On Fri Mar 20 13:57:42 2026 +0000, Francis De Brabandere wrote:
I think this can maybe be solved on the parser side but with a lot more duplication. The current result is what I ended up with after many iterations. It's also already validated and in use on the vpinball project. I fear that some of these issues just don't have a clean solution. But I can have another go on the parser side. Maybe in a different MR with the same tests. Then we can compare. After having another look I don't think we can handle this without the help of the lexer prediction/parenthesis tracking because bison can't look ahead. A GLR parser can do this, eg https://github.com/uwol/proleap-vb6-parser/blob/main/src/main/antlr4/io/prol...
The current solution works, there's good test coverage. Nothing keeps somebody to come up with a cleaner solution later. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10244#note_133122