From LPMeissner@msn.com Thu Aug 1 20:29:47 1996 Received: from upsmot01.msn.com (upsmot01.msn.com [204.95.110.78]) by dkuug.dk (8.6.12/8.6.12) with ESMTP id UAA21269 for ; Thu, 1 Aug 1996 20:29:18 +0200 Received: from upmajb02.msn.com (upmajb02.msn.com [204.95.110.74]) by upsmot01.msn.com (8.6.8.1/Configuration 4) with SMTP id LAA25800 for ; Thu, 1 Aug 1996 11:21:32 -0700 Date: Thu, 1 Aug 96 18:32:28 UT From: "Loren Meissner" Message-Id: To: sc22wg5@dkuug.dk Subject: Preliminary minutes of Dresden WG5 meeting This is a very preliminary draft. I thought it might be useful to distribute something prior to the X3J3 meeting. Please let me know if I have spelled your name wrong or made gross errors in content. Please note that the formal actions of WG5 are contained in the Resolutions already distributed separately by David Muxworthy (WG5-N1219). See also the Convenor's informal report (Miles Ellis, dated 30 July 1996). WG5 MEETING - DRESDEN 22 TO 26 JULY 1996 ATTENDANCE: 25 delegates, 1 observer Convenor: Miles Ellis Austria: Gerhard Schmitt; David Schmitt (observer) Canada: Whitman Wright Germany: Manuela Zuern, Uwe Kuester, Wolfgang Walter, Karl-Heinz Rotthaueser, Andrea Coriand, Christian Weber, Michael Hennecke, Andrew Pitonyak Japan: Masayuki Takata, Minoru Tanaka Sweden: Lars Mossberg UK: Malcolm Cohen, Steve Morgan, Keng Low, David Muxworthy US: David Epstein, Jerrold Wagener, Kurt Hirchert, Baker Kearfott, Loren Meissner, Tom Lahey, Keith Bierman CONVENOR'S REPORT SC22 ballot on Technicnal Corrigendum 3 ends 3 October. Final ballot on Fortran 95 will begin as soon as processing of TC3 is completed. Convenor appointments for this meeting: Secretary, Meissner; Librarian, Hirchert; Drafting committee, Muxworthy, Weber, Wagner, Takata, Mossberg, and Schmitt. MINUTES of San Diego meeting were approved with minor typographical corrections. NATIONAL ACTIVITY REPORTS were presented by Austria, Canada, Germany, Japan, Sweden, UK, and US. (See document numbers ... ) REPORT OF THE DEVELOPMENT BODY FOR FORTRAN 2000 was presented by Wagener on behalf of X3j3. LIAISON WITH OTHER SC22 WORKING GROUPS: Programming Language Standards: Effect of multiple character sets needs to be considered. DEFECT MANAGEMENT No further defect management for Fortran 90 is contemplated; further work in this area will be relative to Fortran 95. This assumes that TC3 and Fortran 95 will be approved. Procedures are being developed for accepting defect reports from individuals, from member bodies, and from WG5. Fortran 95 defect management is expected to follow substantially the same procedures as were used for Fortran 90. VARYING STRINGS We have a work item to modify the "informative" part to conform to Fortran 95. INTEROPERABILITY WITH C (technical report) Hennecke The principal technical issue remaining is "bind" vs. "map to". DATA TYPE ENHANCEMENTS (technical report) Cohen (known informally as PARAMETERIZED DERIVED TYPES) At issue is assignment of derived types that contain allocatable components. CONDITIONAL COMPILATION Epstein WG5 has previously suggested that a project in this area be directed toward preparation of an additional "Part 3" of the Fortran standard. The "Fortran-like" approach was presented at this point; presentation of the "fpp" approach was deferred until later in the meeting. Straw votes: 16 prefer the "Fortran-like" approach 2 prefer the "fpp" approach to COCO 4 undecided If and when COCO is standardized it should be ... 7 optional 9 mandatory 6 undecided PARAMETERIZED DERIVED TYPES (possible technical report) Morgan A technical issue involves the syntax for parameter value inquiries; two suggested approaches are "parameters as components" and "intrinsic types". Straw votes: 3 prefer the "components" model 6 prefer the "intrinsic types" model 13 undecided INTERVAL ARITHMETIC Kearfott The principal question is whether this item should be processed as a Fortran 2000 requirement or as an additional "Part" of the standard. FLOATING-POINT EXCEPTION HANDLING (technical report) Bierman Completion at this meeting is anticipated. CONDITIONAL COMPILATION Bierman The "fpp" approach was presented. ADDITONAL FORTRAN 2000 REQUIREMENTS Zuern presented N1187; Kuester presented N1186; Wright presented N1206. FLOATING-POINT EXCEPTION HANDLING Bierman, Weber Straw votes: Where IEEE leaves behavior processor-dependent, this TR should do the same. 2 yes 10 no 11 undecided For the following straw votes, delegates were asked not to vote "undecided" so that a final draft can be prepared at this meeting. (N1195r1 page 15): In case of integer overflow during conversion from decimal string, the processor must signal "IEEE INVALID" 12 yes 6 no (N1195r1 page 16): If no value of a variable depends upon the process, whether the exception is signalling is processor-dependent 14 yes 4 no Overflow and Divide-by-zero must be supported for non-IEEE processors. In the current TR draft, names for these exceptions always begin with "IEEE". Should this notation be retained? 10 yes 7 no The current TR draft must be changed to recognize the following rules: A called procedure must not change the caller's halting or rounding modes. A procedure must not reset a callers exception flag (but may set it). A procedure must not use exception flags as hidden input parameters. Straw votes: These rules are to be enforced ... 9 automatically by the processor 6 by the programmer These rules apply ... 13 to all procedures 5 only to PURE procedures DATA TYPE ENHANCEMENTS Cohen It is suggested that concerns recently expressed by X3J3 be addressed by means of additional informative text (see N1197). Straw Vote: Revise N1196 by adding proposed text at end of section 3.4 20 yes 0 no CONDITIONAL COMPILATION Epstein, Bierman Straw Votes: 9 The approach suggested in N1192 ("Fortran-like") 6 The approach suggested in N1208 ("fpp") Repeated with request that delegates not abstain: 13 The approach suggested in N1192 ("Fortran-like") 6 The approach suggested in N1208 ("fpp") Proceed with Conditional Compilation project? 15 yes 5 no Convenor states that this project will be pursued as an additional "Part" of the Standard, based on the "Fortran-like" approach. INTEROPERABILITY WITH C (Hennecke) Straw Votes: Should permit Fortran module variables to access C global variables 18 yes 1 no 1 undecided C "char *" should map to Fortran's ... 11 character array 4 derived type 7 undecided 17 retain BIND and add MAP_TO (at least for intrinsic types) 1 retain only BIND 2 provide only "MAP_TO" PARAMETERIZED DERIVED TYPES US position is that this project has not met the time-critical requiremwnt for conversion to a TR. Convenor reviewed the requirements for a TR. Straw Vote: 11 Convert this project to a TR 8 Process as a high-priority Fortran 2000 requirement 0 Do not include in Fortran 1 Undecided Convenor states that insufficient consensus exists for SC22 processing as a technical report. This project will be given further consideration as a Fortran 2000 requirement. INTERVAL ARITHMETIC A straw vote at the previous WG5 meeting favored processing as an optional "Part". Straw Votes: 11 Process within Part 1 of the Standard 8 Process as a separate "Part" of the Standard 0 Do not include in Fortran 2 Undecided 0 Process as a Technical Report 19 Process as a Fortran 2000 requirement 0 Undecided Convenor states that this project will be processed as a possible Fortran 2000 requirement. FORTRAN 2000 REQUIREMENTS A set of specific requirements should be adopted. Those to which WG5 is already rather firmly committed will consume most of the available time. These are: Floating-point exception handling (TR); Enhanced Derived Type Facilities (TR); Interoperability with C (TR) Derived-type I/O Asynchronous I/O Procedure Variables or Pointers to Procedures Interval Arithmetic Parameterized Derived Types Minor Technical Enhancements Additional requirements may be added later but only with full recognition of implications for the "Strategic Plan" timetable. Obect-oriented features do not appear to be well enough in hand for firm action at this meeting. As an experiment, WG5 will hold an extra meeting in February 1997, in conjunction with the X3J3 meeting. WG5 subgroups will be formed to report on additional requirements at that meeting. The existing "requirements data base" will be closed and archived; further proposals will be processed via different channels. New subgroups may wish to review the existing data base to ensure that important proposals do not become lost. Straw Votes: 4 Complete further refinement of F2000 requirements at this meeting 9 Form subgroups to complete further refinement before February 1997 10 Undecided Should Condition Handling (beyond the Floating Point Exception Handling TR) be a high-priority F2000 requirement? 3 yes 14 no 6 undecided WG5 POSITION ON ISO COPYRIGHT POLICY Straw Vote: Retain status quo? 20 yes 1 no 0 undecided Incorporate this issue in a Resolution? 14 yes 3 no 6 undecided POSIX BINDING TO FORTRAN (IEEE STANDARD) This IEEE standard will expire in 1997 unless it is extended. Straw Vote: Should WG5 request extension? 1 yes 4 no (many) undecided RESOLUTIONS See N1219