[anyingiit]

What & why

Fixes both defects in #586:

• Decorated blank runs are drawn. to_svg trimmed each run's trailing blanks and skipped a run left empty, decoration and all, so an underlined empty form field (printf 'Name: \033[4m \033[0m|') vanished from the SVG while to_html drew it. A run with underline or strikethrough now keeps its blanks, so the line is drawn under them as a terminal draws it. That includes the trailing blanks of a decorated run that has text: an underlined "ab " is now underlined across all five cells, as in to_html, where before the line stopped after "ab". Undecorated runs are trimmed as before. Concealed text still draws nothing in the SVG; to_html draws the underline of a concealed underlined run, and I left that existing difference alone.
• No characters XML forbids. escape_xml now writes every character outside the XML 1.0 Char production (C0 controls other than tab, LF and CR, plus U+FFFE and U+FFFF) as U+FFFD, one for one so a cell stays one column. A window title holding \x01 or a cell holding U+FFFF no longer makes an SVG no parser opens. to_html shares the function, so it gets the same treatment.

Tests in crates/termlens/tests/export.rs: a decorated-blanks screen, U+FFFF in a cell (both built with Screen::parse, no PTY), and \x01 in an OSC 0 title through the emit fixture (ignored on Windows with the usual #149 reason, like the other fixture-driven export tests). Each checks the output with a small assert_well_formed_xml helper: every character is an XML Char and the tags nest. That is not a full parse (entities and attribute syntax aren't checked); I also parsed the test SVGs with xmllint by hand and they are well-formed. I kept it dependency-free rather than adding an XML parser as a dev-dependency. If you'd prefer a real parse (as asked in #376), I'm happy to switch it to roxmltree or quick-xml.

Closes #586

Motivation / disclosure: I had some spare AI-assistant quota (Claude Code) and am using it to try to help projects with open good-first-issues. The change was prepared with Claude Code and verified as listed below; per CONTRIBUTING §6 the commit carries no AI attribution and I'm the author of record under the DCO. If it doesn't fit, isn't up to your bar, or you'd simply rather not take it, please feel free to close it, no hard feelings at all.

Checklist

• Linked an issue (to_svg drops underlined blank fields and writes characters XML forbids #586)
• Tests added/updated for the change: three new tests; all three fail with only render.rs reverted and pass with it
• cargo fmt --all and cargo clippy --workspace --all-targets --all-features are clean; cargo test --workspace --all-features passes
• All commits are signed off (check-dco.sh main..HEAD: OK)
• No AI attribution trailers (check-no-ai-attribution.sh main..HEAD: OK)
• CHANGELOG.md updated under [Unreleased] → Fixed
• Snapshot changes (if any) were reviewed with cargo insta review, not blind-accepted (none in this PR: no existing snapshot had a decorated blank run or a forbidden character)

[vyncint]

Thanks — both fixes are right. Verified beyond the tests: the decorated runs render correctly (I rasterised an underlined empty field and a struck-through "ab " with Inkscape; both lines draw where a terminal draws them, thanks to the existing xml:space="preserve"), and the SVG passes a real XML parser. The dependency-free well-formedness helper is fine. Its new PTY test was clean across a 15-shard stress run.