[test] Sync in *san configurations from chromium.

build/standalone.gypi

Index: build/standalone.gypi
diff --git a/build/standalone.gypi b/build/standalone.gypi
index 8bafda91d13e1c0be964602ac7e480b3ce051e7e..38b2b65875987889ef031fb0e85adcf6e26e3cc4 100644
--- a/build/standalone.gypi
+++ b/build/standalone.gypi
@@ -288,69 +288,99 @@
     ],
   },
   'conditions': [
-    ['asan==1 and OS!="mac"', {
+    ['OS!="mac"', {

[earthdok, 2015/05/22 19:57:25]

Following common.gypi, you should probably change this to "posix and not mac".
As it is now, it's not obvious why Mac is special.

[Michael Achenbach, 2015/05/26 08:26:35]

Done. But I'm not convinced that our posix variable is meaningful. Our gyp is a
bit of a mess and the posix variable seems not to be set in a nested scope and
then copied out.

[earthdok, 2015/05/26 13:36:54]

Try flipping the values this variable receives. My experiment shows that it is
meaningful.

[Michael Achenbach, 2015/05/26 14:20:29]

Indeed.

       'target_defaults': {
-        'cflags_cc+': [
-          '-fno-omit-frame-pointer',
-          '-gline-tables-only',
-          '-fsanitize=address',
-          '-w',  # http://crbug.com/162783
-        ],
-        'cflags!': [
-          '-fomit-frame-pointer',
-        ],
-        'ldflags': [
-          '-fsanitize=address',
-        ],
-      },
-    }],
-    ['tsan==1 and OS!="mac"', {
-      'target_defaults': {
-        'cflags+': [
-          '-fno-omit-frame-pointer',
-          '-gline-tables-only',
-          '-fsanitize=thread',
-          '-fPIC',
-          '-Wno-c++11-extensions',
-        ],
-        'cflags!': [
-          '-fomit-frame-pointer',
-        ],
-        'ldflags': [
-          '-fsanitize=thread',
-          '-pie',
-        ],
-        'defines': [
-          'THREAD_SANITIZER',
-        ],
-      },
-    }],
-    ['msan==1 and OS!="mac"', {
-      'target_defaults': {
-        'cflags_cc+': [
-          '-fno-omit-frame-pointer',
-          '-gline-tables-only',
-          '-fsanitize=memory',
-          '-fsanitize-memory-track-origins=<(msan_track_origins)',
-          '-fPIC',
-        ],
-        'cflags+': [
-          '-fPIC',
-        ],
-        'cflags!': [
-          '-fomit-frame-pointer',
-        ],
-        'ldflags': [
-          '-fsanitize=memory',
-        ],
-        'defines': [
-          'MEMORY_SANITIZER',
-        ],
-        'dependencies': [
-          # Use libc++ (third_party/libc++ and third_party/libc++abi) instead of
-          # stdlibc++ as standard library. This is intended to use for instrumented
-          # builds.
-          '<(DEPTH)/buildtools/third_party/libc++/libc++.gyp:libcxx_proxy',
+       'conditions': [
+          # Common options for AddressSanitizer, LeakSanitizer,
+          # ThreadSanitizer and MemorySanitizer.
+          ['asan==1 or lsan==1 or tsan==1 or msan==1', {
+            'target_conditions': [
+              ['_toolset=="target"', {
+                'cflags': [
+                  '-fno-omit-frame-pointer',
+                  '-gline-tables-only',
+                ],
+                'cflags!': [
+                  '-fomit-frame-pointer',
+                ],
+                'ldflags!': [
+                  # Functions interposed by the sanitizers can make ld think
+                  # that some libraries aren't needed when they actually are,
+                  # http://crbug.com/234010. As workaround, disable --as-needed.
+                  '-Wl,--as-needed',

[earthdok, 2015/05/22 19:57:25]

Do not add this unless the problem manifests in standalone v8. Looking at the
referenced issue, it looks like we don't even need this in Chromium any longer.

[Michael Achenbach, 2015/05/26 08:26:35]

Done.

+                ],
+              }],

[earthdok, 2015/05/22 19:57:25]

Need to define the ADDRESS_SANITIZER symbol here, as well as in the Mac config:
https://code.google.com/p/chromium/codesearch#chromium/src/build/common.gypi&...

(This is not done uniformly with other tools because the ASan symbol is defined
across multiple platforms, whereas the other tools are Linux-only.)

(Please ignore the "MEMORY_SANITIZER_INITIAL_SIZE" symbol in line 2808 - this is
not relevant to V8.)

[Michael Achenbach, 2015/05/26 08:26:35]

Done. See above. Btw: codesearch links get outdated quite quickly in
common.gypi. I try to use fixed git hashes:
https://chromium.googlesource.com/chromium/src/+/11784f3a12/build/common.gypi...

I assume MEMORY_TOOL_REPLACES_ALLOCATOR is also not important?

[earthdok, 2015/05/26 13:36:54]

It's not referenced in V8 code, so no. It would not mind if this changed in the
future, though. MEMORY_SANITIZER_INITIAL_SIZE needs to die, on the other hand.

[Michael Achenbach, 2015/05/26 14:20:29]

I'd add it as soon as someone wants to reference it.

+            ],
+          }],
+          ['asan==1', {
+            'target_conditions': [
+              ['_toolset=="target"', {
+                'cflags': [
+                  '-fsanitize=address',
+                  # TODO(earthdok): Re-enable. http://crbug.com/427202

[earthdok, 2015/05/22 19:57:25]

The referenced issue does not affect standalone V8, so this TODO should look the
same as the TODOs for TSan, MSan blacklists.

Generally speaking, I don't think you want to have blacklists. If you do, you
would have to either pull them from Chromium, or keep them in sync with the
blacklists in Chromium.

[Michael Achenbach, 2015/05/26 08:26:35]

Removing these TODOs for now.

+                  #'-fsanitize-blacklist=<(asan_blacklist)',
+                ],
+                'ldflags': [
+                  '-fsanitize=address',
+                ],
+              }],
+            ],
+          }],
+          # TODO(machenbach): Add sanitizer coverage.
+          ['lsan==1', {
+            'target_conditions': [
+              ['_toolset=="target"', {
+                'cflags': [
+                  '-fsanitize=leak',
+                ],
+                'ldflags': [
+                  '-fsanitize=leak',
+                ],
+              }],

[earthdok, 2015/05/22 19:57:25]

define LEAK_SANITIZER?

[Michael Achenbach, 2015/05/26 08:26:35]

Done.

+            ],
+          }],
+          ['tsan==1', {
+            'target_conditions': [
+              ['_toolset=="target"', {
+                'cflags': [
+                  '-fsanitize=thread',
+                  # TODO(machenbach): Implement this for v8.
+                  # '-fsanitize-blacklist=<(tsan_blacklist)',
+                ],
+                'ldflags': [
+                  '-fsanitize=thread',
+                ],
+                'defines': [
+                  'THREAD_SANITIZER',
+                ],
+              }],
+            ],
+          }],
+          ['msan==1', {
+            'target_conditions': [
+              ['_toolset=="target"', {
+                'cflags': [
+                  '-fsanitize=memory',
+                  '-fsanitize-memory-track-origins=<(msan_track_origins)',
+                  # TODO(machenbach): Implement this for v8.
+                  # '-fsanitize-blacklist=<(msan_blacklist)',
+                ],
+                'ldflags': [
+                  '-fsanitize=memory',
+                ],
+                'defines': [
+                  'MEMORY_SANITIZER',
+                ],
+              }],
+            ],
+            # Share this between all *san configs with like in common.gypi.
+            'dependencies': [
+              # Use libc++ (third_party/libc++ and third_party/libc++abi) instead of

[earthdok, 2015/05/22 19:57:25]

Those pathnames are no longer correct after the move to buildtools, please fix.

[Michael Achenbach, 2015/05/26 08:26:35]

Done.

+              # stdlibc++ as standard library. This is intended to use for instrumented

[earthdok, 2015/05/22 19:57:25]

Please fix: "intended to be used for".

[Michael Achenbach, 2015/05/26 08:26:35]

Done.

+              # builds.
+              '<(DEPTH)/buildtools/third_party/libc++/libc++.gyp:libcxx_proxy',
+            ],
+          }],
         ],
       },
     }],