STR (with the NVDA screen reader):
- In NVDA menu -> Settings -> Preferences -> Speech:
- Set the Synthesizer to eSpeak NG.
- Enable Automatic language switching.
- Open this test case:
data:text/html,<img alt="123" lang="de" src="https://via.placeholder.com/10x10">
- Press down arrow to read the line.
- Expected: NVDA speaks "einhundertdreiundzwanzig" (123 in German).
- Actual: NVDA speaks "one hundred twenty three" in English.
In bug 1757127, I implemented support for Accessible::Language using the text atributes cache. However, images aren't HyperTextAccessibles, so they don't have text attributes. We'll need special handling for this. We can either:
- Cache language in the main cache for this case.
- Create a text attributes cache with just this one attribute. While a bit wasteful, it does make the retrieval code simpler.
Either way, we'll need a different code path in BundleFieldsForCache.