[raphamorim]

OSC 7, OSC 8 and OSC 9;9 carry verbatim payloads, but the OSC parser split them on every ; and kept at most 16 params. OSC 8 then used only the third param as the URI, so https://x.dev/a;b opened as https://x.dev/a, and OSC 7 / 9;9 rejoined the split params in the handler, which still lost anything past the 16th ;.

The parser now stops splitting once it reaches the verbatim part of these sequences, keeping the ; as data: after 7;, after 8;<params>;, and after 9;9;. Other OSCs (including 9;4 progress) split as before. The check runs only on a ; while the param count is 1 or 2, so the hot path is unchanged. join_params is gone.

OSC 8 handling:

• the URI is everything after the second ;, however many ; it contains
• a URI with invalid UTF-8 is decoded lossily and starts a new link, instead of the old text staying attached to the previous link
• an empty URI always closes the link, also when an id is given; an empty id= counts as no id
• hint labels, hover and click open the URI exactly as the program sent it; post-processing (trailing punctuation, unmatched brackets) now only applies to regex matches

Cleanup: the regex hint path uses trim_match_tail instead of its duplicate post_process_hyperlink_uri, and the duplicate copy in screen/mod.rs is removed. The old tests moved over, plus a case for a leading unmatched delimiter.

Tests: parser tests for OSC 8, 7 and 9;9 payloads with more than 16 ; and for 9;4 still splitting; terminal tests for ; in an OSC 8 URI, invalid UTF-8 and 8;id=foo; closing.

Docs (post-processing scope, OSC 8 URI rule, changelog) are on the osc8-verbatim-uri branch of rioterm.com.