Change field_trials type to hashes in permission report proto

chrome/common/safe_browsing/permission_report.proto

Index: chrome/common/safe_browsing/permission_report.proto
diff --git a/chrome/common/safe_browsing/permission_report.proto b/chrome/common/safe_browsing/permission_report.proto
index 59dc519e40172ecf2d29e4e4cadd0167b282f421..604b2214dfd4c589a325d26e1efdc9c4904a1a4e 100644
--- a/chrome/common/safe_browsing/permission_report.proto
+++ b/chrome/common/safe_browsing/permission_report.proto
@@ -34,9 +34,7 @@ message PermissionReport {
   optional SourceUI source_ui = 6;
 
   // The relevant field trials enabled for this report.
-  // TODO(stefanocs): Update field_trials field to store hashes instead of
-  // string.
-  repeated string field_trials = 7;
+  repeated FieldTrial field_trials = 7;
 
   // Platform
   enum PlatformType {
@@ -89,4 +87,21 @@ message PermissionReport {
     AUDIO_CAPTURE = 8;
     VIDEO_CAPTURE = 9;
   }
+
+  // Description of a field trial or experiment that the user is currently
+  // enrolled in.

[raymes, 2016/06/08 03:00:11]

nit: fill 80 char lines

[stefanocs, 2016/06/08 03:47:22]

Done.

+  // All metrics reported in this upload can potentially be influenced by the
+  // field trial.
+  // This message is copied from 

[raymes, 2016/06/08 03:00:11]

nit you can add a new line above this for a new paragraph

[stefanocs, 2016/06/08 03:47:22]

Done.

+  // //src/components/metrics/proto/system_profile.proto
+  // since the file cannot be imported by the current build system.

[raymes, 2016/06/08 03:00:11]

Hmm do you know why this is? Is it hard to fix? Should we file a bug about it? 

[stefanocs, 2016/06/08 03:47:22]

Yes, I discussed this issue with Johan yesterday. We think the problem is that
the build system seems to be using a python script located at
//src/tools/protoc_wrapper/protoc_wrapper.py to compile the proto files and this
script only set the proto_path argument (which is the import path) to the
directory of the containing folder and not to the //src. Maybe we should file a
bug about this issue.

[raymes, 2016/06/09 01:44:17]

Do you think it would be hard to change the script? Maybe we should even just
add a hardcoded list of paths to the script for cases like this?

[stefanocs, 2016/06/09 07:45:09]

I have tried a simple fix by adding //src to proto_path, it was able to find the
imported file but it wasn't able to build the proto because the generated
function name is different.

For example when importing "components/metrics/proto/system_profile.proto" in
this proto:

A forward declaration of a function named
"protobuf_AddDesc_components_2fmetrics_2fproto_2fsystem_5fprofile_2eproto"
is generated by permission_report.proto.
But the function implementation generated by system_profile.proto is named
"protobuf_AddDesc_system_5fprofile_2eproto".

The cause is possibly because the protos are not compiled within the same level.
A more complicated changes to build system might be needed.

+  message FieldTrial {
+    // The name of the field trial, as a 32-bit identifier.

[raymes, 2016/06/08 03:00:11]

nit: fill 80 chars

[stefanocs, 2016/06/08 03:47:22]

Done.

+    // Currently, the identifier is a hash of the field trial's name.
+    optional fixed32 name_id = 1;
+
+    // The user's group within the field trial, as a 32-bit identifier.
+    // Currently, the identifier is a hash of the group's name.
+    optional fixed32 group_id = 2;
+  }
 }