dgh%dgh@Sun.COM (David Hough) (12/08/87)
I have finished the first draft of a document on this subject and shipped it off to the ANSI C committee, meeting this week at the Driskill Hotel in Austin. I will continue to sharpen and expand my commentary based on feedback from colleagues until the next formal public review period. So far the following have indicated agreement with most of what I've proposed: Larry Breed ibmpa!lmb W. J. Cody cody@anl-mcs.arpa Zhishun Alex Liu zliu%hobbes@berkeley.edu K-C Ng kcng@sun.com Gene Spafford spaf@purdue.edu Jim Valerio omepd!radix!jimv If you are interested in troff source, please send me mail - don't post to news! You won't need to post if you remember to include a bullet-proof usenet or arpa return address. Please specify if you are interested in the comments submitted to X3J11 on 7 December 1987 or the ever-evolving most current version. If you're not sure, here's the executive summary. troff source is almost 60K bytes and thus unsuitable for posting: The proposed C standard suffers numerical shortcomings - many inherited from its precursors - in the areas of expression evaluation and floating-point exception handling, particularly in the library of elementary transcendental func- tions. The following comments are included: Comment #1, Section 3.2.1.4: round conversions between floating types Comment #2, Section 3.3.3.3: don't define unary + and - with formulas Comment #3, Section 4.7: SIGFPE means floating point Comment #4, Section 4.9.6.1: printf %+ displays signed zero Comment #5, Section 4.9.6.1: printf %f distinguishes exact zero Comment #6, Section 4.9.6.2: scanf requires pushback > 1 Comment #7, Section 4.10.1.4: strtod/atof are mathematical functions Comment #8, Section 4.10.2: require two random number generators Comment #9, Section A.5: forbid implicit narrowing conversions Comment #10, Section 3.5.2.3: facilitate forcing rounding Comment #11, Section 3.3: opaque parentheses by default Comment #12, Section 3.4: opaque parentheses defer constant expression evaluation Comment #13, Section 3.3.3.3: respect opaque parentheses in expression evaluation Comment #14, Section 2.2.4.2: <float.h> has too many names, not enough information Comment #15, Section 2.2.4.2: long double > minimal float Comment #16, Section 4.9.6: printf/scanf duality for non-model numbers Comment #17, Section 4.5: specific mathematical library functions Comment #18, Section 4.13.2: standard functions predefine generic operators Comment #19, Section 4.5: make numerical exception handling uniform David Hough ARPA: dhough@sun.com UUCP: {ucbvax,decvax,allegra,decwrl,cbosgd,ihnp4,seismo}!sun!dhough