Update trim_data to deal with locale fallback failure for unit

scripts/trim_data.sh

 #!/bin/bash
 # Copyright (c) 2014 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.
 
+set -e
 
 # Remove entries currently not used in Chromium/V8.
 function filter_locale_data {
   echo Removing unncessary categories in ${localedatapath}
   for langpath in ${localedatapath}/*.txt
   do
     echo Overwriting ${langpath} ...
     sed -r -i \
       '/^    characterLabel\{$/,/^    \}$/d
        /^    AuxExemplarCharacters\{.*\}$/d
@@ skipping 161 matching lines @@
   for i in ${dataroot}/unit/*.txt
   do
     echo Overwriting $i ...
     sed -r -i \
       '/^    units(|Narrow|Short)\{$/, /^    \}$/ {
          /^    units(|Narrow|Short)\{$/ p
          /^        (duration|compound)\{$/, /^        \}$/ p
          /^    \}$/ p
          d
        }' ${i}
+
+    # Delete empty units,units{Narrow|Short} block. Otherwise, locale fallback
+    # fails. See crbug.com/707515.
+    sed -r -i \
+      '/^    units(|Narrow|Short)\{$/ {
+         N
+         /^    units(|Narrow|Short)\{\n    \}/ d

[Daniel Erat, 2017/04/11 21:18:10]

sorry, my sed is a bit rusty/nonexistent. i think that this line is deleting
opening units lines followed by closing units lines, but i'm not sure what the
earlier pattern or the 'N' ("Read/append the next line of input into the pattern
space.") does. can you explain it for me?

[jungshik at Google, 2017/04/11 21:53:17]

    unitsNarrow{
    }

Given an input like the above, 

this line will match the first input line above and add it to the pattern space.

 
'/^    units(|Narrow|Short)\{$/ {

"N" will append the next input line ("    }") to the pattern space, making it
read like 
    unitsNarrow{\n    \}

Then, " /^    units(|Narrow|Short)\{\n    \}/" will match the pattern space and
will delete it starting a new cycle (processing the next input line. 

If the regex is not matched, the pattern space is just printed out and sed will
move on to process the next input line.

See  http://www.grymoire.com/Unix/Sed.html#uh-51 that I found after writing the
above. :-) 

+      }' ${i}
   done
 }
 
 # big5han and gb2312han collation do not make any sense and nobody uses them.
 function remove_legacy_chinese_codepoint_collation {
   echo "Removing Big5 / GB2312 / UniHan collation data from Chinese locale"
   target="${dataroot}/coll/zh.txt"
   echo "Overwriting ${target}"
   sed -r -i '/^        (uni|big5|gb2312)han\{$/,/^        \}$/ d' ${target}
 }
@@ skipping 12 matching lines @@
 filter_currency_data
 filter_region_data
 remove_legacy_chinese_codepoint_collation
 filter_unit_data
 
 # Chromium OS needs exemplar cities for timezones, but not Chromium.
 # It'll save 400kB (uncompressed), but the size difference in
 # 7z compressed installer is <= 100kB.
 # TODO(jshin): Make separate data files for CrOS and Chromium.
 #remove_exemplar_cities