Windows: use wide char interfaces

Exchange file names and command line parameters with the system using
wchar_t interfaces: allows preserving values which can be reversibly
transcoded in the current multibyte charset (which can't be UTF-8). Store
all file paths internally in UTF-8

src/filters/rclexec1.py

@@ -61,7 +61,7 @@ class Executor(RclBaseHandler):
             return True, postproc.wrapData()
         else:
             try:
-                fullcmd = cmd + [filename]
+                fullcmd = cmd + [rclexecm.subprocfile(filename)]
                 proc = subprocess.Popen(fullcmd,
                                         stdout = subprocess.PIPE)
                 stdout = proc.stdout

src/filters/rclexecm.py

@@ -29,13 +29,25 @@ import shutil
 import getopt
 import rclconfig
 
-PY3 = sys.version > '3'
+PY3 = (sys.version > '3')
+_mswindows = (sys.platform == "win32")
 
 def makebytes(data):
     if type(data) == type(u''):
         return data.encode("UTF-8")
     return data
 
+def subprocfile(fn):
+    # On Windows PY3 the list2cmdline() method in subprocess assumes that
+    # all args are str, and we receive file names as UTF-8. So we need
+    # to convert.
+    # On Unix all list elements get converted to bytes in the C
+    # _posixsubprocess module, nothing to do
+    if PY3 and _mswindows:
+        return fn.decode('UTF-8')
+    else:
+        return fn
+
 my_config = rclconfig.RclConfig()
 
 ############################################

src/filters/rclpdf.py

@@ -479,7 +479,8 @@ class PDFExtractor:
             print("RECFILTERROR HELPERNOTFOUND pdftotext")
             sys.exit(1);
 
-        self.filename = params["filename:"]
+        self.filename = rclexecm.subprocfile(params["filename:"])
+
         #self.em.rclog("openfile: [%s]" % self.filename)
         self.currentindex = -1
         self.attextractdone = False