[Cfp-interest] draft of meeting notes for 19 April 2012
David Hough CFP
pcfp at oakapple.net
Thu Apr 19 14:36:25 PDT 2012
19 Apr 2012 CFP Meeting Notes
Attendees: David, Fred, Jim, Rajan, Joel
Action items from 15 March -
Jim revised part I draft and submitted to WG, based on March meeting discussion.
Next meeting Thursday 10 May.
Issues in draft:
Issue 1: Is requiring freestanding support for part of a header OK?
New idea for C standard. No adverse comment yet from WG14
In document now.
Consensus no change.
Issue 2: Break F.1 into three paragraphs for clarity.
Consensus: OK. No substantive changes.
Issue 3: Better to restrict the format specifiers that strfrom must support?
Don't need other stuff for simple data transfer.
Omits many formatting options and "*" precision at run time and ellipsis -
allowed but not required. Means specifying subset that programmer can't
extend to get full function.
Consensus: add ellipsis to minimum subset, allow implementation by snprintf.
Not require * precision.
AI: Jim to write up.
Issue 4: Do we need wide character strfrom? (hosted only, freestanding
doesn't have wide char)
Rajan: many customers using wide characters.
First consensus: add wide char support.
But: superfluous, other ways of doing this on hosted implementations.
AI: Jim to remove issue.
Issue 5: Appropriate for FP spec to specify integer strfrom functions?
Rajan: no, stick to FP.
Jim: no code portability issue, only a vague sense of incompleteness,
not sufficient reason.
Consensus: remove integer strfrom from draft, superfluous on hosted imps.
AI: Jim to remove issue.
ISSUE 6: The macro names FP_CEIL, etc. might suggest results will be in floating types.
Jim: change to different macros to avoid confusion.
AI: Jim to document a different set of macros.
ISSUE 7: Should the 7.12 specification of llogb require a domain error for finite out-of-range cases?
C comm will resolve DR that will decide.
AI: Jim to remove issue.
ISSUE 8: Forward references need to be added throughout.
Handle when finally finished, but move the issue to beginning of document.
AI: Jim to move issue to front.
Fred's issues:
1) fsqrt domain error occurs if the argument is less than zero
agreed - wording from C11
AI: Jim to change.
2) snan initializer for thread local storage - same as for static initializer
7.12#5
Jim: is this already implicit, so we don't need to state? C11 uses
"initializer of static duration" for static var and thread local.
Also uses "static or thread storage duration".
AI: Fred to find right C11 wording for snan initialization for thread local
storage.
3) canonicalize function: needs footnote that two arguments can point to same
object, in which case canonicalize in place.
do simple objects ever overlap partially? can happen in unions.
AI: Jim to add footnote addressing issue.
What to do next?
Part 2 of spec is to be Decimal. Different kind of job:
update of existing spec. Lots of formatting required.
Fred has a proposed document of changes.
Part 3 is interchange and extended. Should be mostly done already.
Don't want to inundate WG14 with lots of little updates.
AI: Jim to talk to John Benito to determine best approach to developing
other parts,
interaction with WG14, and posting document for public review.
AI: Jim to investigate moving decimal document into approved template.
All done in Word. Nobody else volunteered to help. Jim hopes to hear
from such a volunteer.
Leftover issue raised by Joel:
Jim's note:
http://mailman.oakapple.net/pipermail/cfp-interest/2012-April/000334.html
tgmath and constant rounding mode spec:
How does specification of constant rounding mode affect tgmath functions?
We stated that they are affected.
We've also stated that the functions so affected are those listed in Table 2
whose macro expansion has not been suppressed. So what happens when macros
are suppressed?
We are pinning down C11 behavior that wasn't fully defined.
Joel: tgmath might have to reimplement macros instead of rewrapping.
We need to work on this issue next time.
AI: Joel to write up a note.
More information about the Cfp-interest
mailing list