Save ouytput images to a temp director which will be erased

Also removed hardcoded temp path in benchmark itself.

benchmark/foundation/benchrunner.py

@@ -1,4 +1,5 @@
 import os
+import tempfile
 import time
 import subprocess
 
@@ -86,8 +87,10 @@ def benchmarkScene(ctx, scene):
     logger.DEBUG("File to use: {}" . format(blendfile))
     # Get command for rendering.
     # TODO(sergey): Create some temp folder.
+    tmpdir = tempfile.TemporaryDirectory(prefix="benchmark_")
     cfra = util.queryCurrentFrame(blendfile)
-    command = constructBenchmarkCommand(ctx, scene, blendfile, "/tmp/", cfra)
+    command = constructBenchmarkCommand(
+            ctx, scene, blendfile, tmpdir.name, cfra)
     logger.DEBUG("Command for rendering: {}" . format(command))
     progress.step('WARM_UP')
     logger.INFO("> Warm-up round, making sure everything is ready " +

benchmark/space/__init__.py

@@ -3,6 +3,7 @@ import json
 import os
 import platform
 import sys
+import tempfile
 from threading import Thread, Lock
 import blf
 import bpy
@@ -533,6 +534,7 @@ class BENCHMARK_OT_run(bpy.types.Operator):
     timer = None
     progress_provider = None
     logger_provider = None
+    tmpdir = None
 
     def setup_sink(self):
         self.progress_provider = ProgressProviderSink()
@@ -607,6 +609,7 @@ class BENCHMARK_OT_run(bpy.types.Operator):
         global_progress_status = None
         current_progress = 0.0
         context.area.tag_redraw()
+        self.tmpdir = None
 
     def modal(self, context, event):
         if event.type == 'TIMER':
@@ -632,6 +635,8 @@ class BENCHMARK_OT_run(bpy.types.Operator):
         compute_device = context.scene.compute_device
         device_type, device_name = compute_device.split(":")
 
+        self.tmpdir = tempfile.TemporaryDirectory(prefix="blender_benchmark_")
+
         # Before doing anything, make sure we have all sinks set up, so we do
         # not miss any progress report.
         self.setup_sink()
@@ -649,6 +654,7 @@ class BENCHMARK_OT_run(bpy.types.Operator):
         ctx.device_type = device_type
         ctx.device_name = device_name
         ctx.single_compute_scene = True
+        ctx.image_output_dir = self.tmpdir.name
         # Set this to True when having multiple GPUs of same name and only
         # one of the mis to be enabled. Or when requesting GPU render without
         # specifying GPU name.