I happened to look at the spelling logs and saw it was still erroring quite a log. The implementation has two problems related to stream chunk handling on the output from aspell. Two fixes:
set the stdout encoding to utf8 instead of buffer, so that utf8 characters don't break across stream chunks
extend the matcher to allow for aspell's 'end of data' marker being split across two chunks
Description
I happened to look at the spelling logs and saw it was still erroring quite a log. The implementation has two problems related to stream chunk handling on the output from aspell. Two fixes:
I am cautiously hopeful this will fix the problems in https://github.com/overleaf/issues/issues/2007
Screenshots
NA
Related Issues / PRs
https://github.com/overleaf/issues/issues/2007
Review
Potential Impact
Spelling only, so low - can roll back if there is a problem.
Manual Testing Performed
Accessibility
NA
Deployment
NA
Deployment Checklist
Metrics and Monitoring
Spelling error logs in stackdriver
Who Needs to Know?
cc @mserranom