Sam Hardwick

> I'm facing the exact same problem. I'm on Quod Libet 4.5.0, Manjaro (both KDE Plasma & awesomewm), and I use fcitx5. Did you find a solution to this? Only...

I've been testing the proposed change mostly in the FiNer project, which is the most complex pmatch project I know of. One very important thing to understand about character classes...

Hey, sorry I missed this issue - are you aware of the --tokenize-multichar option? From hfst-tokenize --help: -m, --tokenize-multichar Tokenize multicharacter symbols (by default only one utf-8 character is tokenized...

Yeah, I guess that would be a logical place for it.

Kevin, do I understand correctly that the difference was between Ins() and non-Ins() cases? In that case, let me explain something (that you might already know, but let me say...

Hang on - you're not asking me, are you, @unhammer? Or is the Ins() behaviour still wrong?

Sorry I haven't gotten around to this yet - been busy with other things - but I'll look into it now. As for #400, the general case of epsilon alignment...

Could this be a precedence issue? I didn't check with your analyser, but testing with omorfi led me to think what you maybe want is eg. morphology & [ incondform:[?\*]...

Hey, thanks a ton for the report! This problem is fortunately completely avoidable, though there is a bug here that needs fixing too. First, some details you don't really want...

Ahh, unfortunately I've never run any hfst stuff on Windows, I don't really know. Where do you get the segfault from, though? "Fatal Python error" sounds like maybe something is...