Sparse IO: Allow failover to network in debug builds

net/disk_cache/simple/simple_entry_impl.cc

Index: net/disk_cache/simple/simple_entry_impl.cc
diff --git a/net/disk_cache/simple/simple_entry_impl.cc b/net/disk_cache/simple/simple_entry_impl.cc
index 72d6d4c7e104a03db5499460443afac22d7d20f9..94997efb538c315465458d36d2a7e4cf889c424d 100644
--- a/net/disk_cache/simple/simple_entry_impl.cc
+++ b/net/disk_cache/simple/simple_entry_impl.cc
@@ -496,7 +496,11 @@ int SimpleEntryImpl::ReadyForSparseIO(const CompletionCallback& callback) {
   DCHECK(io_thread_checker_.CalledOnValidThread());
   // TODO(gavinp): Determine if the simple backend should support sparse data.
   NOTIMPLEMENTED();
+#if defined(OS_ANDROID)

[rvargas (doing something else), 2013/08/28 21:35:47]

There should not be an ifdef here.

[gavinp, 2013/08/28 21:41:25]

This was specifically requested by the TPM to make this CL mergeable. Perhaps
the thing to do is add a TODO() to remove this #ifdef after the change has baked
on canary for a day and then been merged?

[pasko, 2013/08/28 21:56:28]

I don't like this ifdef either, and maybe this one is a bit of extreme, just
wanted the TPM to feel cozy. I left a TODO in http_cache_transaction.cc to
remove this conditionalizing, and it then did not make much sense to replicate
that TODO at every hunk in the patch.

[rvargas (doing something else), 2013/08/28 22:33:40]

I just commented on the bug. I understand the other ifdef... this one is out of
place IMO. (but don't take this comment as blocking the CL in any way, because
it doesn't break anything)

+  return net::ERR_NOT_IMPLEMENTED;
+#else
   return net::ERR_FAILED;
+#endif
 }
 
 SimpleEntryImpl::~SimpleEntryImpl() {