 (along with quality = 4), which is enough for black and white text-based tests. Setting scaleMode to noscale helps a lot when taking a screenshot.

[britt-j]

Thanks for the info, I'm sorry if I missed that anywhere else.

I do have a few questions though, if you don't mind

I usually set outliers = 0 and tolerance = 128 (along with quality = 4)

Do you mean the sample_count in with_renderer?, because I can't find anything about quality in the test readme nor in the other tests in the edittext folder

Regarding the output, it's best if the output comes from the Flash Player, and not Ruffle.

I'm sorry if I've misunderstood what image tests are for, but I thought they were moreso regression tests for Ruffle specifically rather than accuracy tests?
I'm still going to take a screenshot from FP for this test, I'm just wondering

In general, I recommend using fonts that are vastly different from the default ones. I personally use my own crafted fonts and commit them (which you can also just embed to the SWF), see for instance: https://github.com/ruffle-rs/ruffle/blob/e9926f9a2e4f518a4095d9c1643440031033f14f/tests/tests/swfs/visual/fonts/leading_define_font/TestFont.ttf.

So it'd be alright if I used that font you linked, or should I make a different one for this?

[kjarosh]

Do you mean the sample_count in with_renderer?, because I can't find anything about quality in the test readme nor in the other tests in the edittext folder

Yes, I'm sorry, I meant sample_count.

I'm sorry if I've misunderstood what image tests are for, but I thought they were moreso regression tests for Ruffle specifically rather than accuracy tests? I'm still going to take a screenshot from FP for this test, I'm just wondering

Well, I guess there's no universal definition. Some tests in fact do that due to the fact that it's often extremely hard to get pixel-perfect output in Ruffle. In the case of simple text tests, I've put a lot of effort into making text near-pixel-perfect, so it should be fairly easy. Not to mention that we won't have to update the test when something changes in Ruffle, it should automatically pass (provided that something changes in the right direction ;) ).

So it'd be alright if I used that font you linked, or should I make a different one for this?

Sure thing, you can simply embed this TTF in JPEXS and you're good to go. No need to create a custom font. Creating a custom font would be beneficial if you wanted to test something specific related to that font, but here we just want to check whether the text renders. Just remember that this font defines only 4 characters if I remember correctly (a, b, c, d).

If you need some help with something just let me know, I can polish up the test :)

[britt-j]

@kjarosh don't know if the force pushes sent a notification, but if not, this should finally be ready now

[kjarosh]

Perfect! Thank you!