________________________________ From: owner-sv-ec@server.eda.org [mailto:owner-sv-ec@server.eda.org] On Behalf Of Daniel Mlynek Sent: Wednesday, September 10, 2008 4:55 AM To: sv-ec@server.eda-stds.org Subject: [sv-ec] 6.24.1 static cast after mantis 1384 fix There is bunch of inconsitency and unclerance after mantis 1384 fix LRM draft 6 : 6.24.1 Description of static cast: "If the expression is assignment compatible with the casting type, then the cast shall return the value that a variable of the casting type would hold after being assigned the expression. If the expression is not assignment compatible with the casting type, then if the casting type is an enumerated type, the behavior shall be as described as in 6.19.4, and if the casting type is a bit-stream type, the behavior shall be as described in 6.24.3." The sentence is still open - what should happened if dest and source is not assignment compatybile and casting type is not an enum not bit-stream type - is it possible? What about classes - are they bit stream - is not so obvious [DR] Assignments are illegal without assignment compatibility (6.22.3). Some classes are bit-stream types if certain conditions are met (6.24.3). Anything else is not defined and therefore not legal. 6.24.3: "An associative array type or class shall be illegal as a destination type." (this was allowed in 1800-2005 now draft 6 bans it) Both above are contradictory - becaue 1st allows to cast to a type which is assignment compatybile (which is possible for class handles - ie cast to super class) while 2nd forbids static cast to class type. IMHO there are two solutions : [DR] They are not contradictory. It is singling out a special case that needs to be made illegal since there is no semantic definition for creating the index to the associative array element. - cast to class should be fdisallowed in all cases - sentence from 6.24.3 should be moved to 6.24.1 - cast between class handles should be allowed - it should not perform any kind of check on compile/elab time - but on run time it should triggger fatal errors if casted types are not compatybile - this should be described in 6.24.3. [DR] This is what $cast does. Cast where source is a non-class bit stream class and dectination is a class handle should be forbidden in 6.24.3[DR] The purpose of defining classes as bit-stream types was specifically to allow packing/unpacking of class members with non-class bitstream types. I personally was against this as most classes will be disqualified by containing local or protected properties, as well as containing properties that the user would not want to include in their pack/unpack streams. I thought there was some restriction about the class as a destination type saying the class must be constructed before moving the bits. We defined this for the streaming operators, but not for bit-stream casts. Mantis 2473 added as an addition an example - what should be the behaviour of tool in below example class B; reg x; function f(); $display("B",x); endfunction endclass class A extends B; int y; function f(); $display("B",y); endfunctionendclass module top; A a;[DR] = new(); B b;[DR] = new(); [DR] A and B are both legal bit-stream class types. B has 1 bit, and A has 33 bits. initial begin b = 'B(a); /// should it be legal?? this is ok according to 6.24.1 but offends 6.24.3[DR] Illegal because the number of bits don't match a = 'A(b); /// should it be legal?? this offends both : 6.24.1 and 6.24.3[DR] Illegal because the number of bits don't match 'B(a).f(); /// should it be legal?? this show the only sensible usage of casting to base class type but is this ok according to syntax???[DR] not legal for a number of reasons. Syntax does not allow expressions in selects and cast to a destination class needs to be constructed. end endmodule DANiel -- This message has been scanned for viruses and dangerous content by MailScanner <http://www.mailscanner.info/> , and is believed to be clean. -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.Received on Wed Sep 10 07:07:23 2008
This archive was generated by hypermail 2.1.8 : Wed Sep 10 2008 - 07:07:53 PDT