[Findit] Fetch DEPS from buildspec/ instead of trunk for chrome official builds.

appengine/findit/crash/fracas.py

Index: appengine/findit/crash/fracas.py
diff --git a/appengine/findit/crash/fracas.py b/appengine/findit/crash/fracas.py
index 90600dcee03b9482fa1e194de1144eabf7b5ed11..0e005aeb58d682c1b1d17ba1ad315b8daa0cf0c5 100644
--- a/appengine/findit/crash/fracas.py
+++ b/appengine/findit/crash/fracas.py
@@ -4,21 +4,23 @@
 
 """Process crashes from Chrome crash server and find culprits for them."""
 
+from common import chromium_deps
+from crash import detect_regression_range
+from crash import findit_for_crash
+from crash.fracas_parser import FracasParser
 
-def FindCulpritForChromeCrash(  # Not implemented yet pylint: disable=W0613
-    channel, platform, signature, stack_trace,
-    crashed_version, versions_to_cpm):
+
+def FindCulpritForChromeCrash(signature, platform,
+                              stack_trace, crashed_version, historic_metadata):
   """Finds culprits for a Chrome crash.
 
   Args:
-    channel (str): The channel name, could be 'dev', 'canary', 'beta', etc.
     platform (str): The platform name, could be 'win', 'mac', 'linux',
-        'android', 'ios', etc.
+      'android', 'ios', etc.
     signature (str): The signature of a crash on the Chrome crash server.
     stack_trace (str): A string containing the stack trace of a crash.
     crash_version (str): The version of Chrome in which the crash occurred.
-    versions_to_cpm (dict): Mapping from Chrome version to crash per million
-        page loads.
+    historic_metadata (dict): Mapping from Chrome version to historic metadata.

[stgao, 2016/05/05 17:32:28]

Is it still a dict?

[Sharu Jiang, 2016/05/05 20:26:45]

Done.

 
   Returns:
     (analysis_result_dict, tag_dict)
@@ -34,7 +36,6 @@ def FindCulpritForChromeCrash(  # Not implemented yet pylint: disable=W0613
             "revision": "commit-hash",
             "code_review_url": "https://codereview.chromium.org/ISSUE",
             "project_path": "src/v8",
-            "project_name": "v8",
             "author": "who@chromium.org",
             "time": "2015-08-17 03:38:16",  # When the revision was committed.
             "reason": "A plain string with '\n' as line break to explain why",
@@ -48,22 +49,50 @@ def FindCulpritForChromeCrash(  # Not implemented yet pylint: disable=W0613
     The tag dict are allowed key/value pairs to tag the analysis result for
     query and monitoring purpose on Findit side. For allowed keys, please
     refer to crash_analysis.py and fracas_crash_analysis.py:
-      For results with normal culprit-finding algorithm:
-        {
+      For results with normal culprit-finding algorithm: {
           'found_suspects': True,
           'has_regression_range': True,
           'solution': 'core_algorithm',
-        }
-      For results using git blame without a regression range:
-        {
+      }
+      For results using git blame without a regression range: {
           'found_suspects': True,
           'has_regression_range': False,
           'solution': 'blame',
-        }
-      If nothing is found:
-        {
+      }
+      If nothing is found: {
           'found_suspects': False,
-        }
+      }
   """
-  # TODO (katesonia): hook the analysis logic up here.
-  return {'found': False}, {'found_suspects': False}  # pragma: no cover.
+  crash_deps = chromium_deps.GetChromeDependency(crashed_version, platform)
+  stacktrace = FracasParser().Parse(stack_trace, crash_deps)
+
+  regression_deps_rolls = {}
+  regression_versions = detect_regression_range.DetectRegressionRange(
+      historic_metadata)
+
+  if regression_versions:
+    good_version, bad_version = regression_versions

[Martin Barbella, 2016/05/05 16:37:37]

How about last_good_version, first_bad_version? Makes it a bit easier to parse
for someone who isn't as familiar with the algorithm.

[Sharu Jiang, 2016/05/05 20:26:45]

Done.

+    # Get regression deps and crash deps.
+    regression_deps_rolls = chromium_deps.GetDEPSRollsDict(
+        good_version, bad_version, platform)
+
+  culprit_results = findit_for_crash.FindItForCrash(
+      stacktrace, regression_deps_rolls, crash_deps)
+
+  # TODO(katesonia): Enable dependency classifier and component classifier.
+  suspected_dep = ''
+  suspected_components = []
+
+  return (
+      {
+          'found': True,

[Martin Barbella, 2016/05/05 16:37:37]

Is this correct?

[Sharu Jiang, 2016/05/05 20:26:45]

No, updated it.

+          'suspected_dep': suspected_dep,
+          'suspected_components': suspected_components,
+          'culprits': culprit_results,
+      },
+      {
+          'found_suspects': bool(culprit_results),
+          'has_regression_range': bool(regression_versions),
+          'solution': 'core_algorithm',

[stgao, 2016/05/05 17:32:28]

Is it always 'core_algorithm'? What if there is no regression range detected --
is git blame used instead?

[Sharu Jiang, 2016/05/05 20:26:45]

Right now, we don't support the git blame yet because people has concerns about
the git blame ranking. Once we enable the git blame, I will modify this.

[stgao, 2016/05/05 21:02:46]

In that case, if we don't have regression range, should we still run the
analysis above in #79?

[Sharu Jiang, 2016/05/05 23:03:45]

In FindItForCrash, we checkout if the regression_deps_rolls is empty or not,
besides, I plan to add the git blame code inside FindItForCrash, so I still want
to run this function.

[stgao, 2016/05/06 19:11:24]

As long as analysis is not run and empty result is returned for crash without
regression range, we are good.

+      }
+  )