Make url_request_context dumps include context names

base/trace_event/memory_infra_background_whitelist.cc

Index: base/trace_event/memory_infra_background_whitelist.cc
diff --git a/base/trace_event/memory_infra_background_whitelist.cc b/base/trace_event/memory_infra_background_whitelist.cc
index ae74322040f45043af0183f8fda5b32223910170..524a3d6192a69a5a181fb7a02f276120b313158c 100644
--- a/base/trace_event/memory_infra_background_whitelist.cc
+++ b/base/trace_event/memory_infra_background_whitelist.cc
@@ -69,10 +69,12 @@ const char* const kAllocatorDumpNameWhitelist[] = {
     "net/http_network_session_0x?/stream_factory",
     "net/sdch_manager_0x?",
     "net/ssl_session_cache",
-    "net/url_request_context_0x?",
-    "net/url_request_context_0x?/http_cache",
-    "net/url_request_context_0x?/http_network_session",
-    "net/url_request_context_0x?/sdch_manager",
+    "net/url_request_context",
+    "net/url_request_context/%s",
+    "net/url_request_context/%s/0x?",
+    "net/url_request_context/%s/0x?/http_cache",
+    "net/url_request_context/%s/0x?/http_network_session",
+    "net/url_request_context/%s/0x?/sdch_manager",
     "web_cache/Image_resources",
     "web_cache/CSS stylesheet_resources",
     "web_cache/Script_resources",
@@ -178,6 +180,16 @@ bool IsMemoryAllocatorDumpNameWhitelisted(const std::string& name) {
     }
   }
 
+  std::string url_request_context_prefix = "url_request_context/";
+  size_t pos = stripped_str.find(url_request_context_prefix);
+  if (pos != std::string::npos) {
+    size_t replace_start = pos + url_request_context_prefix.length();
+    size_t replace_end = stripped_str.find('\\', replace_start);
+    if (replace_end == std::string::npos)
+      replace_end = stripped_str.length();
+    stripped_str.replace(replace_start, replace_end - pos, "%s");

[ssid, 2017/03/27 23:02:34]

No this is exactly what we are trying to avoid in this file. The reason why we
whitelist is to avoid reporting PII data of the users. At some point in future
if you decide to make the "name" of url_request as the url name, then this would
end up uploading the url names which is a violation of the user privacy.
The original patch did it correctly with all the possible strings listed here. 
Also please look at the nit i added later:
https://codereview.chromium.org/2772283003/diff/1/base/trace_event/memory_inf...

The issue with "main" and "main_media" can be solved by adding "main/" and
"main_media/" I guess?

[xunjieli, 2017/03/28 01:04:56]

Done. 

Just to clarify: URLRequestContext is the thing that makes URLRequests. The
Context won't ever have a url attached to it.  

+  }
+
   for (size_t i = 0; g_allocator_dump_name_whitelist[i] != nullptr; ++i) {
     if (stripped_str == g_allocator_dump_name_whitelist[i]) {
       return true;