Print out the generated suppressions for Dr.Memory reports

The output format is very close to the one used by other tools, it should be
parseable by the buildbot log parsers.

Still TODO: the generated suppressions are too long since they disregard the
-callstack_truncate_below flag, see
http://code.google.com/p/drmemory/issues/detail?id=700

TBR=bruening
Review URL: http://codereview.chromium.org/8771003

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@112443 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 76 insertions, 11 deletions

diff --git a/tools/valgrind/drmemory_analyze.py b/tools/valgrind/drmemory_analyze.py
index 87c1310..db0e374 100755
--- a/tools/valgrind/drmemory_analyze.py
+++ b/tools/valgrind/drmemory_analyze.py
@@ -9,6 +9,7 @@
 
 from collections import defaultdict
 import common
+import hashlib
 import logging
 import optparse
 import os
@@ -17,6 +18,48 @@ import subprocess
 import sys
 import time
 
+class DrMemoryError:
+  def __init__(self, report, suppression, testcase):
+    self._report = report
+    self._testcase = testcase
+
+    # Chromium-specific transformations of the suppressions:
+    # Replace 'any_test.exe' and 'chrome.dll' with '*', then remove the
+    # Dr.Memory-generated error ids from the name= lines as they don't
+    # make sense in a multiprocess report.
+    supp_lines = suppression.split("\n")
+    for l in xrange(len(supp_lines)):
+      if supp_lines[l].startswith("name="):
+        supp_lines[l] = "name=<insert_a_suppression_name_here>"
+      if supp_lines[l].startswith("chrome.dll!"):
+        supp_lines[l] = supp_lines[l].replace("chrome.dll!", "*!")
+      bang_index = supp_lines[l].find("!")
+      d_exe_index = supp_lines[l].find(".exe!")
+      if bang_index >= 4 and d_exe_index + 4 == bang_index:
+        supp_lines[l] = "*.exe" + supp_lines[l][bang_index:]
+    self._suppression = "\n".join(supp_lines)
+
+  def __str__(self):
+    output = self._report + "\n"
+    if self._testcase:
+      output += "The report came from the `%s` test.\n" % self._testcase
+    output += "Suppression (error hash=#%016X#):\n" % self.ErrorHash()
+    output += "{\n%s\n}\n" % self._suppression
+
+    # TODO(timurrrr): docs on suppressing?
+    #output += ("  For more info on using suppressions see "
+    #           "http://dev.chromium.org/developers/how-tos/using-valgrind#TOC-Suppressing-Errors")
+    return output
+
+  # This is a device-independent hash identifying the suppression.
+  # By printing out this hash we can find duplicate reports between tests and
+  # different shards running on multiple buildbots
+  def ErrorHash(self):
+    return int(hashlib.md5(self._suppression).hexdigest()[:16], 16)
+
+  def __hash__(self):
+    return hash(self._suppression)
+
 
 class DrMemoryAnalyzer:
   ''' Given a set of Dr.Memory output files, parse all the errors out of
@@ -24,6 +67,7 @@ class DrMemoryAnalyzer:
 
   def __init__(self):
     self.known_errors = set()
+    self.error_count = 0;
 
   def ReadLine(self):
     self.line_ = self.cur_fd_.readline()
@@ -36,19 +80,42 @@ class DrMemoryAnalyzer:
       self.ReadLine()
     return result
 
-  def ParseReportFile(self, filename):
+  def ParseReportFile(self, filename, testcase):
     ret = []
 
+    # First, read the generated suppressions file so we can easily lookup a
+    # suppression for a given error.
+    supp_fd = open(filename.replace("results", "suppress"), 'r')
+    generated_suppressions = {}  # Key -> Error #, Value -> Suppression text.
+    for line in supp_fd:
+      # NOTE: this regexp looks fragile. Might break if the generated
+      # suppression format slightly changes.
+      m = re.search("# Suppression for Error #([0-9]+)", line.strip())
+      if not m:
+        continue
+      error_id = int(m.groups()[0])
+      assert error_id not in generated_suppressions
+      # OK, now read the next suppression:
+      cur_supp = ""
+      for supp_line in supp_fd:
+        if supp_line.startswith("#") or supp_line.strip() == "":
+          break
+        cur_supp += supp_line
+      generated_suppressions[error_id] = cur_supp.strip()
+    supp_fd.close()
+
     self.cur_fd_ = open(filename, 'r')
     while True:
       self.ReadLine()
       if (self.line_ == ''): break
 
-      match = re.search("^Error #[0-9]+: (.*)", self.line_)
+      match = re.search("^Error #([0-9]+): (.*)", self.line_)
       if match:
-        self.line_ = match.groups()[0].strip() + "\n"
-        tmp = self.ReadSection()
-        ret.append("".join(tmp).strip())
+        error_id = int(match.groups()[0])
+        self.line_ = match.groups()[1].strip() + "\n"
+        report = "".join(self.ReadSection()).strip()
+        suppression = generated_suppressions[error_id]
+        ret.append(DrMemoryError(report, suppression, testcase))
 
       if re.search("SUPPRESSIONS USED:", self.line_):
         self.ReadLine()
@@ -73,7 +140,7 @@ class DrMemoryAnalyzer:
     to_report = []
     self.used_suppressions = defaultdict(int)
     for f in filenames:
-      cur_reports = self.ParseReportFile(f)
+      cur_reports = self.ParseReportFile(f, testcase)
 
       # Filter out the reports that were there in previous tests.
       for r in cur_reports:
@@ -91,12 +158,10 @@ class DrMemoryAnalyzer:
 
     logging.error("Found %i error reports" % len(to_report))
     for report in to_report:
-      if testcase:
-        logging.error("\n%s\nNote: observed on `%s`\n" %
-                      (report, testcase))
-      else:
-        logging.error("\n%s\n" % report)
+      self.error_count += 1
+      logging.error("Report #%d\n%s" % (self.error_count, report))
     logging.error("Total: %i error reports" % len(to_report))
+    sys.stderr.flush()
     return -1