[cfp-interest 4020] New CFP potential items and my proposed responses - Part 2

RAJAN BHAKTA rbhakta at us.ibm.com
Fri Nov 13 14:25:26 PST 2026


Note that I did not put in the following issues from N3977 (C23 issues) since it seemed self-evident or something we already have on the queue and not something we may have changed (unlike issue 1046).

I also did not list issue 1086, since it is something I’ve never seen (unsigned infinity) in practice and was intending to leave it to others in case they had better views, but given Josha’s response to the first email I figured someone may have experience with it and I would appreciate any thoughts beyond my ignorant ones.

    1059: <fenv.h> types underspecified
      OK to leave it as such since no issues in practice.
      If it must be specified, suggestions:
      Question 1: Yes.
      Question 2: Yes.
      Question 3: Kept open. No need to specify one way or another.
      Question 4: Kept open. No need to specify one way or another.
      Question 5: Kept open. No need to specify one way or another. In practice, fenv_t is likely not modifiable, while there others are (defined as an integral type underneath).
      Question 6: Kept open. No need to specify one way or another.
      Question 7: Kept open. No need to specify one way or another. Only practical way is to have representation-usable for fexcept_t and femode_t (for assignment, copying, etc.).
      Question 8: N/A.  But kept open. No need to specify one way or another.
      Question 9: N/A.  But kept open. No need to specify one way or another.

    1085: Non-finite values in arithmetic constant expressions
      Seems to be the same question we have been discussing on what “Range of Representable Values” means.

    1086: Questions about infinity
      Disagree that negative and positive infinity can be independently. I read the “likewise” as a way of restating the previous part of the sentence so it applies to for positive infinity. But I can see the other reading that the issue reports. This could be fixed and makes questions 1 and 2 irrelevant. I also took you could have either unsigned or signed infinities, not both. Another thing that could be fixed making question 3 irrelevant.
      Question 4: Not specified in the range, so no.
      Question 5: No, since it is unsigned. Just like unsigned integers.
      Question 6: Text seems clear it is only for positive or negative infinities, so yes, it should return 0.
      Question 7: Text seems clear so yes, it is implementation-defined.

Regards,

Rajan Bhakta
​CFP

From: cfp-interest <cfp-interest-bounces at ucbtest.org> on behalf of RAJAN BHAKTA <rbhakta at us.ibm.com>
Date: Friday, November 13, 2026 at 3:52 PM
To: cfp-interest at oakapple.net <cfp-interest at oakapple.net>
Subject: [EXTERNAL] [cfp-interest 4018] WG14 meeting next week - New CFP potential items and my proposed responses

This Message Is From an External Sender
This message came from outside your organization.
Report Suspicious<https://us-phishalarm-ewt.proofpoint.com/EWT/v1/AdhS1Rd-!-_FVP3geXk9bWls5zoTAUdkpmsZvcB3tmrox2yI4098yJbJVxz7QH9nUPKAcXWHJDkDQC9W3CTXKrmz2_vpbqFBzc7SkCi9i6JuU9zkr2il5TVq426m6GwzF3OtQr4HCf8BYr_uExg$>

Hello,

In the WG14 meeting starting next week, we will be going over C2Y and C23 issues. Some of the ones that I noticed touch our domain (CFP) are listed here along with my proposed responses. I do intend to bring these up in the next CFP meeting as well, but if I can get early input, I can incorporate it into the meeting responses if we get to these issues.

C2Y issue log, r2 [https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3978.htm<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.open-2Dstd.org_jtc1_sc22_wg14_www_docs_n3978.htm&d=DwMFAg&c=BSDicqBQBDjDI9RkVyTcHQ&r=XTIN0AwPKAf2bIKNxlzdeqggAytE5h-JEusWpx9MXwc&m=xNaKMFZAYlsSt3ueqVQdIlrQVC3ofr7mLcZSH_kAdouhKbkdZExwVZJNN6_wuoV7&s=DIVlPOnRQtchxk6zz4dcPE4F_Ry18k_x71wMZ0l7iRU&e=>]
    1100: Should be "What constitutes range errors in non-default rounding modes are implementation-defined."

C23 issue log, r7 [https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3977.htm<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.open-2Dstd.org_jtc1_sc22_wg14_www_docs_n3977.htm&d=DwMFAg&c=BSDicqBQBDjDI9RkVyTcHQ&r=XTIN0AwPKAf2bIKNxlzdeqggAytE5h-JEusWpx9MXwc&m=xNaKMFZAYlsSt3ueqVQdIlrQVC3ofr7mLcZSH_kAdouhKbkdZExwVZJNN6_wuoV7&s=aalrk6foA2kRrzL_TXEigKMYL8XgBPbCIhYLEWmjFyQ&e=>]
    1038: Intent was behavior of use of those types. (float_t, double_t)
      Question 1: Intent is that it should be the behaviour in any use of the types, not any program at all.
      Question 2: Yes it can be DFP, and if float and double are base 10 types it could work.
      Question 3: Yes, but the same restrictions/conditions apply: Implementation defined behaviour since not real floating types.
      We can restrict what they types are to real floating types since that seems to be the only practical choice, but that does restrict freedom and is a normative change that does not match the intent.

    1046: Seems suggested correction is good. Not sure what changed, but we didn't notice. (powr(y,x) -> powr(x,y))

    1062: Intent is to allow it if the implementation wants (efficiency). (SNaNs cause exceptions on assignment?)
      Question 1: Implementation defined should still take precedence. This is the most likely case due to efficiency. Most implementations likely do byte-wise or chunk-wise copy on hardware so it would be a burden to require tracking contents and triggering the signal. Matches existing behaviour as well for implementations that I know of.
      Question 2: Like question 1. Most likely will be like a trap representation. Signals should only occur only on a read, not assignment.
      Question 3: No standard library function should produce a SNaN other than ones that directly modify it (like setting payloads). Uninitialized objects being a SNaN (Ex. malloc'd storage) and then assigned should signal if a read is done just like trap representations.

    1074: Interaction between freestanding and feature test macros with conflicting specifications
      Question 1: Yes, if __STDC_IEC_60559_COMPLEX__ is not defined, the header is not required. If __STDC_IEC_60559_COMPLEX__ is defined, then that should override the clause 4 since this is a conditionally normative Annex (G) which overrides the main body with its changes.
      Question 5: This is a broken implementation as the implementation said no complex types but includes a standard header that includes complex types.
      Question 6: Yes, and yes for both parts to the question. It is for the effective setting of access to the floating-point environment, not the textual one. The current wording seems to be clear in any setting of "on" (or "default" when default -> on) precludes it from the entire program, not just the translation unit or section.

Regards,

Rajan Bhakta
CFP

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20261113/f183f3e7/attachment-0001.htm>


More information about the cfp-interest mailing list