[Findit] Extract failed output nodes from stdio log of compile step.

appengine/findit/waterfall/extractors.py

Index: appengine/findit/waterfall/extractors.py
diff --git a/appengine/findit/waterfall/extractors.py b/appengine/findit/waterfall/extractors.py
index b0f6639c2775407366bf0aa1f22d572cc94bab4c..a79a734f3c900bb7c5f55c749c6751345ef106f0 100644
--- a/appengine/findit/waterfall/extractors.py
+++ b/appengine/findit/waterfall/extractors.py
@@ -217,6 +217,7 @@ class CompileStepExtractor(Extractor):
     signal = FailureSignal()
     failure_started = False
     is_build_command_line = False
+    failed_output_nodes = set()
 
     if (master_name == self.MAC_MASTER_NAME_FOR_COMPILE and
         bot_name in self.IOS_BUILDER_NAMES_FOR_COMPILE):
@@ -242,6 +243,28 @@ class CompileStepExtractor(Extractor):
         if line.startswith(self.FAILURE_START_LINE_PREFIX):
           if not failure_started:
             failure_started = True
+          line = line[len(self.FAILURE_START_LINE_PREFIX):]
+          while line:

[chanli, 2016/04/30 02:24:58]

Is it possible that the line is like:
FAILED: a.o "something in quotes" b.o "some other thing in quotes"?

[stgao, 2016/05/01 05:12:28]

In theory, this is possible.
But in practice, when with quote, it would be more like:
FAILED: "some target with space".

And this is handled by the code below.

+            quote_index = line.find('"')

[lijeffrey, 2016/04/30 03:44:25]

nit: it seems since this is a temporary solution, maybe extract this into a
separate function that takes line and returns a set that will be
failed_output_nodes?

i.e. 

def _GetOutputNodesFromLine(line):
  failed_output_nodes = set()
  while line:
    quote_index = line.find('"')
    ...
  return failed_output_nodes

then later

line = line[len(self.FAILURE_START_LINE_PREFIX):]
failed_ouptut_nodes = _GetOutputNodesFromLine(line)
is_build_command_line = True
...


This way the function can be more easily tested and removed when this temporary
solution is no longer needed

[stgao, 2016/05/01 05:12:28]

This is a good idea. 
Done, but with additional change.

+            if quote_index < 0:
+              sub_part = line
+              remaining_part = None
+            else:
+              sub_part = line[:quote_index]
+              match_quote_index = line.find('"', quote_index + 1)
+              if match_quote_index < 0:  # Unexpected format.
+                failed_output_nodes.clear()
+                break
+              failed_output_nodes.add(
+                  line[quote_index + 1: match_quote_index])
+              remaining_part = line[match_quote_index + 1:]
+            line = remaining_part
+
+            for node in sub_part.split(' '):
+              node = node.strip()
+              if node:
+                failed_output_nodes.add(node)
+
           is_build_command_line = True
           continue
         elif is_build_command_line:
@@ -270,6 +293,7 @@ class CompileStepExtractor(Extractor):
           # either within the compile errors or is a ninja error.
           self.ExtractFiles(line, signal)
 
+    signal.failed_output_nodes = sorted(failed_output_nodes)
     return signal