Explicitly ask for stack frame pointers on Debug posix builds.

build/config/compiler/BUILD.gn

 # Copyright (c) 2013 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 import("//build/config/android/config.gni")
 import("//build/config/chrome_build.gni")
 import("//build/config/compiler/compiler.gni")
 import("//build/config/nacl/config.gni")
 import("//build/toolchain/cc_wrapper.gni")
 import("//build/toolchain/toolchain.gni")
@@ skipping 213 matching lines @@
     # This is here so that all files get recompiled after a clang roll and
     # when turning clang on or off. (defines are passed via the command line,
     # and build system rebuild things when their commandline changes). Nothing
     # should ever read this define.
     defines += [ "CR_CLANG_REVISION=$clang_revision" ]
   }
 
   # Non-Mac Posix compiler flags setup.
   # -----------------------------------
   if (is_posix && !(is_mac || is_ios)) {
-    if (enable_profiling) {
-      # Explicitly ask for frame pointers. Otherwise they are omitted when
-      # any optimization level is used (and Android debug builds use -Os).
+    if (enable_profiling || is_debug) {
+      # Explicitly ask for frame pointers, otherwise they are omitted when
+      # any optimization level is used (Android debug builds use -Os.)
+      # This prevents stack frames from working when profiling, and can cause
+      # crashes when unwinding stack frames (crbug.com/636489).

[Dirk Pranke, 2016/08/22 20:03:48]

Do we have any idea what the perf impact of -fno-omit-frame-pointer is? I've
been thinking that maybe it should just be on unconditionally in non-official
builds.

Alternatively, we might want to make it be a configurable argument; the goal of
either choice would partially be to make sure stack traces are as
debuggable/useful as possible.

[vmiura, 2016/08/22 20:32:04]

I don't have data but I expect the impact is fairly small.  The
'enable_profiling' flag partly overlaps with that.  Perhaps a follow up?

       cflags += [ "-fno-omit-frame-pointer" ]
+
       if (!is_debug) {
         cflags += [ "-g" ]
 
         if (enable_full_stack_frames_for_profiling) {
           cflags += [
             "-fno-inline",
             "-fno-optimize-sibling-calls",
           ]
         }
       }
@@ skipping 1019 matching lines @@
       # Don't emit the GCC version ident directives, they just end up in the
       # .comment section taking up binary size.
       "-fno-ident",
 
       # Put data and code in their own sections, so that unused symbols
       # can be removed at link time with --gc-sections.
       "-fdata-sections",
       "-ffunction-sections",
     ]
 
+    if (is_debug) {
+      # Explicitly ask for frame pointers, otherwise they are omitted when
+      # any optimization level is used and can cause crashes when unwinding
+      # stack frames (crbug.com/636489).
+      common_optimize_on_cflags += [ "-fno-omit-frame-pointer" ]
+    }
+
     common_optimize_on_ldflags += [
       # Specifically tell the linker to perform optimizations.
       # See http://lwn.net/Articles/192624/ .
       "-Wl,-O1",
       "-Wl,--gc-sections",
     ]
 
     if (!using_sanitizer) {
       # Functions interposed by the sanitizers can make ld think
       # that some libraries aren't needed when they actually are,
@@ skipping 281 matching lines @@
 
 if (is_ios || is_mac) {
   # On Mac and iOS, this enables support for ARC (automatic ref-counting).
   # See http://clang.llvm.org/docs/AutomaticReferenceCounting.html.
   config("enable_arc") {
     common_flags = [ "-fobjc-arc" ]
     cflags_objc = common_flags
     cflags_objcc = common_flags
   }
 }