## 1.0.3 - fix: Bypass the dispatcher and call the real Flash operations directly when the stack resides in internal DRAM. This prevents deadlocks that could arise from nested calls to flash operation interfaces (e.g. specifically, calling a flash operation from within a flash operation already executing under the scheduler) and avoids unnecessary task scheduling overhead. ## 1.0.2 - fix: Bypass the dispatcher and call the real Flash operations directly when it is not initialized yet, so early-boot Flash access (e.g. partition loading and core dump probing during system startup). ## 1.0.1 - feat: Added `spi_flash_mmap` / `spi_flash_munmap` to the set of Flash APIs intercepted by the dispatcher, so memory-mapping can also be safely invoked from tasks whose stacks live in PSRAM. - fix: Serialize dispatcher requests with an internal mutex so the shared request/result slot is always exclusively owned by a single caller, even when wrappers are called concurrently from different tasks. - fix: Fix on ESP32 the allocator could place these objects into IRAM, causing a `LoadStoreError` on byte accesses. - deprecate: `esp_flash_dispatcher_config_t::queue_size` is kept for backward compatibility only and no longer affects behavior; the dispatcher now uses a single shared slot regardless of the configured value. ## 1.0.0 - Initial release of `esp_flash_dispatcher`. Intercepts `esp_flash_read`, `esp_flash_write`, `esp_flash_write_encrypted`, `esp_flash_erase_region`, and `esp_flash_erase_chip`, and executes them on a dedicated background task whose stack lives in internal RAM, so application tasks can keep their stacks in PSRAM without breaking Flash operations.