[hejsztynx]

Summary

Closes #811

This PR addresses a couple of iOS issues:

• we had an undesirable html length check, which caused a string below the length of 13 to not be parsed as HTML, e.g. <p>a</p>
• if you passed </body>123<body>, where the closing and opening body tags are swapped, the whole app crashed
• in case of an exception during parsing, we now catch that exception and fall back to a plain text insertion. Such approach was already used in EnrichedText parser, but not in the input one

Aside from that, it introduces a modification into our normalizer flow. Now it's consistent across all three platforms and works as follows:

• if the value is wrapped with <html></html> tags, we strip them and extract html from the <body> tag, if there is one
• now we look at the useHtmlNormalizer flag, and we normalize the current html if needed
• at last, we try to parse our processed html

Previously the main issue was that even with useHtmlNormalizer set to true, we didn't normalize the html when it was wrapped with <html> tags - we want to always normalize it, as <html> tags are not a guarantee of any sort that the structure is ours.

This PR also introduces one fix, caught by an android regression test. Empty <codeblock></codeblock> was normalized into <codeblock><p></p></codeblock>, which is something that android's parser does not handle correctly, so normalizer was adjusted - <p></p> is not emitted when empty. The iOS behavior effectively has not changed in this manner.

I've run all the e2e tests, they pass.

Test Plan

Try out the cases I listed above on iOS, with both EnrichedTextInput and EnrichedText components. Run the e2e tests on android/ios, experiment with html parsing, test the html round-trip with useHtmlNormalizer set to true and false.

Screenshots / Videos

iOS bugs that were present:

Screen.Recording.2026-10-14.at.17.02.45.mov

Screen.Recording.2026-10-14.at.17.03.58.mov

Compatibility

OS	Implemented
iOS	✅
Android	✅
Web	✅

Checklist

• E2E tests are passing
• Required E2E tests have been added (if applicable)

[Copilot]

🟡 Changes recommended

The fixed failure modes lack regression coverage in the existing iOS Maestro suite.

1 open finding

• Add iOS Maestro coverage for short and reversed-body HTML parsing · New

What changed in this PR

Improves iOS HTML preprocessing resilience and fallback behavior.

Changes:

• Supports short HTML strings.
• Safely handles reversed body tags.
• Catches preprocessing failures and preserves raw input.

File	Description
ios/​inputHtmlParser/​InputHtmlParser.mm	Adds exception handling.
ios/​htmlParser/​HtmlParser.mm	Fixes length and body-range parsing.
ios/​EnrichedTextInputView.mm	Guards failed preprocessing.

🧠 Review effort: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🟡 Changes recommended

Unresolved crash and content-corruption risks, together with inconsistent iOS fragment parsing, block approval.

5 open issues

• Skip HTML normalization when clipboard input is nil · New
• Avoid raw body substring matching in Android HTML extraction · New
• Parse unwrapped fragments when HTML normalization is disabled · New
• Escape decoded attribute values in Gumbo serialization · New
• Use HTML-aware body boundaries to avoid matching comments · New

1 resolved since last review

• Add iOS Maestro coverage for short and reversed-body HTML parsing

_{Review effort: Balanced}