GN: Use icf=all instead of icf=safe

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 338 matching lines @@
   } else if (is_android && is_clang &&
              (current_cpu == "mipsel" || current_cpu == "mips64el")) {
     # Let clang find the ld.bfd in the NDK.
     _rebased_android_toolchain_root =
         rebase_path(android_toolchain_root, root_build_dir)
     ldflags += [ "--gcc-toolchain=$_rebased_android_toolchain_root" ]
   }
 
   if (is_posix && (use_gold || (use_lld && !is_nacl)) && !using_sanitizer &&
       !(is_android && use_order_profiling)) {
-    # TODO(crbug.com/576197) - gcc on x86 platforms + gold + icf=all
-    # doesn't currently work. Once it does, use icf=all everywhere.
-    # Additionally, on Android x86 --icf=safe seems to cause issues as well.
-    # Additionally, on cast Android x86, --icf=all causes issues.
+    # On cast Android x86, --icf=all causes issues.
     if (is_android && current_cpu == "x86") {
       ldflags += [ "-Wl,--icf=none" ]
-    } else if (is_clang || (current_cpu != "x86" && current_cpu != "x64")) {
+    } else {

[Dirk Pranke, 2016/07/15 21:08:42]

won't this still break linux/gcc x86? From my reading of 576197, the underlying
gold issue hasn't been fixed yet.

[stevenjb, 2016/07/15 21:12:07]

I didn't read through all of it, I assumed that since it was resolved it was,
well, fixed.

If it isn't resolved on linux (which I assume includes chromeos?) we need to
reopen the issue or open another issue, update this comment, and prioritize it,
since a 5% size increase is pretty awful.

[llozano, 2016/07/15 21:12:27]

this has been in use by chromeos for a while but it has not been tested on linux
or Android. I would prefer we figure out why the option was overriden by GN and
allow to make it a ChromeOS thing only.

[Dirk Pranke, 2016/07/15 21:30:05]

I agree that we should update the comment, possibly to refer to the binutils bug
instead of the chromium bug
(https://sourceware.org/bugzilla/show_bug.cgi?id=17704).

It looks like the bug was reported after we had switched the linux build to GN,
so I'm guessing we just never applied the "fix" to the GYP files, and so the
GYP-based CrOS builds were unaffected and stayed on --icf=all.

I'm fine w/ you changing the code so that you get --icf=all on CrOS x86 if you
think it's safe to do so, just don't break the other platforms where we still
think it isn't safe :). (and, no pun intended). 

[stevenjb, 2016/07/15 21:36:49]

So, I don't really understand the problem or the solution well enough to
meaningfully update the comment.

llozano@, if you or someone on your team could provide a better patch that would
be great, otherwise someone can l-g-t-m this, just whatever it takes to get this
fix in before Chrome OS 54 goes to Beta, thanks!

       ldflags += [ "-Wl,--icf=all" ]
-    } else if (!is_android) {
-      ldflags += [ "-Wl,--icf=safe" ]
     }
   }
 
   if (linux_use_bundled_binutils) {
     cflags += [ "-B$binutils_path" ]
   }
 
   # Clang-specific compiler flags setup.
   # ------------------------------------
   if (is_clang) {
@@ skipping 1187 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
   }
 }