content: Allow software compositing on Chrome OS if requested

content/browser/compositor/gpu_process_transport_factory.cc

Index: content/browser/compositor/gpu_process_transport_factory.cc
diff --git a/content/browser/compositor/gpu_process_transport_factory.cc b/content/browser/compositor/gpu_process_transport_factory.cc
index 673074c04b6d73ab55077a9f0565db1cbdc7a68f..eb302d7db1d7a529595f5400f0d0e5baee25e642 100644
--- a/content/browser/compositor/gpu_process_transport_factory.cc
+++ b/content/browser/compositor/gpu_process_transport_factory.cc
@@ -178,6 +178,9 @@ CreateOverlayCandidateValidator(gfx::AcceleratedWidget widget) {
 }
 
 static bool ShouldCreateGpuOutputSurface(ui::Compositor* compositor) {
+  if (base::CommandLine::ForCurrentProcess()->HasSwitch(switches::kDisableGpu))

[danakj, 2015/07/07 23:42:45]

Hm, yes.. but I guess it means you have 2 places using this switch now:

https://code.google.com/p/chromium/codesearch#chromium/src/content/browser/gp...

and this makes things harder to reason about. What do you think John? Should the
#if CHROMEOS maybe live in the GpuDataManagerImplPrivate and this could always
respect the CanUseGpuBrowserCompositor?

Right now if you used --disable-gpu you'd end up disabling all GPU things except
the compositor, which is inconsistent and weird, because we make this #ifdef
decision over here instead of in the GpuDataManager.

I'm a little worried about making disable-gpu work on chromeos too. Suddenly
people may start using it in ways you didn't intend here. For example we had a
bunch of browser tests running chromeos with software compositing completely
accidentally before.

How hard is it to use a patched version of chrome in the cases where you'd want
to use this flag?

+    return false;
+
 #if defined(OS_CHROMEOS)
   // Software fallback does not happen on Chrome OS.
   return true;