Don't use m32 cflag for x86_64.

gyp/common_variables.gypi

Index: gyp/common_variables.gypi
diff --git a/gyp/common_variables.gypi b/gyp/common_variables.gypi
index f89c08b95e2f5a6879f43b4577c1fac67c4a5687..5b5d2ebf013849484fb2a2b04eae014459014f52 100644
--- a/gyp/common_variables.gypi
+++ b/gyp/common_variables.gypi
@@ -82,7 +82,7 @@
         }, {
           'skia_poppler_enabled%': 0,
         }],
-        [ 'skia_os in ["linux", "freebsd", "openbsd", "solaris", "mac"] or skia_arch_type == "arm64"', {
+        [ 'skia_os in ["linux", "freebsd", "openbsd", "solaris", "mac"] or "64" in skia_arch_type', {

[mtklein, 2015/02/11 18:15:05]

So are we going to get "x86" and "x86_64"  as skia_arch_type consistently
through all the builds, or only for Android builds?  Am I thinking right that on
desktops we'll still see "x86" and skia_arch_width=64?  That seems pretty
confusing.

[scroggo, 2015/02/11 20:12:10]

I would like it to be the case for all builds (although I have no control over
people who manually set their gyp defines... Do we ever do that on the bots?).
I think you are correct.

If we were consistent, this line could just say

['"64" in skia_arch_type',

If I understand this nesting process, that would mean we'd need to set the
default to x86_64 in the variables 3 dict above for these OSs.

Which means I'd need to add one deeper level which defines skia_os, and in level
3 do this check to switch to x86_64. Does that sound correct?

[scroggo, 2015/02/11 20:15:26]

And now of course I'm not sure of that. It appears that we already both set and
condtion on skia_os in level 2...

[scroggo, 2015/02/12 15:35:45]

I made that change in https://codereview.chromium.org/919003002, which builds on
this one. PTAL.

[scroggo, 2015/02/12 15:55:47]

Pulled in that change here. PTAL.

           'skia_arch_width%': 64,
         }, {
           'skia_arch_width%': 32,