[SwastikTripathi]

htmlToText only adds line breaks around the block tags in BLOCK_TAGS_REGEX, and that list stops at HTML 4. Sibling HTML5 sections were glued together, which is the merging of distinct text that the function's docs say it tries to avoid:

HTML	Before	After
<section>Call 555-1234</section><section>9876 Main St</section>	Call 555-12349876 Main St	Call 555-1234\n9876 Main St
<header>Head</header><main>Body</main><footer>Foot</footer>	HeadBodyFoot	Head\nBody\nFoot
A<hr>B	AB	A\nB
<dl><dt>Term</dt><dd>Def</dd></dl>	TermDef	Term\nDef
<details><summary>Sum</summary>Body</details>	SumBody	Sum\nBody
<tr><th>H1</th><th>H2</th></tr>	H1H2	H1\tH2\t

The first row shows the practical effect: social.parseHandlesFromHtml runs on this text, so a phone number at the end of one <section> runs into the next section's digits.

The list now also has the HTML5 sectioning and grouping elements a browser lays out as blocks (article, aside, section, nav, header, footer, main, hgroup, figure, figcaption, details, summary, dialog, menu), plus hr, dt, dd, legend and caption. <th> cells get the same tab separator as <td>.

Added a test with those cases to test/utils/cheerio.test.ts. None of the new tags appear in the existing inputs (including html-to-text-test-data.ts and the social tests), so their expected output doesn't change.

Verification: I couldn't run the repo's vitest suite here. I ran the new test's cases, plus several existing cases, through the published @crawlee/utils build, once as released and once with the same two-line change applied to its compiled cheerio.js (the block-tag and td code is the same as on master). As released, all 7 new cases fail. With the change, all 7 pass, both bare and wrapped in a full <html><body> document as checkHtmlToText does, and the existing cases I retyped give the same output as before. oxfmt --check passes on both files.

🤖 Generated with Claude Code

[barjin]

lgtm, thank you @SwastikTripathi !

[barjin]

lgtm, thank you @SwastikTripathi !