Enhancement/505760246/apk integration

AGENTS.md

@@ -107,3 +107,55 @@ python butler.py format
 
 This will format the changed code in your current branch.
 It's possible to get into a state where linting and formatting contradict each other. In this case, STOP, the human will fix it.
+
+## Remote Swarming Development & Validation Workflow
+
+This section outlines the step-by-step cycle for developing, deploying, launching, and verifying changes on remote Swarming bots.
+
+### Step 1: Local Development & Formatting
+Make your code changes in your feature branch. Always run formatting and linting locally before committing:
+```bash
+pipenv run python butler.py format
+pipenv run python butler.py lint
+```
+
+### Step 2: Push & Deploy to Remote `dev`
+For changes to run on remote Swarming bots, they must be committed, merged, and pushed to the remote **`dev`** branch:
+1. Commit and push your feature branch:
+   ```bash
+   git add <modified_files>
+   git commit -m "Your description"
+   git push origin <your-feature-branch>
+   ```
+2. Merge into `dev` and push:
+   ```bash
+   git checkout dev
+   git pull origin dev
+   git merge <your-feature-branch>
+   git push origin dev
+   git checkout <your-feature-branch>
+   ```
+3. **⚠️ Crucial Rebuild Wait Time**: After pushing to `dev`, you **MUST wait 25 to 30 minutes** before triggering any Swarming tasks. This gives the remote Google Cloud Storage (GCS) builder enough time to pull your new commit, compile the binaries, and package them into the deployment ZIP bundle (`linux-3.zip`) fetched by the bots.
+
+### Step 3: Preprocess & Launch the Swarming Task
+Once the deployment package has finished rebuilding on GCS:
+1. Trigger the preprocess pipeline and launch a new Swarming task:
+   ```bash
+   python3 scratch/preprocess_and_launch.py
+   ```
+2. Note the generated **Swarming Task ID** (e.g. `791f445b26114a10`) and the task URL printed in the stdout.
+
+### Step 4: Live Monitoring & Log Retrieval
+1. **Live Monitor**: Open `scratch/monitor_swarming_task.py`, update `task_id` with your new Task ID, and run the script to stream the live console output:
+   ```bash
+   python3 scratch/monitor_swarming_task.py
+   ```
+2. **Download High-Resolution Logs**: Once the task terminates:
+   * Look at the live monitor output to identify the **assigned Bot Name** (e.g. `lin-192-g582`) and the final state (`COMPLETED` or `BOT_DIED`).
+   * Open `scratch/read_logs.py`, update the `bot_name` and adjust the time filter window (e.g. `timestamp >= "YYYY-MM-DDTHH:MM:SSZ"`), then run:
+     ```bash
+     pipenv run python scratch/read_logs.py
+     ```
+   * This downloads all high-resolution Stackdriver bot and logcat streams into `scratch/bot_logs.txt`.
+3. **Analyze**: Inspect `scratch/bot_logs.txt` using grep/editors to verify your changes (such as JNI prints or fuzzer loop outputs).
+

src/clusterfuzz/_internal/batch/service.py

@@ -256,7 +256,8 @@ def is_remote_task(command: str, job_name: str) -> bool:
   """
   try:
     _get_specs_from_config(
-        [remote_task_types.RemoteTask(command, job_name, None)])
+        [remote_task_types.RemoteTask(command, job_name, None)],
+        should_check_regions=False)
     return True
   except ValueError:
     return False
@@ -300,15 +301,15 @@ def _get_config_names(batch_tasks: List[remote_task_types.RemoteTask]):
   return config_map
 
 
-def _get_subconfig(batch_config, instance_spec):
+def _get_subconfig(batch_config, instance_spec, should_check_regions=True):
   all_subconfigs = batch_config.get('subconfigs', {})
   instance_subconfigs = instance_spec['subconfigs']
 
   queue_check_regions = batch_config.get('queue_check_regions')
-  if not queue_check_regions:
-    logs.info(
-        'Skipping batch load check because queue_check_regions is not configured.'
-    )
+  if not should_check_regions or not queue_check_regions:
+    logs.info('Skipping batch load check. '
+              f'should_check_regions: {should_check_regions}, '
+              f'queue_check_regions: {queue_check_regions}')
     weighted_subconfigs = [
         WeightedSubconfig(subconfig['name'], subconfig['weight'])
         for subconfig in instance_subconfigs
@@ -343,8 +344,8 @@ def _get_subconfig(batch_config, instance_spec):
   return all_subconfigs[chosen_name]
 
 
-def _get_specs_from_config(
-    batch_tasks: List[remote_task_types.RemoteTask]) -> Dict:
+def _get_specs_from_config(batch_tasks: List[remote_task_types.RemoteTask],
+                           should_check_regions: bool = True) -> Dict:
   """Gets the configured specifications for a batch workload."""
   if not batch_tasks:
     return {}
@@ -382,7 +383,8 @@ def _get_specs_from_config(
     # This saves us time and reduces fragementation, e.g. every linux fuzz task
     # run in this call will run in the same zone.
     if config_name not in subconfig_map:
-      subconfig = _get_subconfig(batch_config, instance_spec)
+      subconfig = _get_subconfig(batch_config, instance_spec,
+                                 should_check_regions)
       subconfig_map[config_name] = subconfig
 
     should_retry = instance_spec.get('retry', False)

src/clusterfuzz/_internal/bot/fuzzers/engine_common.py

@@ -290,7 +290,9 @@ def find_fuzzer_path(build_directory, fuzzer_name):
   for root, _, files in shell.walk(build_directory):
     for filename in files:
       if (legacy_name_prefix + filename == fuzzer_name or
-          filename == fuzzer_filename):
+          filename == fuzzer_filename or
+          (filename.endswith('.apk') and fuzzer_name in filename) or
+          filename.endswith('.apk')):
         return os.path.join(root, filename)
 
   # This is an expected case when doing regression testing with old builds