Add histogram for image errors.

content/browser/loader/resource_dispatcher_host_impl.cc

Index: content/browser/loader/resource_dispatcher_host_impl.cc
diff --git a/content/browser/loader/resource_dispatcher_host_impl.cc b/content/browser/loader/resource_dispatcher_host_impl.cc
index f3a441b8a5a2a0e7d28e2d62382feb5762d39621..857d165613352d95d0dcf52bb34d462449c37007 100644
--- a/content/browser/loader/resource_dispatcher_host_impl.cc
+++ b/content/browser/loader/resource_dispatcher_host_impl.cc
@@ -756,6 +756,13 @@ void ResourceDispatcherHostImpl::DidFinishLoading(ResourceLoader* loader) {
               kAllNetErrorCodes, arraysize(kAllNetErrorCodes)));
     }
   } else {
+    if (info->GetResourceType() == ResourceType::IMAGE) {
+      UMA_HISTOGRAM_CUSTOM_ENUMERATION(
+          "Net.ErrorCodesForImages",
+          -loader->request()->status().error(),
+          base::CustomHistogram::ArrayToCustomRanges(
+              kAllNetErrorCodes, arraysize(kAllNetErrorCodes)));

[Ilya Sherman, 2013/04/26 23:59:10]

Can you use a sparse histogram here?  (The corresponding macro is
UMA_HISTOGRAM_SPARSE_SLOWLY.)

[jar (doing other things), 2013/04/27 00:48:58]

+1

On 2013/04/26 23:59:10, Ilya Sherman wrote:

[James Simonsen, 2013/04/27 00:57:07]

If we're going to do that, we should update all of them in this file. I thought
this one was okay because it matched the rest of the file.

[jar (doing other things), 2013/04/27 01:00:53]

SGTM

The list has gotten long (so memory utilization is high), and this histogram
*should* be rarely augmented (and when it is, it is dwarfed by the compute cost
that got us here typically), so this "map and lock" based histogram is now a
better choice.  The good news is that existing histograms (with 1-wide buckets)
should be completely compatible, so there is no need to change the names, or the
histograms.xml file.


On 2013/04/27 00:57:07, James Simonsen wrote:

[Stephen, 2013/04/27 03:02:41]

OK so IIUC the code here should be
UMA_HISTOGRAM_SPARSE_SLOWLY("Net.ErrorCodesForImages",
-loader->request()->status().error()); and the other CUSTOM_ENUMERATION macros
should be updated similarly.  Is that right?

On 2013/04/27 01:00:53, jar wrote:

[Ilya Sherman, 2013/04/27 23:06:23]

Assuming that the expression "-loader->request()->status().error()" gives the
net error code, then yes :)
 
Yeah, sparse histograms are relatively new, so there's a lot of existing
histograms like these that would be good to move over.

+    }
     // This enumeration has "2" appended to distinguish it from older versions.
     UMA_HISTOGRAM_CUSTOM_ENUMERATION(
         "Net.ErrorCodesForSubresources2",