Add litsupport module to collect QEMU plugin metrics

litsupport/CMakeLists.txt

@@ -15,6 +15,7 @@ set(LITSUPPORT_FILES
   modules/microbenchmark.py
   modules/perf.py
   modules/profilegen.py
+  modules/qemulog.py
   modules/remote.py
   modules/run.py
   modules/run_under.py

litsupport/modules/qemulog.py

@@ -0,0 +1,51 @@
+"""test-suite/lit plugin to collect metrics from QEMU logs.
+
+If you are running the test suite under QEMU and are generating logs
+that contain metrics (typically from a plugin), you can use this
+module to:
+
+1. save the logs to a file
+2. parse any colon (:) separated metrics and collect them in the
+metrics
+
+For example, the example plugin libinsn.so outputs dynamic instruction
+counts in the following format:
+
+cpu 0 insns: 42
+total insns: 1024
+
+If you enable the plugin and tell QEMU to log it:
+
+-DTEST_SUITE_RUN_UNDER='qemu -plugin libinsn.so -d plugin'
+-DTEST_SUITE_EXTRA_LIT_MODULES=qemulog
+
+You can then compare those instruction counts later:
+
+./compare.py results.json -m 'total insns'
+
+"""
+
+from litsupport import shellcommand
+from litsupport import testplan
+from litsupport.modules import run_under
+
+
+def _mutateCommandLine(context, commandline):
+    context.qemulog = context.tmpBase + ".qemulog"
+    cmd = shellcommand.parse(commandline)
+    cmd.envvars.update({"QEMU_LOG_FILENAME": context.qemulog})
+    return cmd.toCommandline()
+
+
+def _getOutput(context):
+    with open(context.qemulog, "r") as f:
+        result = dict()
+        for line in f.read().splitlines():
+            [name, count] = line.split(":")
+            result[name] = int(count)
+        return result
+
+
+def mutatePlan(context, plan):
+    plan.runscript = testplan.mutateScript(context, plan.runscript, _mutateCommandLine)
+    plan.metric_collectors.append(_getOutput)