[spine-cpp] SkeletonRenderer does not start a mask if clipping attachment attached to the slot with 0 alpha.

Hi,

Our team uses the spine-cpp runtime in our engine. Recently, we've updated our runtime from spine-c version 3.8.99 to spine-cpp version 4.2.36. While porting the code, it became clear that it would be easier to rewrite it from scratch.

For the most part, I used the spine-sfml C++ implementation as a reference. However, since we use custom rendering code, we cannot use SkeletonRenderer directly, but I used it as a guide for porting the rendering logic. To my surprise, after testing some animations re-exported for version 4.2, the new implementation had an issue with clipping: some masks were completely skipped. After some investigation, it became clear that the problem was along these lines:

/// for each slot in draw order
Slot &slot = *skeleton.getDrawOrder()[i];
Attachment *attachment = slot.getAttachment();
if (!attachment) {
    clipper.clipEnd(slot);
    continue;
}

// Early out if the slot color is 0 or the bone is not active
if (slot.getColor().a == 0 || !slot.getBone().isActive()) {
    clipper.clipEnd(slot);
    continue;
}

Some of our clipping attachments were attached to slots with an alpha value equal to 0, so they were skipped due to the early-out condition, and hence clipping did not work properly. I added a temporary workaround for this issue by checking the attachment type before continuing to the next attachment, but it would be nice to see a standardized solution, especially for those who use SkeletonRenderer as their main renderer.

...
auto& draw_order = m_skeleton->getDrawOrder();
for (size_t i = 0; i < m_skeleton->getSlots().size(); i++)
{
    auto& slot = *draw_order[i];
    auto* attachment = slot.getAttachment();
    auto attachment_type = GetSlotAttachmentType(slot); // Returns 1 out of 4: eNull, eRegion, eMesh or eClipping

    if ((!attachment || cpp::is_equal(slot.getColor().a, 0.0F) || !slot.getBone().isActive())
        && attachment_type != AttachmentType::eClipping)
    {
        m_clipping.clipEnd(slot);
        continue;
    }
....
}
...

I assume that my workaround is correct because it mimics the Spine editor behavior for the animation. Unfortunately, I cannot provide an example animation, as it was created within the studio, and I am not allowed to publish it.

EsotericSoftware / spine-runtimes

[spine-cpp] SkeletonRenderer does not start a mask if clipping attachment attached to the slot with 0 alpha. #2635