mirror of
https://gitee.com/johng/gf
synced 2026-06-06 16:21:40 +08:00
a4b80e8680
## Summary Fix two bytea data corruption issues in the PostgreSQL driver: 1. **READ path** (fixes #4677): `CheckLocalTypeForField` and `ConvertValueForLocal` had no case for plain `bytea` type, causing it to fall through to the Core layer which incorrectly mapped it to `LocalTypeString`. Binary data was then converted to string via `gconv.String()`, corrupting the bytes on retrieval. 2. **WRITE path** (fixes #4231): `ConvertValueForField` applied PostgreSQL array syntax conversion (`[` → `{`, `]` → `}`) to all slice types including `[]byte` for bytea columns, corrupting bytes `0x5B` (`[`) and `0x5D` (`]`) on insertion. ## Changes - **`contrib/drivers/pgsql/pgsql_convert.go`**: - `CheckLocalTypeForField`: Add `case "bytea"` → `LocalTypeBytes` - `ConvertValueForLocal`: Add `case "bytea"` to preserve `[]byte` as-is - `ConvertValueForField`: Skip `[]`→`{}` replacement for `[]byte` with `bytea` field type - **`contrib/drivers/pgsql/pgsql_z_unit_convert_test.go`**: - Add unit tests for `bytea` type in `CheckLocalTypeForField`, `ConvertValueForLocal`, and `ConvertValueForField` - **`contrib/drivers/pgsql/pgsql_z_unit_issue_test.go`**: - Add `Test_Issue4677`: End-to-end round-trip test with various binary data (including 0x00, 0x5B, 0x5D, 0xFF) - Add `Test_Issue4231`: Targeted test for 0x5D byte corruption on write ## Test plan - [x] `Test_CheckLocalTypeForField` - bytea returns `LocalTypeBytes` - [x] `Test_ConvertValueForLocal` - bytea preserves `[]byte` as-is - [x] `Test_ConvertValueForField` - bytea skips array syntax replacement - [x] `Test_Issue4677` - full DB round-trip with binary data - [x] `Test_Issue4231` - write path preserves 0x5B/0x5D bytes - [x] Full pgsql test suite passes with no regressions closes #4677 closes #4231 ref #4689