Don't check the result of new for NULL.

net/tools/crash_cache/crash_cache.cc

Index: net/tools/crash_cache/crash_cache.cc
diff --git a/net/tools/crash_cache/crash_cache.cc b/net/tools/crash_cache/crash_cache.cc
index 700e822aeba6cf76a6c423742b47907a0319b21a..72addfbb004eed03e8dc97eaefe14c5b6a005b7b 100644
--- a/net/tools/crash_cache/crash_cache.cc
+++ b/net/tools/crash_cache/crash_cache.cc
@@ -268,7 +268,7 @@ int LoadOperations(const base::FilePath& path, RankCrashes action,
   // Work with a tiny index table (16 entries).
   disk_cache::BackendImpl* cache = new disk_cache::BackendImpl(
       path, 0xf, cache_thread->message_loop_proxy().get(), NULL);
-  if (!cache || !cache->SetMaxSize(0x100000))
+  if (!cache->SetMaxSize(0x100000))

[rvargas (doing something else), 2013/09/03 19:01:06]

Can we leave this as it was? This is a tool, not part of the released product
(so the allocator doesn't necessarily crash when returning null), and the
purpose of the tool is to crash the cache at specific points, so crashing here
instead of returning an error is arguably the wrong behavior.

[Nico, 2013/09/03 19:17:33]

Have you checked that new actually returns null on error? According to the C++
standard it doesn't, and at least with libstdc++ new throws in -fno-exception
builds too as far as I know.

[rvargas (doing something else), 2013/10/16 19:01:23]

As of vs2010 this also crashes with C++ exceptions disabled. So, nevermind.

     return GENERIC;
 
   // No experiments and use a simple LRU.