solidjs / solid / 35838850243

73%

main: 89%

push

github

fix(compiler): scope every hole that can build JSX, not only props.children (#3567) (#3599)

JSX passed through a prop other than `children` (a layout's `header`, an
object of slots, a context getter) hydrated with permuted keys whenever an
id-allocating hole followed it in the same template. Both compilers' hole-
scope predicate treated only `.children` member reads as able to build
hydratable content, so `{props.header}` compiled to a transparent insert:
the server reserved the scoped sibling's slot at registration (ssr-arg
evaluation) and resolved the header thunk later inside the `ssr()` walk,
while the client allocated in statement order. Same id count, permuted —
the server elements stayed unclaimed and the handler bound to a detached
copy.

The predicate is now a denylist: a dynamic hole takes a scope unless its
value is provably a primitive (literals, template literals, unary, binary,
update and private-`in` expressions). Property reads of any name, optional
calls, the left operand of `||`/`??`, sequence and assignment expressions,
and array literals with hydratable entries are scoped, identically in the
Babel plugin and the native compiler. The Babel dom generate also
classified the hole after `transformNode` had rewritten `a && b` into a
memo ternary; it now classifies the source expression like the ssr
generate. Cost: a member-read text hole reserves one id slot on both sides,
the same as `{count()}` always has.

Verification: seven parity-harness scenarios, compiler pins and Babel-vs-
Oxc parity probes, plus a 22-scenario server-render → hydrate regression
pair (test/harness/slot-hydration-3567.tsx) that asserts zero hydration
warnings, every server `_hk` element claimed, and that a click on the
server-rendered button reaches the handler. Known remaining gap, pinned as
`test.fails`: a bare identifier bound to a function returning JSX is still
unscoped on both sides and misaligns when a scoped hole follows; a dev-only
"unscoped hole moved t... (continued)

592 of 865 branches covered (68.44%)

Branch coverage included in aggregate %.

873 of 1121 relevant lines covered (77.88%)

19.75 hits per line