Need reviewer on addition of whitespace normalizer script

[Jonathan Gibbons]

Kelly,

Is there a reason you don't use "expand"?

For a while now I've been using the following simple script to fix the 
whitespace in my files before I commit a changeset.

-- Jon



#!/bin/sh

# Find and fix whitespace errors in files that would otherwise be caught 
and rejected by the
# OpenJDK Mercurial "jcheck" facility.
# 1. Tabs in files are expanded to spaces
# 2. Trailing whitespace is removed
# 3. Add final newline if it is missing

# With no args, script uses "hg status" to determine modified and new files.
# Otherwise, script scans files and directories given on command line 
and fixes
# *java, *.g, *.properties, *.xml files.

if [ $# = 0 ]; then
   files=$(hg status --modified --added --no-status)
else
   files="$*"
fi

find $files -name SCCS -prune -o \( -name \*.java -o -name \*.g -o -name 
\*.properties -o -name \*.xml \) -print |
     while read f ; do
         updated=0

         # check for tabs or trailing whitespace, fix if found
         if egrep '    |( $)' $f > /dev/null ; then
             expand $f | sed -e 's/[     ]*$//' > $f~ && mv $f~ $f
             updated=1
         fi

         # check for final newline, fix if not found
         if perl -ne 'END { exit 1 if $nl; } $nl = /\n$/' $f ; then
             echo >> $f
             updated=1
         fi

         # log update
         if [ $updated = 1 ]; then
             echo $f
         fi
     done




On 03/02/2012 03:41 PM, Kelly O'Hair wrote:
> A TAB takes you to a specific TAB spot, default is one TAB stop every 8 characters in a line.
> So the conversion is not just 'replace TAB with N characters'.
> This is one of the issues with TABs, they aren't as predictable as people might think, especially
> when mixed with spaces or placed in any location other than the beginning of the line all by
> themselves.
>
> It should not change the indenting, but that's with the assumption that the TABs follow the
> 8 character spacing. And that of course depends on how you are viewing the source. :^(
>
> -kto
>
> On Mar 2, 2012, at 3:04 PM, David Holmes wrote:
>
>> On 3/03/2012 3:12 AM, Kelly O'Hair wrote:
>>> I don't understand the question. It only changes TAB characters, removes trailing whitespace on lines,
>>> and duplicate blank lines at the end of the file.
>> I think the issue is what does it replace a TAB with? 4 spaces for JDK or 2 spaces for Hotspot?
>>
>> David
>>
>>> -kto
>>>
>>> On Mar 2, 2012, at 1:00 AM, Staffan Larsen wrote:
>>>
>>>> Does this handle the difference between indents in HotSpot (indent 2) vs the JDK (indent 4)?
>>>>
>>>> /Staffan
>>>>
>>>> On 1 mar 2012, at 22:32, Kelly O'Hair wrote:
>>>>
>>>>> Need reviewer. Adding the whitespace normalizer script as a convenience to the jdk developers.
>>>>>
>>>>> 6625113: Add the normalize and rmkw perl script to the openjdk repository or openjdk site?
>>>>> http://cr.openjdk.java.net/~ohair/openjdk8/normalizer-script/webrev/
>>>>>
>>>>> Probably a little history is warranted here. This script was originally written to normalize the
>>>>> whitespace in the jdk7 sources as they entered the Mercurial repositories in "changeset 0".
>>>>> It's been modified since then very slightly. I can't recall who wrote it (please speak up if you know)
>>>>> but it has been a valuable tool and I've had this CR to add it to the make/scripts directory for some time.
>>>>>
>>>>> The SCCS keyword removed (rmkw) was less useful, and I decided that it did not deserve being added.
>>>>>
>>>>> Why whitespace normalization? This was decided a long time ago when we had a raft of complaints from
>>>>> people viewing the sources with different tools and getting different views based on the TABs and trailing
>>>>> blanks or trailing newlines. So we decided to normalize on no TABs, no trailing blanks on lines, and
>>>>> no more than one blank line at the end of the file. This script was used to do that normalization.
>>>>>
>>>>> -kto
>>>>>