Fix the regression problem after using unsafe appendChild

This is a hacky way to fix the regression problem after using the unsafe
appendChild. We need to convert the merged doc back and forth to make
sure it builds indexes for the later pretty-print.

Bug: 1115727
Change-Id: I9225999aec8132e3a069fe79cacc8b1377e99af7
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2354206Reviewed-by: Steven Holte <holte@chromium.org>
Commit-Queue: Steven Holte <holte@chromium.org>
Auto-Submit: Weilun Shi <sweilun@chromium.org>
Cr-Commit-Position: refs/heads/master@{#798176}

tools/metrics/histograms/merge_xml.py

@@ -131,7 +131,12 @@ def MergeTrees(trees):
           # Sort the histograms by name and return a single <histograms> node.
           CombineHistogramsSorted(doc, trees) +
           GetElementsByTagName(trees, 'histogram_suffixes_list')))
-  return doc
+  # After using the unsafe version of appendChild, we see a regression when
+  # pretty-printing the merged |doc|. This might because the unsafe appendChild
+  # doesn't build indexes for later lookup. And thus, we need to convert the
+  # merged |doc| to a xml string and convert it back to force it to build
+  # indexes for the merged |doc|.
+  return xml.dom.minidom.parseString(doc.toxml())
 
 
 def MergeFiles(filenames=[], files=[]):