Skip to content

@marko/runtime-tags@6.3.17

Choose a tag to compare

View all tags
@github-actions github-actions released this 22 Jul 16:41
@marko/runtime-tags@6.3.17
090d516

Patch Changes

  • #3527 be7dea6 Thanks @DylanPiercey! - Give a cause-specific error when two-way binding (value:=x) to a value from array destructuring, explaining that it has no change handler and suggesting object destructuring or an explicit Change attribute, instead of the generic "Unable to bind to value."

  • #3523 d0e32de Thanks @DylanPiercey! - Emit the bare value for a text-only tag or html-comment whose body is a lone dynamic interpolation (<title>${x}</title>), letting the text sink coerce it instead of generating a redundant ${_to_text(x)} template wrapper.

  • #3531 74490b6 Thanks @DylanPiercey! - Fix a controlled <select> with a void (undefined/null) value selecting a different <option> on the server than the client: SSR now normalizes a void value to "" and marks a value="" option selected, matching DOM instead of letting the browser auto-pick the first option.

  • #3407 bba9259 Thanks @DylanPiercey! - Fix a client render where a rejected <await> or a failed lazy load under an ancestor <try @placeholder> left the placeholder shown forever. The failure path now completes the placeholder's await counter (unless it is a placeholder-less or resumed reorder counter), so the placeholder is dismissed and the caught content renders.

  • #3522 a236904 Thanks @DylanPiercey! - Emit a single Class-API interop registration per renderer, instead of one for each tag occurrence, trimming the redundant scriptlets when the same class component is used multiple times in a template.

  • #3524 e6d462a Thanks @DylanPiercey! - Stop emitting a dead scope binding, walk slot, and resume marker for a content attribute on a native tag that also has body content — codegen already drops the attribute (body wins), so analyze now ignores it too.

  • #3517 5025913 Thanks @DylanPiercey! - Fix a hydration mismatch for a dynamic tag with positional args and a fallback body. When the tag resolves to null (or a string), <${tag}(1, 2)>Fallback</> rendered nothing on the server but the fallback body on the client; the server now renders the fallback too.

  • #3519 b19d9dc Thanks @DylanPiercey! - Preserve tag-name case when escaping </style inside <style> raw text, so an author-written </STYLE> in a content:/url(...)/custom-property value is no longer lowercased.

  • #3525 3446c8e Thanks @DylanPiercey! - Flatten a text-only conditional written with the <else if=cond> space syntax into a single placeholder ternary, matching the <else-if=cond> spelling — both now skip the _if runtime, branch scopes, and serialize guards.

  • #3526 2190723 Thanks @DylanPiercey! - Stop shipping a dead _closure_get pending-resume id in DOM output when a closure's subscriber section is not under a <try> @placeholder (only a sibling closure is) — the id had no matching HTML-side registration to look it up.

  • #3518 58bb403 Thanks @DylanPiercey! - Keep </html> in document order when the tag carries a resume marker (e.g. <html lang=input.lang> on a stateful page), instead of emitting it before the resume scripts.

  • #3521 36cf307 Thanks @DylanPiercey! - Speed up in-place keyed list updates. The <for> reconciler rebuilt an O(n) key→scope lookup map on every update even when nothing had moved. It now reuses the branch at each position directly and only builds the map — from the remaining unmatched scopes — once a key first lands out of order.

  • #3535 04e289c Thanks @DylanPiercey! - Restore a two-way-bound radio group (checkedValue:=) to its default member on a native form reset, instead of clobbering the bound value to undefined. On reset the fired handler now adopts the group's reset-restored default rather than the firing (unchecked) radio's value.

  • #3510 c56de42 Thanks @DylanPiercey! - Escape NUL and lone surrogates in a serialized RegExp source. RegExp.source leaves a raw NUL or lone surrogate in the pattern unescaped, and the resume payload embeds the /.../ literal directly into a <script>, where a NUL is folded to U+FFFD and a lone surrogate is unencodable in UTF-8 — silently corrupting the resumed regex (an SSR/CSR mismatch). Such characters are now emitted as \x00 / \uXXXX escapes; paired surrogates and normal escapes are unaffected.

  • #3529 f54e09a Thanks @DylanPiercey! - Shrink the <for> reconciler by dropping a redundant map. A keyed list update already builds a key→scope lookup; it now stashes each scope's old index on that same pass, so the longest-increasing-subsequence step that plans moves no longer allocates and fills a second scope→position Map. Same behavior, one fewer allocation per reordering update, and a slightly smaller runtime.

  • #3513 118c4b6 Thanks @DylanPiercey! - Fix an SSR/CSR mismatch for a <textarea> with a non-string initial value. SSR rendered the value as text content (_escape), while the client set defaultValue via the value= attribute normalization, so true/NaN/0n rendered differently after a client render/navigation than on the server (e.g. <textarea value=computeTotal()> yielding NaN). SSR now coerces a textarea's value the same way — matching the client and <input> — so both agree. String values are unaffected, and the client runtime is unchanged.

  • Updated dependencies [0187289]:

    • @marko/compiler@5.41.4
Assets 2
Loading