fix(viewer): keep unbalanced markdown emphasis literal in PDF reports

[apetcu]

Summary

Generating the PDF report crashed with ValueError: Parse error: saw </b> instead of expected </i> whenever a finding contained a run of asterisks — typically a masked secret such as '******' from Spring Boot's /actuator/env sanitization. The <b><i></b></i> seen in the issue is what _inline_md produced from ****** (literal <b> tags in finding text were already escaped and are fine); ReportLab rejects the crossed tags, so POST /api/report/send failed for the whole report.

Fixes #1171

Root cause

_inline_md applied the **…**, __…__ and *…* substitutions with .+? content, so one pass could open a tag inside another pass's span and the passes interleaved into crossed markup. Any unbalanced emphasis (***x***, **a *b** c*) triggered the same crash.

Changes

• _inline_md: emphasis content may no longer contain its own delimiter or a tag ([^*<>\n] / [^_<>\n]), so the passes cannot cross; ***x*** is handled explicitly as bold-italic; a bare run of asterisks stays literal.
• _para: builds the Paragraph and, if ReportLab still rejects the markup, retries with our <b>/<i>/<font> tags stripped so the text renders instead of failing the whole report. Used for prose, bullets and headings in _markdown_flowables.
• Tests: parametrized crossed-markup cases for _inline_md, _markdown_flowables with unbalanced emphasis, an end-to-end generate_report_pdf with a masked secret in the summary, and the _para fallback. The new tests fail on main with the traceback from the issue.

Verification

• uv run pytest tests/test_report_pdf.py tests/test_fenced_code.py — 25 passed (8 new)
• uv run ruff format --check, uv run ruff check, uv run mypy strix/interface/viewer/report_pdf.py — clean

---

Prepared with AI assistance (Claude Code); reviewed and tested locally by @apetcu.

[greptile-apps]

Greptile Summary

The PR prevents malformed Markdown emphasis from producing crossed ReportLab tags and adds a plain-text fallback for rejected paragraph markup.

• Adds explicit bold-italic parsing and restricts emphasis spans to avoid delimiter interleaving.
• Routes Markdown prose, bullets, and headings through a paragraph-construction fallback.
• Adds regression coverage for masked secrets, unbalanced emphasis, fallback rendering, and end-to-end PDF generation.

Confidence Score: 4/5

The balanced nested-emphasis regression should be fixed before merging because valid narrative Markdown can now render visibly incorrect text in PDF reports.

The delimiter-excluding expressions prevent outer emphasis from matching balanced nested spans, leaving raw Markdown markers in reachable report narratives that previously rendered with nested formatting.

Files Needing Attention: strix/interface/viewer/report_pdf.py, tests/test_report_pdf.py

Important Files Changed

Filename	Overview
strix/interface/viewer/report_pdf.py	Prevents crossed emphasis tags and adds paragraph fallback handling, but the narrowed patterns regress balanced nested Markdown emphasis.
tests/test_report_pdf.py	Adds strong crash-regression coverage, though it does not assert that previously supported balanced nested emphasis remains formatted.

Prompt To Fix All With AI

### Issue 1
strix/interface/viewer/report_pdf.py:424-427
**Nested emphasis renders literally**

When a report narrative contains balanced nested Markdown such as `**bold with *italic* inside**`, the new delimiter-excluding patterns cannot match the outer span, causing raw emphasis delimiters and incomplete formatting to appear in the generated PDF.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

_{Reviews (1): Last reviewed commit: "fix(viewer): keep unbalanced markdown em..." | Re-trigger Greptile}

[greptile-apps]

Nested emphasis renders literally

When a report narrative contains balanced nested Markdown such as **bold with *italic* inside**, the new delimiter-excluding patterns cannot match the outer span, causing raw emphasis delimiters and incomplete formatting to appear in the generated PDF.

Knowledge Base Used: Findings, usage, and reporting

Prompt To Fix With AI

This is a comment left during a code review.
Path: strix/interface/viewer/report_pdf.py
Line: 424-427

Comment:
**Nested emphasis renders literally**

When a report narrative contains balanced nested Markdown such as `**bold with *italic* inside**`, the new delimiter-excluding patterns cannot match the outer span, causing raw emphasis delimiters and incomplete formatting to appear in the generated PDF.

**Knowledge Base Used:** [Findings, usage, and reporting](https://app.greptile.com/strix-org-3/-/custom-context/knowledge-base/usestrix/strix/-/docs/findings-and-reporting.md)

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

[apetcu]

Good catch — balanced nesting had regressed. Addressed in bc1b0b7: bold spans now only require a non-delimiter at each end (so a masked ****** still stays literal) and italic content may contain complete <b>..</b> spans, so **bold with *italic* inside** and *outer **bold** inner* render correctly again; covered by test_inline_md_keeps_balanced_nested_emphasis.

[bearsyankees]

Thanks for the pr @apetcu -- triaging :)