Replace the external ArrayBuffer view and manual UTF-16 scan with native Koffi decoding through a pointer-width buffer. Preserve the original address for CoTaskMemFree.
Exercise production resultPath with real Koffi decoding of live UTF-16 buffers. Cover zero-low-byte BMP characters, surrogate pairs, NUL termination, and strings exceeding 32 KiB; test four- and eight-byte BigInt pointer preservation and the original release address. A negative-control mutation confirms truncating a 64-bit address fails.
Keep the bilingual Agent Note focused on implementation rationale and required memory ownership. Preserve master archive seals during rebase.
Co-authored-by: zhanglunet <[email protected]>
`applyEntryPatches` pushed a patch's `insert` rows into the entry list by
reference, so a later patch that inserted into a group an earlier patch had
inserted mutated the patch's own object. With bundle rows now inserted into
their group by a separate patch, the Loader's rollback of a rejected update
re-applied the previous include config over those mutated objects and hit
`duplicate loader entry id`; the rollback failed and the rejection reason
named it instead of the row that broke. Inserted rows are now cloned before
they join the list, as the function's detachment contract already stated;
the vendored change is logged and pinned by an application-twice test.
The publish guide still told bundle authors that mounted row ids carry the
package prefix and that `!!js` gates should compare names, a scheme this
stack withdrew: ids stay as declared, and a shared id or a repeated one
leaves the bundle out with the reason. It now says so, and names
`dsh.title` and `dsh.plugins` with a pointer to the manifest package that
declares every key. The cordis catalog follows app-boot's changed exports.
`stageOf` parsed the stage out of the Loader's `failed to <stage> loader
entry` message. The vendored wrapper is now an exported `EntryUpdateError`
carrying its `stage`, logged as a local modification, and the contained
group reads the field. `isJsDisabled` wrapped `isJsExpr` behind a literal
`disabled === true` comparison that already excluded expression nodes, so
it goes. The post-boot rejection guard, its documentation, and the
external-bundle module doc now say what they do: the process keeps running
with nothing stopped or attributed to a plugin, and a contained group lets
a failing row's siblings mount rather than failing the bundle as a whole.
The probe's child was a script string that wrote its report to stdout, and
the parent parsed the whole stream as one JSON value: a plugin that merely
printed a line at import could not be probed, and a package that kept a
timer alive after import ran until the timeout. The parent asserted the
parsed value as the report, and the cache read checked only `format`, so a
file holding nothing but `{ "format": 2 }` came back as a complete record.
The child is now its own module, `probe-child.ts`, run through tsx under a
source launch and bundled as `lib/probe-child.js`, and it sends one report
over an IPC channel; stdout is not read at all, and the child is killed as
soon as the report arrived. `parseChildReport` and `parseProbeRecord`
validate the report and the cached record field by field, so an
unrecognized report is a rejection and an unrecognized record is probed
again. `PluginProbe.ok` now states what it is: the main export imported and
cordis is not a second copy; `kind` and `addable[].ok` decide what can be
enabled or added.
`ProfileRuntime.recompose({ reloadBundles: true })` replaced its profile
before the Loader had accepted the update, so a rejected recomposition left
`current`, `layers`, and `originOf` describing a stack that never mounted.
The user patch watchers took a second path through the launcher's
`composeLive`, which wrote the new conflicts into the failure registry
before the update and kept composing from the runtime's stale profile.
Conflict records shared the registry with row failures under a fabricated
entry id and a group id that meant three different things, so the plugin
inventory attributed a conflict to the layer that won the id, and a bundle
disabled after failing kept its import and apply records forever.
The runtime now holds one committed composition — profile, stack ownership,
conflicts — and publishes a candidate only after `entry.update` resolves;
`originOf` reads the committed owners instead of recomputing them. The
watchers call `recompose` through a `reapply` callback, so every live change
goes through the one entry point. Conflicts live on the runtime with their
messages and never enter `pluginFailures`, whose records name rows that
reached the Loader; the inventory lists them under the layer that lost. A
contained group clears its rows' records when it unmounts.
`composeExternalLayer` inserted every id-targeted patch after all inserts,
so a bundle that replaced a group's config and then appended to it lost the
appended row under the `runtime` stage; two inserts into the same built-in
group produced two wrapper groups with one id, which the Loader rejected as
a duplicate and the boot failed; and an id a bundle inserted twice was
indexed once, escaped the ownership check, and failed the boot the same way.
The composition now inserts the bundle's group empty and emits the bundle's
patches in the order written, re-targeting root inserts into the group and
gathering every insert into one built-in group in one wrapper. A repeated id
is a conflict of the bundle with itself: the bundle is left out whole and
recorded like any other conflict, and a built-in layer that repeats an id
fails loud. `composeProfileStack` renders each contained layer once and
returns the owner of every id with the patches and conflicts, and one walker
in `patch-rows.ts` replaces the five recursive scans of inserted rows.
The client namespace service keeps install and remove for its own members
and refuses a Remote method of that name at mount, so the browser could
never reach plugins/install. A spec now reads the reserved names off the
gateway and checks every @Remote name against them.