Fix WebRTC log list errors.

chrome/browser/resources/media/webrtc_logs.js

 // Copyright 2013 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 /**
  * Requests the list of uploads from the backend.
  */
 function requestUploads() {
   chrome.send('requestWebRtcLogsList');
 }
@@ skipping 11 matching lines @@
 
   // Clear any previous list.
   logSection.textContent = '';
 
   for (var i = 0; i < uploads.length; i++) {
     var upload = uploads[i];
 
     var logBlock = document.createElement('div');
 
     var title = document.createElement('h3');
+    // TODO(grunell): Use |capture_time| only. If empty, show "Unknown" or such.
+    // In october 2015, capture time was added to the log list, previously the
+    // local ID was used as capture time. It still needs to be backwards
+    // compatible for some time before changing here.

[tommi (sloooow) - chröme, 2015/10/16 18:24:48]

Should the log include version information?

[Henrik Grunell, 2015/10/19 10:53:24]

I assume you mean the log list, and update an old version to the new format. One
problem is that the list is updated when writing a new entry, and if reading it
here may happen before that. I'm not sure it's worth it to add a browser startup
timer for updating and refactor etc. Or did you mean just adapt what we show to
the version? I don't see that being much better, but I'm open to that.

[tommi (sloooow) - chröme, 2015/10/19 11:11:09]

What I was getting at is that if we could say "version 0 does not have the
capture_time, so we can assume it's not going to be there for that version.
Version 1 however does have it, so assume it's there".  When writing the file,
we always write the latest format.  So yeah, basically we'd know for sure what's
in each version instead of guessing or having a period of testing what fields
might be available. If we add one more, remove or modify, then things will get
hairy if we just guess.

[Henrik Grunell, 2015/10/19 12:43:46]

We'd have to convert the list info the new version in that case to ensure all
lines have the same version. Instead I did so that if no capture time, we fall
back on local ID as capture time if within reasonable bounds, from when the
feature was introduced until Now(). If it isn't, fall back on upload time.
Within some time, a list should  be in the new format anyway. Wdyt?

[tommi (sloooow) - chröme, 2015/10/19 12:51:42]

What about changing capture_time in the C++ code to be the upload time and keep
the js code slightly simpler?
We could then move the version handling to there (C++ that is) if/when we need
it.  Generally I'd prefer to have as few places as possible be aware of the
on-disk format but now it feels like we have two places.

[Henrik Grunell, 2015/10/19 12:58:32]

I've already done that. :) Forgot to mention it last comment.

     title.textContent =
         loadTimeData.getStringF('webrtcLogHeaderFormat',
                                 upload['capture_time'].length != 0 ?
                                     upload['capture_time'] :
                                     upload['upload_time']);
     logBlock.appendChild(title);
 
     var localFileLine = document.createElement('p');
     if (upload['local_file'].length == 0) {
       localFileLine.textContent =
@@ skipping 54 matching lines @@
     }
     logBlock.appendChild(uploadLine);
 
     logSection.appendChild(logBlock);
   }
 
   $('no-logs').hidden = uploads.length != 0;
 }
 
 document.addEventListener('DOMContentLoaded', requestUploads);