4 files changed, 36 insertions, 3 deletions
diff --git a/src/GF/API.hs b/src/GF/API.hs
index 2d23da0f6..5a55f5b1f 100644
--- a/src/GF/API.hs
+++ b/src/GF/API.hs
@@ -26,6 +26,7 @@ import CMacros
 import Transfer
 import qualified Generate as Gen
 
+import Text (untokWithXML)
 import Option
 import Custom
 import ShellState
@@ -208,7 +209,10 @@ optLinearizeTree opts0 gr t = case getOptVal opts transferFun of
     | otherwise = return . unlines . map untok . optIntOrOne . linTree2strings mk g c
   g = grammar gr
   c = cncId gr
-  untok = customOrDefault opts useUntokenizer customUntokenizer gr
+  untok = if False ---- oElem (markLin markOptXML) opts 
+            then untokWithXML unt
+            else unt  
+  unt = customOrDefault opts useUntokenizer customUntokenizer gr
   optIntOrOne = take $ optIntOrN opts flagNumber 1
 
 {- ----
diff --git a/src/GF/Compile/ShellState.hs b/src/GF/Compile/ShellState.hs
index e1e64e85c..4b1e5a8f3 100644
--- a/src/GF/Compile/ShellState.hs
+++ b/src/GF/Compile/ShellState.hs
@@ -122,7 +122,8 @@ updateShellState opts sh ((_,sgr,gr),rts) = do
       a' = ifNull Nothing (return . head) $ allAbstracts cgr0
   abstr0 <- case abstract sh of
     Just a -> do
-      -- test that abstract is compatible
+      -- test that abstract is compatible --- unsafe exception for old? 
+      --- if True oElem showOld opts then return () else
       testErr (maybe True (a==) a') ("expected abstract" +++ P.prt a) 
       return $ Just a
     _ -> return a'
diff --git a/src/GF/Text/Text.hs b/src/GF/Text/Text.hs
index 2fbf97fd3..de29e9026 100644
--- a/src/GF/Text/Text.hs
+++ b/src/GF/Text/Text.hs
@@ -6,7 +6,25 @@ import Char
 -- elementary text postprocessing. AR 21/11/2001
 -- This is very primitive indeed. The functions should work on
 -- token lists and not on strings. AR 5/12/2002
+-- XML hack 14/8/2004; not in use yet
 
+-- does not apply untokenizer within XML tags --- heuristic "< "
+-- this function is applied from top level...
+untokWithXML :: (String -> String) -> String -> String
+untokWithXML unt s = case s of
+  '<':cs@(c:_) | isAlpha c -> '<':beg ++ ">" ++ unto (drop 1 rest) where 
+                  (beg,rest) = span (/='>') cs
+  '<':cs -> '<':unto cs ---
+  [] -> []
+  _ -> unt beg ++ unto rest where
+               (beg,rest) = span (/='<') s
+ where
+   unto = untokWithXML unt
+
+-- ... whereas this one is embedded on a branch
+exceptXML :: (String -> String) -> String -> String
+exceptXML unt s = '<':beg ++ ">" ++ unt (drop 1 rest) where 
+  (beg,rest) = span (/='>') s
 
 formatAsTextLit :: String -> String
 formatAsTextLit = formatAsText . unwords . map unStringLit . words 
@@ -62,3 +80,13 @@ unStringLit s = case s of
   _ -> s
  where
    strlim = (=='\'')
+
+concatRemSpace :: String -> String
+concatRemSpace = concat . words
+{-
+concatRemSpace s = case s of
+  '<':cs -> exceptXML concatRemSpace cs
+  c : cs | isSpace c -> concatRemSpace cs
+  c :cs -> c : concatRemSpace cs
+  _ -> s
+-}
+\ No newline at end of file
diff --git a/src/GF/UseGrammar/Custom.hs b/src/GF/UseGrammar/Custom.hs
index 3fb386c79..60c906fa0 100644
--- a/src/GF/UseGrammar/Custom.hs
+++ b/src/GF/UseGrammar/Custom.hs
@@ -301,7 +301,7 @@ customUntokenizer =
   ,(strCI "code",      const $ formatAsCode)
   ,(strCI "textlit",   const $ formatAsTextLit)
   ,(strCI "codelit",   const $ formatAsCodeLit)
-  ,(strCI "concat",    const $ concat . words)
+  ,(strCI "concat",    const $ concatRemSpace)
   ,(strCI "glue",      const $ performBinds)
   ,(strCI "reverse",   const $ reverse)
   ,(strCI "bind",      const $ performBinds) -- backward compat