fix(deps): the second pull forget to set wasm_hash_download in pulled…

….json (#3674)

* add e2e test to reveal the issue

* fix(deps): the second pull forget to set wasm_hash_download in pulled.json

* changelog

CHANGELOG.md

@@ -80,6 +80,12 @@ dfx.json.  This is now performed by dfxvm.
 When all the arguments are optional, dfx automatically provides a `null` value when no arguments are provided.
 `--always-assist` flag enables the candid assist feature for optional arguments, instead of providing a default `null` value.
 
+### fix(deps): the second pull forget to set wasm_hash_download in pulled.json
+
+When the dependency has been in the cache, `dfx deps pull` forgot to set correct `wasm_hash_download` in `pulled.json`.
+
+It caused the following `init/deploy` commands to fail.
+
 ## Dependencies
 
 ### Replica

e2e/tests-dfx/deps.bash

@@ -656,3 +656,22 @@ Installing canister: $CANISTER_ID_C (dep_c)"
 
   assert_directory_not_exists "deps"
 }
+
+@test "dfx deps pull can set correct pulled.json when the dependency is already in cache" {
+  use_test_specific_cache_root # dfx deps pull will download files to cache
+
+  # start a "mainnet" replica which host the onchain canisters
+  dfx_start
+
+  setup_onchain
+
+  # pull canisters in app project and the dependencies are cached
+  cd app
+  assert_command dfx deps pull --network local
+
+  # the second pull should be able to set the correct pulled.json
+  assert_command dfx deps pull --network local
+
+  # this command will fail if the pulled.json is not correct
+  assert_command dfx deps init
+}

src/dfx/src/commands/deps/pull.rs

@@ -165,6 +165,7 @@ async fn download_and_generate_pulled_canister(
             if hash_cache.as_slice() == hash_on_chain {
                 cache_hit = true;
                 pulled_canister.gzip = gzip;
+                pulled_canister.wasm_hash_download = hex::encode(hash_cache);
                 trace!(logger, "The canister wasm was found in the cache.");
             }
             break;