Re-disable warning 4018 in the GN Windows build.

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")
 if (cpu_arch == "arm") {
   import("//build/config/arm.gni")
 }
 if (is_posix) {
   import("//build/config/gcc/gcc_version.gni")
@@ skipping 550 matching lines @@
 # This is where we disable various warnings that we've decided aren't
 # worthwhile, and enable special warnings.
 
 config("default_warnings") {
   if (is_win) {
     cflags = [
       "/WX",      # Treat warnings as errors.
 
       # Warnings permanently disabled:
 
+      # TODO(GYP) The GYP build doesn't have this globally enabled but disabled

[Peter Kasting, 2014/07/28 11:20:47]

Is "doesn't have" here supposed to read "has"?

Why are these targets (those that build in GN, at least) not also disabling this
warning in the GN build?  It concerns me if the GN build is globally disabling
something that isn't globally disabled in the GYP build.

[brettw, 2014/07/28 18:05:42]

Previously it was globally disabled in the GYP build and then a bunch of targets
individually disabled it. I don't know why this was the case and I'm quire
surprised that removing the global one actually worked.

In GYP we have a bunch of stuff disabled globally and the same things disabled
per-target. I'm not duplicating the per-target ones if it's disabled globally.

I agree that having a divergence is bad. I think the right solution is to either
disable it in GYP globally again or go in and add it to the gn files missing it.

[Peter Kasting, 2014/07/28 19:20:50]

Removing the global one worked because I manually fixed all the errors that it
triggered before I landed the change to remove it.  Ideally individual targets
would not override this, but even if they do, this means we still have the
warning enabled for the majority of our codebase, which is progress over where
we were before.
Disabling it globally in GYP would be strictly worse than today, since that
would mean we won't get the warning coverage for all the targets that don't
individually disable it.

Can you change this to match the GYP build, i.e. re-enable this globally and
then disable it per-target instead?  Or if there's a list of the relevant
targets, send that to me to update, or something?

Basically, let's work together to make sure this is on by default instead of
off.

+      # for a bunch of individual targets. Re-enable this globally when those
+      # targets are fixed.
+      "/wd4018",  # Comparing signed and unsigned values.
+
       # C4127: conditional expression is constant
       # This warning can in theory catch dead code and other problems, but
       # triggers in far too many desirable cases where the conditional
       # expression is either set by macros or corresponds some legitimate
       # compile-time constant expression (due to constant template args,
       # conditionals comparing the sizes of different types, etc.).  Some of
       # these can be worked around, but it's not worth it.
       "/wd4127",
 
       # C4251: 'identifier' : class 'type' needs to have dll-interface to be
@@ skipping 297 matching lines @@
   } else {
     cflags = [ "-g1" ]
   }
 }
 
 config("no_symbols") {
   if (!is_win) {
     cflags = [ "-g0" ]
   }
 }