Reproduction link or steps
https://repl.rolldown.rs/#eNplkMFugzAMhl/FyiWgQbpekdhL9MqFgdmYgjM5oSqivPsSomwdvcWx/fv7/1UMolrFSD3elLPhTaL6qwvRhXL6NuxgBbtMsMHAZgKpTp0ekVxpSC/qy8qGGsLbPtnj0M7awTBT50ZD4Od4gSyHtSEARjczgbTIV+Ryb5ZpR8ILXByP9JH5c3lD24NwZ8g6YDM7tFBHNXmSVZCu3yCCZp5tHwlUeeEVvA8UleMZt0KkVjL7W0ezBwv7CW3aPjI9XzqkkCv3iZRxGGKVVO534EjiAf5vJIzDb4Q5GA/513BZpnejM/mwIX1OT+nL9Ij/MqYwtNri5jGuXv+sXtVZbD/8MK5d
What is expected?
I expect that the lazy import has the (r) => r.default || r to be rewritten to point to the correct named export
What is actually happening?
client-only.js is merged into index.js. The dynamic-import target gets rewritten from ./client-only.js to ./index.js, but the .then(r => r.default || r) accessor is left as-is, so it reads index.js's default export (the entry function):
System Info
- rolldown: 1.0.1
- node: 22.x
- repro reproduces on macOS arm64 and Linux arm64
Any additional comments?
this is another variant of the #9305 dedupe regression that's distinct from the ones already filed, so opening a separate issue for it.
cross-linking:
in particular I think this comment is the same issue, but not on the right issue....
discovered becauase we hit this in upgrading to vite@8.0.13 in nuxt/nuxt#35095
Reproduction link or steps
https://repl.rolldown.rs/#eNplkMFugzAMhl/FyiWgQbpekdhL9MqFgdmYgjM5oSqivPsSomwdvcWx/fv7/1UMolrFSD3elLPhTaL6qwvRhXL6NuxgBbtMsMHAZgKpTp0ekVxpSC/qy8qGGsLbPtnj0M7awTBT50ZD4Od4gSyHtSEARjczgbTIV+Ryb5ZpR8ILXByP9JH5c3lD24NwZ8g6YDM7tFBHNXmSVZCu3yCCZp5tHwlUeeEVvA8UleMZt0KkVjL7W0ezBwv7CW3aPjI9XzqkkCv3iZRxGGKVVO534EjiAf5vJIzDb4Q5GA/513BZpnejM/mwIX1OT+nL9Ij/MqYwtNri5jGuXv+sXtVZbD/8MK5d
What is expected?
I expect that the lazy import has the
(r) => r.default || rto be rewritten to point to the correct named exportWhat is actually happening?
client-only.jsis merged intoindex.js. The dynamic-import target gets rewritten from./client-only.jsto./index.js, but the.then(r => r.default || r)accessor is left as-is, so it readsindex.js'sdefaultexport (theentryfunction):System Info
Any additional comments?
this is another variant of the #9305 dedupe regression that's distinct from the ones already filed, so opening a separate issue for it.
cross-linking:
in particular I think this comment is the same issue, but not on the right issue....
discovered becauase we hit this in upgrading to
vite@8.0.13in nuxt/nuxt#35095