I found that many (all?) string functions have a mulitbyte version. Switching to these will hopefully be enough.

This is implemented in a local branch. Review will wait until T158954: Use XPath to get text nodes related to utterances is done. In solving this, a fair bit of the segmenting was rewritten, which also solves the following tasks:

• T159811: Overlapping highlighting
• T159809: Highlighting fails on bold/italicized utterance

and the following may well be solved (and should be rechecked) when both this and T158954 are done:

• T159669: Empty <text> elements created
• T159671: IndexSizeError triggered and cascades into failing buttons
• T159810: Player button fails to return to stop upon completed recitation of page

Worked on in Wikispeech (Sprint 2017-03-08):

• Local implementation.

To do in Wikispeech (Sprint 2017-03-22):

• Upload patch to gerrit.
• Review.

Change 344616 had a related patch set uploaded (by Lokal Profil; owner: Sebastian Berlin (WMSE)):
[mediawiki/extensions/Wikispeech@master] Calculate correct offsets for unicode characters

https://gerrit.wikimedia.org/r/344616

Change 344616 had a related patch set uploaded (by Sebastian Berlin (WMSE)):
[mediawiki/extensions/Wikispeech@master] Calculate correct offsets for unicode characters

https://gerrit.wikimedia.org/r/344616

Change 344616 merged by jenkins-bot:
[mediawiki/extensions/Wikispeech@master] Calculate correct offsets for unicode characters

https://gerrit.wikimedia.org/r/344616

Mentioned in SAL (#wikimedia-labs) [2017-03-30T10:15:39Z] <Sebastian-WMSE> Deploy latest from Git master: 64cbd96 (T159545, T159811, T159809)