[cfp-interest 3791] JTC1/SC22/WG14/CFP 2026/02/18 Meeting notes

Thu Feb 19 11:06:42 PST 2026

2026/02/18: 10:00 AM PST/6:00 PM UTC
[Please submit proposed changes to these minutes to Jerome or to the group.
Revision changes appear at the bottom.]

*Attendees*

    Jim Thomas, Rajan Bhakta, Jerome Coonen, Damian McGuckin, Joshua
Cranmer, Fred Tydeman, David Hough

*Updated agenda plus new items*

https://cfp-wiki.esi.com.au/pub/CFP/WebHome/CFP%20meeting%20agenda-20260218-update.pdf?_t=1771391603
–
These minutes should be read alongside the agenda, with its many reference
links.

*Previous meeting notes*

https://mailman.oakapple.net/pipermail/cfp-interest/2026-January/003775.html
2025/11/12 Meeting notes

*Study group logistics*

    Next meeting: 18 March 2026, 8:00 AM PDT/3:00 PM UTC -- note return to
the usual time and the change to Daylight Saving.

    ISO Zoom teleconference

    Please notify the group if this time slot does not work.

   Jim: We have a new wiki and password. Damian stepped in when EDG had to
drop our old wiki.

    Damian: The wiki runs on a VM in my company's world with a company
domain. Some wikis have gone to ISO.

    Rajan: Other groups have gone to non-ISO sites, which they find easier
to deal with.

    Jim: Several people did yeoman service to reconstitute the CFP wiki.

    Damian my colleague Peter Harding did most of the work. Peter and I
occasionally work on similar internal projects. The internal framework is
in Perl, part of the challenge. Wyatt Childers was helpful from EDG.

    Group: Started sweepstakes of who wrote Perl least recently. David won
with a 1990s showing. Joshua admitted to Perl work in 2026. (Is *that* the
secret sauce in C++?)

    Rajan: Asked about wikis at the WG14 meeting. There is a push to start
a C foundation to support hosting, email, etc. Looking for companies to
commit to the foundation. Currently, different study groups use different
hosts. The long term goal is to unify under one umbrella.

*C documents*

    The latest C2Y draft is N3783 Jan. 2026
https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3783.pdf

    C23 has been published ISO/IEC 9899, available for purchase.
https://www.iso.org/standard/82075.html

*IEEE 754 liaison*

    Damian: 80-bit floating point, a cornerstone of the Intel 8087 and
754-1985, is likely to disappear into an annex. Lots of work on language of
rounding, of interest to Damian for CFP angle. A new rounding mode,
round-to-odd, is still under discussion.

    David: Certainly a lot of work happening.

    Jerome: Lots of discussion of NaN propagation.

    Rajan: Does CFP have any members among the 754 renewal committee
hierarchy?

    Group: No, but several members keep tabs on the effort.

*C++ liaison*

    Joshua: Next meeting is the week of 23 March. No schedule of papers yet.

*WG14 update*

    Rajan: Sent email with update. All of the CFP papers were
enthusiastically accepted. Some committee members would prefer editorial
matters go straight to the editor. [CFP errs on the side of putting all
interesting changes out for discussion.] Issue 1014 with change to headers
and error handling in TS-4 are "accepted" but in limbo because there is no
plan to update TS-4.

    Jim: Is it possible to bring TS-4 into C2Y?

    Rajan:The issue is lack of implementation experience, as noted by
Joseph et al.

    Group: No knowledge of TS-4 implementation experience.

    Rajan: Robert says will bring up schedule for C2Y at "next meeting",
which is not fully defined. Relevant to CFP: there is no deadline for new
papers yet.

*TS-4 and TS-5 revisions*

    None.

*News*

    Jim: Paul Zimmermann sent an update about the correctly rounded
function effort.

*Carryover action items from last meeting *

   Jim & Jerome: Investigate Joseph Meyers's issue of the conflicting
requirements for double_t to be either double or long double.
        Done

    Damian & Jerome: Look at +/-0 in fromfp, and also the case of a value
rounding to zero.
        Done

    Group: Think about where CFP is headed.
        Carry over

*Action items from last meeting *
    Jim: In the proposal for Annex F floating-point environment updates (re
rounding precision, etc.), change all instances of “to nearest, ties …” to
“to nearest with ties …”. Send the doc to CFP with a timeline for comments
allowing submission for the March WG14 meeting.
        Done

    Rajan: Send email to Mike Cowlishaw asking whether anyone uses the ties
away from zero as the default in decimal. Share with group.
        Rajan: In reply to my question, Mike doesn't know why the option to
use the new rounding mode as the default was added. He doesn't know of any
systems that default to it, but would expect bankers to use it.
        Jim: CFP has them because 754 has them.
        Damian: There has been discussion in the committee about decimal
rounding. Action to research.

    Damian: Contact Fred about extending the list of exact special cases in
Annex F. Assemble a list for CFP to consider.
        Done

     Jim: Draft a suggestion for approaching the fixes to symbols.
        Done

*Discussion of issues*
    CFP IT
        David: Our email reflector is running on 25 year old hardware on 10
year old OS. I'm upgrading to 5 year old hardware able to run current OS.
Might be good to look at a new host for email.
        Jim: Some fragility there.
        David: Can host the archive of mail indefinitely. Current system is
Solaris and target is Ubuntu.
        Damian: We can host email alongside the wiki but i would prefer to
take a version migrated from the current system. Action to take the archive
now for safekeeping, alongside the new wiki.
        Rajan: Yes, the proposed C foundation is intended to host email,
too.
        Damian: Yes , we do back up the wiki. Action to document the backup.

    Conflicting requirements for double_t
         Jim: The email trail details the change from TS-3 to incorporation
into Annex H in C23. One solution is to return to the TS-3 spec. The other
two options work with Annex H as is, but with some tweaks that add
complications. All options imply possible changes for current
implementations. Option 1 has arguably largest impact, though also arguably
most sensible.
        Rajan: The change to Option 1 will bring up issue of why the change
was made in the first place, which should accompany any proposal. Also,
it's a good idea to write to Joseph about the intention here to avoid
surprises.
        Action

    Sign of zero from fromfp functions
        Jerome: Referring to the emails, it's best to look at the fromfp
family of functions from their appearance in Table F.2 , which gives
bindings for a multitude of operations in 60559. There are ten operations
converting floating values to integer types, with 5 modes of rounding and
the choice to signal Inexact or not. The better representation in the the
table would add the language "cast of" in order to get the integer value
into an integer format. The beauty of this C23 mechanism is that the fromfp
family handles the rounding while keeping inf or nan results in floating
form; then the cast delivers the value in integer type, with the
conventions for inf and nan. The original question was about the sign of a
zero result from fromfp. Given the clear connection between fromfp and
current 2's complement systems, it's natural to propose positive or
unsigned zero as the result. This may break some systems implemented since
C23.
        Rajan: The committee has more appetite for breaking things, if we
are moving in the right direction. Adding an example of the problem will
help. Some incompatibility is probably OK.

    More Annex F special cases
        Damian: Fred noticed compound(0, n) = 1.0 in 60559 but not in Annex
F. It's apparently the only such special case not in F.Also, pow(x, 1.0) is
not a special case in 60559, so it shouldn't be among the special cases in
Annex F. I've added the compound(0,n) change to F.
        Jim: This could mean a change to some implementations -- but it
does follow 754.
        Rajan: Should not be a problem.

    Annex F proposal update
         Jim & Damian: The language of rounding between C2Y and what is
evolving in 60559 is a challenge. Action to pursue.
        Damian: We are getting close to making a proposal.

*Other issues*
    None

*Adjournment*

    12:05 PM PST

-----------------------------------------------

*Action items to be carried over*
    Group: Think about where CFP is headed.

*New action items*

      Damian: Investigate the current state of default rounding(s) in 60559
and whatever is being discussed in the committee about changes, especially
in decimal.

     Damian: Arrange to take the email archive from David. Discuss the
status going forward.

    Damian: Document the wiki backup protocol.

    Jim & Jerome: Draft proposal re. double_t semantics and send a note to
Joseph about the proposed solution.

    Jerome & Damian: Write a proposal to repair fromfp and enhance Table
F.2.

    Jim & Damian: Look at the language around rounding direction in
Damian's Annex F paper and make a proposal to CFP about how to go forward.

*Carryover discussion items*
    Minus signs and hyphens, etc.

    Editorial issues F.2.2

    Editorial issues F.10.1

    Editorial issues F.3, etc

    Rounding direction language

    CFP future

    Complex and signaling NaNs

    INFINITY and _Float16

*Signoff*
    Respectfully submitted.

-Jerome Coonen
 650.996.4738
 jcoonen at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20260219/2b76271b/attachment-0001.htm>