PDGGK opened a new pull request, #401: URL: https://github.com/apache/dubbo-go-hessian2/pull/401
**What this PR does**: When a Java provider returns an object with a `Map` field that holds a `LinkedHashMap` (or any other typed map), and that map contains further typed maps, decoding the response into a registered Go struct fails. On current master the error is `the type ref index 0 is out of range` (or index 1, depending on what came before), and the whole response is lost. The Java encoder writes each type name once and refers to it by index afterwards. `decMapByValue`, which decodes a map into a struct field, read the type of a typed map (`'M'`) and threw it away, so it never reached the decoder's type refs. From then on the Go side was one entry behind the Java side, and the next type ref pointed past the end of the list. `decMap` already reads the type with `decMapType`, which records it. `decMapByValue` now does the same. The type itself is still not used there. The byte dump in #351 shows exactly this shape: `parkingService` is written as `M 0x17 "java.util.LinkedHashMap"` and `storeService` as `M 0x91`, a ref to type index 1. The new test `customReplyLinkedHashMapInTreeMapField` makes hessian-lite produce the same bytes (a `TreeMap` field, then two nested `LinkedHashMap`s), and master fails on it with `the type ref index 1 is out of range`. The error text in the issue (`illegal class index @idx 1`) comes from the older release the reporter used, and their dump is truncated, so I could not replay their exact payload. Tests: `TestDecodeLinkedHashMapInMapField` decodes two replies produced by the Java test server: a `LinkedHashMap` field holding two `LinkedHashMap`s (nested maps are type ref 0) and a `TreeMap` field holding two `LinkedHashMap`s (the case from #351). With `map.go` reverted to master, both cases fail with the out-of-range error above. With the change both decode to the expected struct. `go test ./... -race`, `mvn test -f test_hessian/pom.xml`, `golangci-lint run` (v2.4.0), `go fmt` plus `imports-formatter`, and the license header check all pass locally, with JDK 8 and Go 1.25. I did not test against a running Dubbo provider. The replies come from the hessian-lite `Hessian2Output` in `test_hessian`, the same encoder Dubbo uses. **Which issue(s) this PR fixes**: Fixes #351 **Special notes for your reviewer**: The only caller of `decMapByValue` is `decToDest`, used by `decInstance` for struct fields. A type ref that is out of range for the field map itself now returns an error, as it already did in `decMap`, instead of being skipped silently. **Does this PR introduce a user-facing change?**: ```release-note Fix decoding of a struct whose map field contains nested typed maps (for example LinkedHashMap): it failed with "the type ref index N is out of range". ``` -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
