I don't think this is a breaking change in Swash, but there seems to be at least one other package that re-exports zeno, and changing Mask::offset's behavior is technically a breaking change here in zeno. I've decided to go with the more conservative option of just passing both offset and render_offset in from Swash, just to make extra sure there's no breakage.

I tried to use this branch, but it seems the vertical cut offs don't seem to be fixed. Here's what I see in cargo tree:

├── parley v0.3.0 (/tmp/parley/parley)
│   └── swash v0.2.1 (/tmp/swash)
│       └── zeno v0.3.2 (/tmp/zeno)

[/tmp/zeno]$ git branch -v
  main         dbdd3d3 bump patch version for release (0.3.2)
* tight-bounds 696f941 Update documentation for Mask::offset

Here's the image I get with parley's swash_render example:

Do I need to activate this branch somehow to get the fix? Thanks

You'll need to switch Swash to this branch. You can point it at mine:

swash = { git = "https://github.com/valadaptive/swash", branch = "tight-bounds" }

which is the same as the one in dfrg/swash#88, except patched to use this branch of zeno.

Great! That fixed it. Brilliant work.

Result:

The added bonus of Parley's swash_render example being fixed goes away with parley#343 applied. Which is to say that the swash_render issue may have had a different root cause.

With this padding removal here applied (combind with swash#88), swash_render output will end up being visually unchanged.

parley#343	+ zeno#15	+ zeno#15 & swash#88
Looks good	Clipped because of Swash bug	Back to looking good

Isn't removing the padding from the output a breaking change for consumers of Swash? If they were previously accounting for that then they might need to undo their adjustment? Or will that "just work"?

That's what swash#88 is for. If both Swash and Zeno get a patch release, then both will be picked up by cargo update - or neither. (Unless someone goes out of their way to update their lock file with just one.)

There's no compilation issues or indeed any required action by consumers. So all of this can go quite smoothly with patch releases.

Thanks, the reasoning makes sense so let’s go ahead and land this.

Subpixel formats might require some additional padding which might explain why I hacked this in unconditionally. I’ll double check that before releasing.