Difference between revisions of "User:Tephlon"

Revision as of 22:29, 29 August 2007

Very well then, I'll write something here.

Who, me?

I'm tephlon, or rather, that's my nick. Why tephlon? Well, my wife seems to think that no problems ever stick on me, so she calls me "teflonmannen", which is "the teflon man" in Swedish. My own interpretation is that I have a non-stick memory. The "ph" instead of "f" is just... Well, I don't know. Leet? Whatever.

My real name is Stefan, and I'm born in 1976. I'm the maintainer of the Swedish translation since September 2004, and, actually, one of the moderators on the Translations & Internationalization forum, even though I don't usually do or say very much on the forums. I live in Göteborg -- which is "Gothenburg" in, at least, English-speaking countries -- in Sweden, which really is called Sverige. And that's not Switzerland. But you knew that. I hope.

In 2004 I started my fourth year as a PhD student in astrophysics, and had for some time been really fed up with it. At some point during late spring/early summer I was looking through the games at HappyPenguin.org, and I found Wesnoth. (Well, evidently I first saw Wesnoth June 15 that year. I found an old post of mine on LinuxQuestions.org from June 17 where I mentioned that I found Wesnoth "two days ago" :)) I liked it a lot from the start and recommended it to my girlfriend (now my wife :)), who, to my great astonishment, didn't hate computers so much that she couldn't appreciate a good game. So we started playing.

In August the same year, I looked into the Translations forum and noticed that the translation was pretty much unmaintained, and since I couldn't care less about my project I decided to try and do something about the translation. In a short time, the Swedish translation team, which basically consisted of me and Sanna, managed to do some great work on the translation. In fact, the Swedish translation was the first translation at 100%, and on September 11 we could proudly announce the 0.8.4 release in Swedish.

Late in December 2004 I quit my grad student position due to complete disinterest, and was without a job. During the unemployment period, which lasted for 8 months from the beginning of 2005, there was of course lots of time for the translation work.

In October 2005 I started working at an IT security company, and since then, life has not quite been the same. On April 14th 2006, me and my girlfriend got married. She was pregnant (no, we did not marry because of that, we married because we wanted to), and on August 19th our daughter, Amanda, was born.

So, I guess it is quite understandable that the Swedish translation work is a bit slow. It is most often only me working on it, even though new, and some old :), translators drop by and do some work on it. I have started to think that it is time to realize that I'm in another phase of my life, and that I should let the maintainership go to someone else. Don't know what's best though, hanging around until a translator shows up who is willing to take over, or just leave things and hope that someone will take over eventually?

Well, for now I guess I stick around for a while longer...

Thoughts on Translations

I've been meaning to put down my thoughts on translating a game such as this, but this seems a bit harder than I thought.

From the start, I've wanted the translation to be consistent throughout all the text domains. This might seem obvious, but it's harder than it sounds. The msgids in the po-files often come ouf of context, and when playing through a campaign it's not uncommon to stumble over some dialog which sounds really strange. So, I've come to view the translation process as three intertwined phases, or maybe sub-processes; bulk translation, proof-reading, and consistency checking.

Bulk translation

First, there is the work of getting the "bulk text" down. Choosing the word "bulk" might seem a bit condescending, but it's really not, it's just what it is. It is hard, and often boring, work; not very rewarding. During this phase the aim is just to get the strings translated; just going through one string after another, and translate.

Proof-reading

The second part is the proof-reading. Everything which is committed has been proofread at least once. This can be pretty quick, but sometimes one just get stuck on some odd passage. At times it can be weeks before a translation of a single sentence, or even word, is finished, because of the three criteria a msgstr has to fulfill:

It has to have the same meaning as the msgid.
It has to sound good.
It has to be consistent with the translation as a whole.

The first point can be discussed forever; is a more or less literal translation the best, or a complete rewrite which in the end conveys the same message? The Swedish translation goes something in between, and is really dependent on the second point. If a literal translation sounds good, it ought to be used. Sometimes this is not possible, however, so the passage translated has to be reorganized. Then this has to sound good.

What does "sound good" mean, then? Well, firstly, the translation has to use expressions which are actually used in the language one is translating to. Secondly, it has to be written in a way which displays the "rhythm" in what is actually written or perhaps rather -- if it is a dialog -- spoken. If it is a translation of a dialog between two or more characters, the translation has to sound like someone's actually talking.

Let me elaborate...

One thing, which I have to stress -- and this is extremely important -- is that one has to reflect on how people in general express themselves. One has to ask oneself, do people really talk this way? For instance, in English (or at least in Wesnoth campaigns :)) it is quite common to write passages like "this land", "this village", and so on. In Swedish, this would be literally translated as "detta land" and "denna by". In a campaign narrative this is okay, but, BUT, do people talk this way? No, people say "det här landet", "den här byn". Here one also has to think about how the people in a campaign talk. The Elvish Lord might say "detta land", but the average soldier would not.

However, one mustn't take this too far. The passages must still be readable, and one should still use correct language. There are some things that I have avoided at all costs, even though people "talk" this way. Take, for instance, the nominative "they" and the accusitive/dative "them". This is "de" and "dem" in Swedish. (Roughly at least, there might be exceptions, but I can't think of any at the moment.) However, people always say "dom" when they talk, for both these words. Which of cause leads to that people can't distinguish between the two forms. Well, here's a trick for the Swedes: replace "de" with "vi" and "dem" with "oss", and taste the sentence. How does it sound?

Other examples are the Swedish accusative and dative forms "mig" (Eng. "me") and "dig" (Eng. "you"). This could also be written "mej" and "dej". This, we don't do in the Swedish translation; we use "mig" and "dig".

However, there are exceptions. Of course :) Swedish "ska"/"skall", "sådan"/"sån", "någon"/"nåt". Here I have choosen the form that most suits the person talking.

Ah yes, one more thing. When someone says "I will yada yada", I have often seen this translated as "Jag kommer att yada yada". In the majority of cases it is more appropriate to translate this as "Jag ska(ll) yada yada".

Now, back to the matter at hand...

Consistency checking

The third point is more administrative, since it is basically just to look up how a certain string (for instance unit names) has been translated before. At times this too can be quite troublesome, for instance when it comes to words like Guard, Guardsman, Warder and Sentinel, since Swedish has a hard time distinguishing between these.

Lather, rince, repeat

As I'm sure you understand, these three phases don't come linearly. They have to be mixed, minced and reiterated. The perfect translation is the one where you can't tell what is the original text and what is the translation; when you can hear a dialog as though someone is speaking inside your head; when you don't even think about it; when you don't notice what you read; when it effortlessly brings an image to mind, and enhances your own imagination of what is actually happening.

Translation tools

During the time I've maintained the translation, I've constantly developed and improved a few scripts that help me do this. I have two main scripts where the first one (wsync) syncs my local source repository with the main repository, and the second (wcommit) picks out what I've changed since the last time I committed anything. To completely understand them, I think I have to describe my directory structure first...

Well, in my home directory ($HOME), I have a directory called Wesnoth, in which I do all Wesnoth-related work. This directory is called $ROOTDIR in the scripts below, as you can see. In the Wesnoth directory I have a whole bunch of stuff, but the two most important directories are Source and Translations.

In the Source directory I have all the Wesnoth sources, divided into Wesnoth-trunk, Wesnoth-1.2, Wesnoth-1.0 and Wescamp. This is where the "svn update" command changes stuff in the treesync and quicksync functions in the wsync script.

In the Translations directory I use these four directories as well (Wesnoth-trunk, Wesnoth-1.2, Wesnoth-1.0 and Wescamp), but here they contain one directory per domain, which in turn contains the po file, sv.po. These are the ones I edit when I translate. Furthermore, the Translations directory contains the Commit directory, where all tarballs that the wcommit script creates end up. Also, there is a .LastCommit directory, which keeps track of my latest commits, so that I don't commit the same stuff twice.

There is one more script that I want to show here, the pomerge script. I have told the Swedish translators to make sure they "mark" the strings they add or alter when they translate, so that I know what I have to proof-read. This is also written on the SwedishTranslation wiki page. However, I should really update the info there, since there is a much simpler way: the pomerge script. This script coldly assumes that translators do not change in existing translations, they only make untranslated strings translated, and then send updated files to me (as maintainer and proof-reader). The script does the following.

Filter out (with msgattrib) all translated strings out of the "draft" file sent from a translator, and mark all strings as "fuzzy".
Filter out (with msgattrib) all translated strings (also the fuzzy ones) out of the "main" file (the one I work on (with proof-reading or other) or the SVN file).
Merge (with msgcat) these two "cores". Where the main core and the draft core overlap, the main core is used (hence disregarding any corrections made by a translator to an already existing msgstr).
Merge (with msgmerge) the new core file with the old main file, or possibly the pot file.

The result is an updated main file, in which all translator-provided translations are set as "fuzzy". Proof-read and unfuzzy them!

wsync

!/bin/sh

ROOTDIR=$HOME/Wesnoth

SRCDIR=$ROOTDIR/Source
WORKDIR=$ROOTDIR/Translations


[ "x$ROOTDIR" == x ] && exit 1
[ "x$SRCDIR" == x ] && exit 1
[ "x$WORKDIR" == x ] && exit 1

[ -d $SRCDIR ] || mkdir -p $SRCDIR



trap cleanup SIGINT



cleanup () {
    echo -en "\nExiting: running svn cleanup for $SRCDIR/$BRANCH... " && ( cd $SRCDIR/$BRANCH && svn cleanup ) && echo "Done."
    exit 9
}



treesync () {
   [ $# -eq 2 ] || exit 1

   REPOS="$1"
   BRANCH="$2"

   echo -e "Syncing \033[01;32m$BRANCH\033[0m with \033[01;33m$REPOS\033[0m:\n"

   if [ -d $SRCDIR/$BRANCH ] ; then
      ( cd $SRCDIR/$BRANCH && svn update ) \
         || ( echo -e "\nRunning svn cleanup and svn update:\n" && cd $SRCDIR/$BRANCH && svn cleanup && svn update )
   else
      ( cd $SRCDIR && svn checkout $REPOS $BRANCH )
   fi

   find "$SRCDIR/$BRANCH" -type f -name '*.mine' | while read MINE ; do
      svn revert "${MINE%*.mine}"
   done

   echo
}



quicksync () {
   [ $# -eq 2 ] || exit 1

   REPOS="$1"
   BRANCH="$2"

   echo -e "Quick-syncing \033[01;32m$BRANCH\033[0m with \033[01;33m$REPOS\033[0m:\n"

   if [ -d $SRCDIR/$BRANCH/po/ ] ; then
      ( cd $SRCDIR/$BRANCH/po/ && svn update ) \
         || ( echo -e "\nRunning svn cleanup and svn update:\n" && cd $SRCDIR/$BRANCH/po/ && svn cleanup && svn update )
   else
      echo -e "\nRun a complete sync first - not a quick one!\n"
      exit 2 
   fi

   echo
}



posync () {
   local BRANCH="$1"
   local REPLACE="$2"

   local POFILE
   local PODIR
   local POTFILE
   local DOMAIN
   local ACTION

   for POFILE in $(find $SRCDIR/$BRANCH -name 'sv.po' | sort -s) ; do
      PODIR=${POFILE%*/sv.po}
      POTFILE="$(find $PODIR/ -name '*.pot')"
      DOMAIN="$(echo ${POTFILE##*/} | sed 's|\.pot$||')"

      if [ $(echo $POTFILE | grep -c .) -eq 0 ] ; then
         echo "No pot file found in $PODIR."
      else
         printf "\033[01;32m%-42s\033[0m" "$BRANCH/$DOMAIN"

         [ -d $WORKDIR/$BRANCH/$DOMAIN ] || mkdir -p $WORKDIR/$BRANCH/$DOMAIN

         cp $POTFILE $WORKDIR/$BRANCH/$DOMAIN/sv.pot

         ACTION=r

         if cmp -s $WORKDIR/$BRANCH/$DOMAIN/sv.po $POFILE && cmp -s $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po $POFILE ; then
            ACTION=n

         elif [ $WORKDIR/$BRANCH/$DOMAIN/sv.po -nt $POFILE ] && ! cmp -s $WORKDIR/$BRANCH/$DOMAIN/sv.po $POFILE ; then
            printf "\r\033[01;32m%-42s\033[0m %s" "$BRANCH/$DOMAIN" \
               "Work po-file has been modified. [R]eplace, [m]erge or [s]kip? "
            read ACTION
            ACTION=$(echo ${ACTION:0:1} | tr '[A-Z]' '[a-z]')
            while ! echo $ACTION | egrep -q "^(r|m|s)$" ; do 
               printf "\r\033[01;32m%-42s\033[0m %s" "$BRANCH/$DOMAIN" \
                  "Work po-file has been modified. [R]eplace, [m]erge or [s]kip? "
               read ACTION
               ACTION=$(echo ${ACTION:0:1} | tr '[A-Z]' '[a-z]')
            done
         fi

         if [ "x$ACTION" == "xn" ] ; then
            # This is the same as the "r" action, just to keep timestamps/attributes in order

            printf "\r\033[01;32m%-42s\033[0m \033[1m%-15s\033[0m" "$BRANCH/$DOMAIN" "No change"

            if [ -f $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po ] ; then
               cp $POFILE $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po
            fi

            cp $POFILE $WORKDIR/$BRANCH/$DOMAIN/sv.po

         elif [ "x$ACTION" == "xr" ] ; then
            printf "\r\033[01;32m%-42s\033[0m \033[01;32m%-15s\033[0m" "$BRANCH/$DOMAIN" "Replacing..."

            if [ -f $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po ] ; then
               cp $POFILE $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po
            fi

            cp $POFILE $WORKDIR/$BRANCH/$DOMAIN/sv.po

         elif [ "x$ACTION" == "xm" ] ; then
            printf "\r\033[01;32m%-42s\033[0m \033[01;33m%-15s\033[0m" "$BRANCH/$DOMAIN" "Merging..."

            if [ -f $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po ] ; then
               msgmerge -N -q --update $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po $POTFILE
               touch -m -r $POFILE $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po
            fi

            if [ -f $WORKDIR/$BRANCH/$DOMAIN/sv.po ] ; then
               msgmerge -N -q --update $WORKDIR/$BRANCH/$DOMAIN/sv.po $POTFILE
            else
               [ -d $WORKDIR/$BRANCH/$DOMAIN ] || mkdir -p $WORKDIR/$BRANCH/$DOMAIN
               msgmerge -N -q $POFILE $POTFILE > $WORKDIR/$BRANCH/$DOMAIN/sv.po
            fi
            touch -m -r $POFILE $WORKDIR/$BRANCH/$DOMAIN/sv.po 

         else
            printf "\r\033[01;32m%-42s\033[0m \033[01;31m%-15s\033[0m" "$BRANCH/$DOMAIN" "Skipping!"
         fi

         msgfmt --statistics $WORKDIR/$BRANCH/$DOMAIN/sv.po -o /dev/null
      fi
   done

   echo
}



echo

SVNSYNC=treesync
DO_POSYNC=true

while [ "x$1" != "x" ] ; do
    if echo "$1" | grep -q "q" ; then
        SVNSYNC=quicksync
    fi
    if echo "$1" | grep -q "x" ; then
        DO_POSYNC=false
    fi

    shift
done


$SVNSYNC http://svn.gna.org/svn/wesnoth/trunk Wesnoth-trunk
$SVNSYNC http://svn.gna.org/svn/wesnoth/branches/1.2 Wesnoth-1.2
treesync svn://svn.berlios.de/wescamp-i18n Wescamp

$DO_POSYNC || exit 0 

posync Wesnoth-trunk 
posync Wesnoth-1.2 
posync Wescamp

wcommit

#!/bin/sh

ROOTDIR=$HOME/Wesnoth

SRCDIR=$ROOTDIR/Source
WORKDIR=$ROOTDIR/Translations
COMMITDIR=${WORKDIR}/Commit


[ "x$ROOTDIR" == x ] && exit 1
[ "x$SRCDIR" == x ] && exit 1
[ "x$WORKDIR" == x ] && exit 1
[ "x$COMMITDIR" == x ] && exit 1

[ -d $COMMITDIR ] || mkdir -p $COMMITDIR
[ -d $COMMITDIR/Old ] || mkdir -p $COMMITDIR/Old

mv $COMMITDIR/*gz $COMMITDIR/Old 2> /dev/null
mv $COMMITDIR/*bz2 $COMMITDIR/Old 2> /dev/null

TEMPFILE=$(mktemp -t po-XXXXXX) || exit 2

TODAY=$(date +%Y%m%d)




################################################################################
# The podiff function below is a hideous hack to make sure that the work 
# po-file really differs from the source and also from what was last committed.
# The function returns false if the header (msgid "" on top of the po-file) and 
# obsolete (#~) entries but nothing else is different. 
#
# A lot of the ugliness below comes from the crippledness of gettext's msg-
# functions...
#
# So, we want to compare the work po-file with what was last committed and the
# source, but not take differing headers or obsolete entries into account. How
# do we do this? Or rather, how do we _have_ to do it?
# 
# To be able to omit the headers we _have_ to use msgcomm as this is the only 
# gettext function that has an option for this. Then, to not take obsolete 
# entries into account, we have to use the msgattrib function, as this is the 
# only gettext function that takes _this_ into account.
################################################################################

podiff () {
    WORKPO="$1"
    SOURCEPO="$2"
    LASTCOMMITPO="$3"

    ################################################################################
    # Check first if the work po-file has not been altered at all. If it hasn't, 
    # it is completely unnecessary to continue. STATE is true if the work po-file 
    # should be committed.
    ################################################################################

    if diff -q $WORKPO $LASTCOMMITPO 2>&1 > /dev/null || diff -q $WORKPO $SOURCEPO > /dev/null ; then
        STATE=false
    else
        CMPDIR=$WORKDIR/.Compare
        ERRORLOG=/dev/null

        WORKCLONE=$CMPDIR/work-clone.po
        SOURCECLONE=$CMPDIR/source-clone.po
        LASTCOMMITCLONE=$CMPDIR/lastcommit-clone.po

        WORKSTRIPPED=$CMPDIR/work-stripped.po
        SOURCESTRIPPED=$CMPDIR/source-stripped.po
        LASTCOMMITSTRIPPED=$CMPDIR/lastcommit-stripped.po

        if [ -d $CMPDIR ] ; then
            rm $WORKCLONE    $SOURCECLONE    $LASTCOMMITCLONE    2> /dev/null
            rm $WORKSTRIPPED $SOURCESTRIPPED $LASTCOMMITSTRIPPED 2> /dev/null
        else
            mkdir -p $CMPDIR
        fi

        ################################################################################
        # First, we create clones of these three files, since msgcomm complains if the 
        # same file is used as input files.
        #
        # Next, we use msgcomm with the --omit-header option on the original po-file and
        # its clone. It is convenient to remove the obsolete entries with msgattrib and
        # --no-obsolete in this step, as this can be done in a pipe.
        # 
        # There is one important fact here. If --omit-header is used, msgcomm can't cope
        # with extended characters. The same is true for msgattrib and --no-obsolete.
        # This can be remedied by using the --escape/-E option on both msgcomm and 
        # msgattrib.
        ################################################################################

        cp $WORKPO       $WORKCLONE
        cp $SOURCEPO     $SOURCECLONE
        cp $LASTCOMMITPO $LASTCOMMITCLONE

        msgcomm -E --omit-header $WORKPO       $WORKCLONE       2> $ERRORLOG | msgattrib -E --no-obsolete -o $WORKSTRIPPED       2> $ERRORLOG
        msgcomm -E --omit-header $SOURCEPO     $SOURCECLONE     2> $ERRORLOG | msgattrib -E --no-obsolete -o $SOURCESTRIPPED     2> $ERRORLOG
        msgcomm -E --omit-header $LASTCOMMITPO $LASTCOMMITCLONE 2> $ERRORLOG | msgattrib -E --no-obsolete -o $LASTCOMMITSTRIPPED 2> $ERRORLOG

        ################################################################################
        # NOW we can do the real comparison. diff -q returns true if the files are 
        # _equal_. STATE is true if the work po-file should be committed.
        ################################################################################

        if diff -q $WORKSTRIPPED $LASTCOMMITSTRIPPED 2>&1 > /dev/null || diff -q $WORKSTRIPPED $SOURCESTRIPPED > /dev/null ; then
            STATE=false
        else 
            STATE=true
        fi

        rm $WORKCLONE    $SOURCECLONE    $LASTCOMMITCLONE    2> /dev/null
        rm $WORKSTRIPPED $SOURCESTRIPPED $LASTCOMMITSTRIPPED 2> /dev/null
    fi

    $STATE
}



pocommit () {
   BRANCH="$1"

   LBRANCH=$(echo "$BRANCH" | tr '[:upper:]' '[:lower:]')

   NUMBER=$(find $COMMITDIR/Old/ -name "$LBRANCH-sv-$TODAY-*.svndiff.bz2" -printf '%f\n' | sort -g | tail -1 | rev | cut -f1 -d'-' | rev | cut -f1 -d'.')
   [ "x$NUMBER" == "x" ] && NUMBER=0
   NUMBER=$((NUMBER + 1))

   OUTFILENAME="${LBRANCH}-sv-${TODAY}-${NUMBER}.svndiff"
   OUTFILE="$COMMITDIR/$OUTFILENAME"

   > "$OUTFILE"

   echo -e "Checking \033[01;33m$BRANCH\033[0m: "

   for POFILE in $(find $SRCDIR/$BRANCH -name 'sv.po' | sort -s) ; do
      POTFILE="$(find ${POFILE%*/sv.po}/ -name '*.pot')"
      DOMAIN="$(echo ${POTFILE##*/} | sed 's|\.pot$||')"

      printf "    \033[01;32m%-42s\033[0m %-12s\r" "$DOMAIN" "Checking..."

      [ -d $WORKDIR/.LastCommit/$BRANCH/$DOMAIN ] || mkdir -p $WORKDIR/.LastCommit/$BRANCH/$DOMAIN
      [ -f $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po ] || touch $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po

      ################################################################################
      # I have forgotten why I do "-o $TEMPFILE" below instead of "--update". Let's 
      # ponder...
      #
      # The reason for doing this msgmerge in the first place, is to make the 
      # .LastCommit and the work po-files have correct (the same) line-wrapping, so as 
      # to make a proper comparison in the following long if statement.
      #
      # So, my guess is that I - once upon a time - realized that msgmerge doesn't 
      # touch the po-file if it is up-to-date with the pot-file, and I want to correct 
      # the line-wrapping regardless of whether the po-file matches the pot-file or 
      # not. In fact, it is irrelevant if the po- and pot-files are matching. What 
      # _is_ relevant, is to make a proper comparison between the .LastCommit po-file 
      # and the work po-file.
      #
      # What does line-wrapping have to do with anything? Well, kbabel doesn't wrap 
      # the msgstr lines at all, whereas msgmerge does line-wrapping. And every 
      # po-file that is submitted to the SVN is msgmerged - or at least it is bound to 
      # be when the next pot-update is being made.
      ################################################################################

      msgmerge -N -q -o $TEMPFILE $WORKDIR/$BRANCH/$DOMAIN/sv.po $POTFILE && mv $TEMPFILE $WORKDIR/$BRANCH/$DOMAIN/sv.po || exit 3
      msgmerge -N -q -o $TEMPFILE $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po $POTFILE && mv $TEMPFILE $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po || exit 3

      if [ -f $WORKDIR/$BRANCH/$DOMAIN/sv.po ] && podiff "$WORKDIR/$BRANCH/$DOMAIN/sv.po" "$POFILE" "$WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po" ; then
         cp $WORKDIR/$BRANCH/$DOMAIN/sv.po $POFILE
         cp $WORKDIR/$BRANCH/$DOMAIN/sv.po $WORKDIR/.LastCommit/$BRANCH/$DOMAIN/sv.po

         printf "    \033[01;32m%-42s\033[0m %-12s" "$DOMAIN" "Adding: "
         msgfmt --statistics $WORKDIR/$BRANCH/$DOMAIN/sv.po -o /dev/null
      else
         printf "    \033[01;32m%-42s\033[0m %-12s\n" "$DOMAIN" "No changes."
      fi
   done

   if ( cd $SRCDIR/$BRANCH/ && find . -type f -name 'sv.po' -exec svn diff {} \; >> "$OUTFILE" ) ; then
      ( cd $SRCDIR/$BRANCH/ && find . -type f -name 'sv.po' -exec svn revert {} \; > /dev/null )
   fi

   if [ -s "$OUTFILE" ] ; then
      bzip2 "$OUTFILE"
      echo "Changes added to ${OUTFILE}.bz2."
   else
      rm "$OUTFILE" 2> /dev/null
      echo "No changes in $BRANCH."
   fi

   echo
}




pocommit Wesnoth-trunk
pocommit Wesnoth-1.2
#pocommit Wesnoth-1.0
pocommit Wescamp

pomerge

#!/bin/sh


usage () {
    echo -e "\nUsage: $(basename $0) <draft po> <main po> [outfile]\n"
    exit 0
}

[ $# -ge 2 ] && [ $# -le 3 ] || usage

DRAFT="$1"
MAIN="$2"
[ "x$3" == "x" ] && OUT="/tmp/out.po" || OUT="$3"

if [ ! -f "$DRAFT" ] ; then
    echo "$DRAFT does not exist."
    exit 1
fi

if [ ! -f "$MAIN" ] ; then
    echo "$MAIN does not exist."
    exit 1
fi

if [ -f "$OUT" ] ; then
    echo -n "$OUT already exists. Overwrite? [yN]"
    read ANSWER
    [ "x$ANSWER" == "xy" ] || exit 1
fi



COREDRAFT="/tmp/coredraft.po"
COREMAIN="/tmp/coremain.po"
JOINEDCORES="/tmp/joinedcores.po"
TEMPOUT="/tmp/tempout.po"

rm "$COREDRAFT" "$COREMAIN" "$JOINEDCORES" 2> /dev/null

echo -en "Filtering draft...   \t"
msgattrib --set-fuzzy --translated -o "$COREDRAFT" "$DRAFT"
msgfmt --statistics -o /dev/null "$COREDRAFT"

echo -en "Filtering main...    \t"
msgattrib --translated -o "$COREMAIN" "$MAIN"
msgfmt --statistics -o /dev/null "$COREMAIN"

echo -en "Joining cores...     \t"
msgcat --use-first -o "$JOINEDCORES" "$COREMAIN" "$COREDRAFT"
msgfmt --statistics -o /dev/null "$JOINEDCORES"

echo -n "Merging... "
msgmerge --no-fuzzy-matching -q -o "$TEMPOUT" "$JOINEDCORES" "$MAIN"
echo "Done."

mv "$TEMPOUT" "$OUT"

echo

echo -en "Main file:   \t"
msgfmt --statistics -o /dev/null "$MAIN"

echo -en "Merged file: \t"
msgfmt --statistics -o /dev/null "$OUT"

rm "$COREDRAFT" "$COREMAIN" "$JOINEDCORES" /tmp/out.po 2> /dev/null