pogrep may output two po headers

[claudep]

Version: trunk

When I'm running pogrep (specifically with invert-match), I'm getting a resulting po file with two headers. We already discussed it with Friedel on IRC. A quick-and-dirty patch:

--- /home/users/claudep/pogrep.py   2012-03-13 12:21:34.000000000 +0000
+++ python-packages/lib/python2.6/dist-packages/translate/tools/pogrep.py   2011-06-22 17:47:35.000000000 +0000
@@ -216,6 +216,7 @@
     def filterfile(self, thefile):
         """runs filters on a translation file object"""
         thenewfile = type(thefile)()
-        thenewfile.units = []
       thenewfile.setsourcelanguage(thefile.sourcelanguage)
       thenewfile.settargetlanguage(thefile.targetlanguage)
       for unit in thefile.units:

Another try at higher level:

--- translate/storage/base.py   2011-05-15 20:14:19.000000000 +0200
+++ /home/claude/.../translate/storage/base.py  2012-03-13 14:17:21.759808031 +0100
@@ -543,6 +543,9 @@
         @param unit: The unit that will be added.
         """
         unit._store = self
-        # Prevent two header units in the same store
-        if unit.isheader() and len(self.units) and self.units[0].isheader():
-            self.units[0] = unit
-        else:
-            self.units.append(unit)
- 
     self.units.append(unit)
  
   def addsourceunit(self, source):

[friedelwolff]

I also got this now with a simple grep (not inverted match).

[friedelwolff]

From a quick look, I think your second approach is good, except that it should only be done for PO clases (we have a few). So we could maybe put this in pocommon, but at least cpo.py currently also overrides addunit, so we might want to fix it there as well.

[claudep]

This is still an issue in 2.0. I see that Friedel fixed a similar issue for posegment in a4ac704 and 81ae442.