Rod

We'll need a font corpus to feed it. After a quick glance at https://github.com/google/oss-fuzz/blob/master/projects/freetype2/Dockerfile, perhaps `*.[ot]t[cf]` from: 1. https://github.com/unicode-org/text-rendering-tests 1. https://github.com/googlefonts/color-fonts 1. https://github.com/harfbuzz/harfbuzz/tree/main/test https://google.github.io/oss-fuzz/getting-started/new-project-guide/rust-lang/ has Rust-specific setup instructions.

If you have sources it might be better to transform the sources. If you only have a binary then you could likely write a [FontTools](https://github.com/fonttools/fonttools) python tool that takes in...

Id really like to see this pop out on a profile or some sort of measurement before making the code more complicated for it. On Tue, Sep 19, 2023, 10:51 ...

> This isn’t something we can easily duplicate on our end at the moment. Naively I would have thought we could either precompute the shaping and freetype result and commit...

> we need a measure of accuracy of final rendering output Maybe we can redirect this issue to focus on how we make that easy.

A concrete example would help.

I suggest we start with only the case where all variants support; the user having to drill down to the specific variant seems pretty reasonable when that's not the case.

Use of Google team audit files, crucial for this to be truly useful for us, appears to be blocked on https://github.com/mozilla/cargo-vet/issues/394.

394 above is resolved and https://hg.mozilla.org/integration/autoland/rev/1c29e191de1c shows use of the Chrome audit file.

https://github.com/google/supply-chain aggregates Google audits.