ISAC: Minutes of meeting held on 04 May 2006

Please review these minutes.

Chuck Swart



Minutes of ISAC meeting held via telecom on 04 May 2006

Present: Peter Ashenden, Larry Soule, Lance Thompson, Ajay Verikat

Absent: Jim Lewis

Next Meeting Thursday May 25, 2006, 7 pm Pacific Daylight Time
             (Friday May 26, 2006, 2 am GMT)

Topic: IR 2092 Type conversions don't allow for null arrays

Several issues were uncovered.

a. When the type mark denotes an unconstrained array subtype we need a
uniform rule which is compatible with IR 2075 for generating an
appropriate null index range for cases in which the corresponding
expression has a null range. The most reasonable is to pick the
direction from the index base type as in IR 2075, but for null ranges
to pick the left bound of the index constraint to be the right value
of the base type and the right bound to be the left value of the base
type.

b. In section 7.3.5 a Sentence needs to be rewritten: "If the type
mark denotes an unconstrained array type, then, for each index
position <where the index range is non-null>, a check is made that eh
bounds of the result belong to the corresponding index subtype of the
target type.

c. There are some esoteric cases in which a null array can't be
formed, because there are no applicable index values to create a null
range.  This problem can be solved by requiring that all enumeration
types have at least two values.

d. IR 2038 item 10 points out a problem in section 7.3.1 Literals
which shows that the rules aren't always applicable for determining
null array bounds corresponding to a string literal. IR 2092 will also
incorporate this case.

ACTION: Chuck to revise, then all to vote.

TOPIC: Managing LRM changes.

The LRM is currently being revised for the next language version.
Certain sections are very unstable, reflecting a multitude of language
changes. Therefore, it is sometimes very difficult for IR analyzers to
produce appropriate rewording. So it was decided that for stable parts
of the LRM, wording changes will be based on 1076-2002 (Instead of
change draft standards). For unstable parts of the LRM the IR analyzer
will produce clear intent for proposed changes, but will not suggest
precise wording.

ACTION: All to take into account when analyzing IRs.

TOPIC: IR 2093  Static type conversions and qualified expressions

The proposed changes include definitions of locally and globally
static unconstrained array types. If these definitions are accepted,
then there will be a need to significantly revise the semantics of
case expressions. In addition, there may be other problems, so the
proposal will be revised to remove these troublesome cases.

ACTION: Chuck to revise, then all to vote.

TOPIC: IR 2038 Minor semantic errors

Several items were discussed. These include:


Item 3. Inaccurate wording on assignment where target is (index into)
function call returning an access type.

One possibility is to simply remove the sentence in question.

Ajay (and others) want to better understand the case in question.
Chuck has potential rewording and also will produce an example for
this particular construct.

Item 6. Technical problem with alias and fact that types are often
anonymous.

IR 2058 covers this.

Items 13, 14 possible lack of matching elements rules in some cases
of signal, variable assignments.

IR 2013 may cover this.

ACTION: Chuck to supply example and possible rewording to Ajay. Ajay
to revise, all to read the revised IR.


ITEM: Possible ISAC meeting at DAC

DAC is July 28-28(?). We would like to see if a meeting is feasible.

ACTION: All to consider.