[Cfp-interest 1652] Re: Meeting minutes

[Fred J. Tydeman]

On Wed, 17 Jun 2020 13:09:30 -0700 (PDT) Fred J. Tydeman wrote:
>
>Corrections welcome
There were some corrections sent to me.  Here are the updated minutes.

C Floating-Point teleconference of 2020/06/17

Meeting logistics
 Note taker:  Fred Tydeman

Attendees:
 Jim Thomas (host)
 Fred Tydeman (note taker)
 David Olsen
 Ian McIntosh
 Mike Cowlishaw
 Damian McGuckin
 Rajan Bhakta
 David Hough

Carry over action items
 David: duplicated CFP messages - mailman was updated - done

Action Items
 Jim: Let WG14 know that 60559:2020 was published - done
 Jim: Get new paper numbers about 60559:2020 - done
 All: review Part 3 as annex - done
 Jim: Check WANT macro issue in CFP 1581 - open

Next meeting - July 22

C++ liaison
 C++ [David Olsen]
 Paper out for review [CFP 1638].  Would like feedback from us.
 Jim:  How to constants with suffixes?
 David: Can do some.

C2X integration
 New editors working on rebuilding WG14 N2478.

AI details

 David: Mailman has been updated.  Things appears to be fixed.

 Jim: WG14 has been told about ISO/IEC 60559:2020 publication.

 Jim: New paper numbers for IEEE 2019 has been done

 All: Review Part 3 as Annex
  Links in minutes do not work.  Depends upon how create PDF from Word.
    new AI: Need ideas on how to convert Word to good PDF with links.
    cutepdf -- might work
    google docs -- has some problems
  Missing some conversions = CFP email 1640. Problem is binary-decimal
    conversions; need support for hex in all cases.
    %a: binary to hex can be exact.  Then hex to decimal can be done
    with one rounding error (with enough hex digits).

    Jim: email with Micheal Hack: common struct with both binary and
    decimal of same value with enough precision.  Jim should send to Hack
    about use of hex.

    %a is already part of C and can be big enough for exact representation.

    New AI Jim:  add hex to annex draft.

    This annex part 3 depends upon Annex F.  Should that be optional?
    For example allow, IBM S/360 hex can be float and double, and also
    IEEE-754 FP types via extended types.  No need in market place.

 Jim: Check WANT macro issue of CFP 1581. Still Open.

Other Issues

 Missing 'const' in *payload.  Editorial or new paper?  
 AI: Fred - write paper saying missed.

 Expression eval footnote.
 AI: Fred - write paper on DEC_EVAL_METHOD.  Nothing about static vs auto.

 Email: CFP 1643: Rnd ties away issue.
 AI: Jim: rewrite.

 David Olsen: Promotions for small binary interchange types to double.
 Jim: email 1634 has proposed words.
 Fred: Might be issue with float imaginary vs _Float32 imaginary.  
 AI: Jim+Fred: no prototype vs varargs prototype.  Find where in C where
 float is promoted to double.

 intmax_t removal: *from* return FP type same as argument.
 Fred: compoundn, rootn: use large long long int implies result is 0,
 1, or inf.  [Corrected after meeting: only true for rootn]
 Jim: when longer int is converted to long long int, result is
 unspecified when out of range.
 Fred: Use FP type instead of intmax_t.  
 David: Needs to be an integer.  
 NOTE: Needs to be on agenda for next meeting.

 WANT macros annex F only functions.  
 Mixture to impl defines macros and user WANT macros.  Need synopsis
 for both binary and decimal.

 Fred: Maybe also for part4a decimal math functions: asinpid32 needs
 both decimal and extra functions.
 NOTE: Needs to be on agenda for next meeting.

Others
 None.


---
Fred J. Tydeman        Tydeman Consulting
tydeman at tybor.com      Testing, numerics, programming
+1 (702) 608-6093      Vice-chair of PL22.11 (ANSI "C")
Sample C99+FPCE tests: http://www.tybor.com
Savers sleep well, investors eat well, spenders work forever.