Add doc

MPL.md

@@ -339,7 +339,9 @@ If the current token is still null, then the search is successful if the file's
 
 - If the transient's first search is a TokenSearch of a non-transient identifier
 
-Then the search is successful if the current token matches the identifier. 
+
+
+Then the search is successful if the current token matches the identifier.
 
 This token can be declared after the transient line. That's because the transient token can be build from anywhere, and thus the current token can be anything. Plus, there is no risk of a left-recursion since the token is not build here, it is just checked.
 
@@ -353,6 +355,15 @@ Thus we need to create the referenced token with the current one. To prevent lef
 
 Other tokens are build exactly as it would be as a classic token.
 
+### Transient with only previous search
+
+If a TokenSearch referencing a transient token is preceded with a '#', then lines are read up to the current line, instead of the transient token's declaration line. 
+
+	Identifier = #Token
+
+This is useful when some token can only use those declared before him. It can only be used on transient token. It can be used on the first search, even though it doesn't have any effect.
+
+
 ### Strings
 
 String are considered as transient, as they can't be a current token and generally follow the same construction rules as transient tokens do.

src/compiler/Compiler.java

@@ -176,7 +176,7 @@ public class Compiler
 				if (!found)
 					report(new Info(InfoModel.tokenUndeclared, token));
 				else if (first && foundAfterThisLine)
-					if ((currentLine.header.trans != null) || !tokenTransient)
+					if ((currentLine.header.trans != null) == tokenTransient)
 						report(new Info(InfoModel.firstTokenUndeclared, token));
 				
 				if (!tokenTransient && tokenSearch.trans != null)