Minutes of ISAC meeting held via telecom on 06 January 2005 Present: Peter Ashenden, Jim Lewis, Larry Soule, Chuck Swart, Ajay Varikat Absent: Deepak Pant Next Meeting: Thursday 20 January 2005, 7pm Pacific Time TOPIC: IR2004 Definition of SLA doesn't make sense The discussion led to the conclusion that any practical use of SLA required that the operator be overloaded. Therefore, the analysis, ( no change to the language) was accepted and this IR is now ISAC Approved. ACTION: Chuck to forward to VASG chair. TOPIC: IR2049 Circular definition of an event on a signal In examining the motivation behind the circular definition it was found useful to have access to the VHDL93 LCSs.Note added: The LCSs are located at eda.org. ACTION: Ajay to analyze this IR. Chuck or Peter to make LCSs accessible via web. TOPIC: IR2050 Definition of S'Last_Value was apparently broken in 1993 Discussion revealed that the definition of S'Last_Value for composites was changed in VHDL-93 to reflect the last value of the composite when any element had a change in value. In VHDL-87 the last value of a composite was the aggregate of the element last values. We discovered that the famous "Annotated VHDL-93 LRM" existed on the eda web site, and Peter made it accessible to isac members. Discussion revealed that there are three properties of S'Last_Value: 1. It must be sensitive to changes which occur during delta cycles. 2. The last value of a composite must be the value of the composite before any change to any element. It is not the aggregate of subelement scalar last values. 3. S'Last_Value must defined prior to the first change in S. The VHDL-2002 definition satisfies the first two requirements but not the third. ACTION: Peter to analyze. TOPIC: IR2051 Path_name and instance_name do not allow for protected types ACTION: Chuck to analyze. TOPIC: IR2052 Path_name and instance_name don't deal with operator symbols This problem might be resolved by replacing simple_name with designator. ACTION: Chuck to analyze. TOPIC: IR2053 Minor Typos in VHDL 2002 part 2 ACTION: Peter to send in more typos. TOPIC: IR2054 Individ. assoc. rules for array formal are not valid The IR points out that there are two sets of rules for determining the appropriate index ranges when subelement association is used with unconstrained formals. The first set makes sense--the second appears to have been added in error. Also, The rules for type matching for subelement association together with type conversions are confusing if not wrong. ACTION: Larry to analyze. TOPIC: IR2023 Add predefined array types for integer, boolean, real and time Peter contacted the VHDL news group and the few responders indicated that there was little or no concern about potential compatibility problems. It might be a good idea to also contact VHDL 200x groups about this issue, although it seems that at some of these types have already been incorporated into the proposed revision to package standard. ACTION: Peter to update analysis, followed by ISAC vote. TOPIC: IR2025 "Generate" for sequential code This functionality is most useful if implemented at the lexical level, otherwise syntax errors could be analyzed in dead code. ACTION: Peter to update analysis, then IR will be forwarded to M&P TOPIC: IR2026 Upward propagating parameters This is a M&P issue. ACTION: Chuck to forward. TOPIC: IR2028 Clarify simulation cycle. Full analysis of this is postponed until input is received from VHPI.