J3/06-275 Date: 10 August 2006 To: J3 From: Dick Hendrickson Subject: clause 2 problems References: J3/06-007 Questions and comments that are more than editorial. [2:4] bullet 6, doesn't IEEE and clause 9/10 specify rounding behavior and also SQRT accuracy? [2:9] Says a program is standard conforming if it uses the forms described herein. Does that need to be changed because of C interoperability? We hardly describe that here. [2:9] Related to the above, I think there is confusion about how a main program is described. [9:10] and [9:15] say a main program is a Fortran main program and don't say anything about a C main. [12:21] says must have exactly one main program, which I would take to be the English equivalent to the bnf main-program. [15:11] says "if a program contains a main program...". [2:14] Does VOLATILE affect the way things are interpreted? Basically, the standard provides no clue about volatile. [13:11] Don't we now allow internal procedures to be passed as actual arguments? No longer local. Can they be pointed to by a procedure pointer? or type bound? [298:4] related issue. [298:4] says an internal proc can't be passed, 12.5.1.6 says it can. [13:20] Do we need finalizers or generic type bound procedures here? [15:table 2.2] Add a column for derived type definitions, they're the only missing scoping unit. [17:3+] I agree with Malcolm's issue 1 [18:3] Add "This is the required behavior if there are no co-array things used." [19:35} This says that dummy arguments are not local variables of a subroutine. 16.3.1 [488:5], say they are. Also, [33:16] says a macro dummy is a local thing. [24:7] Doesn't intrinsic also apply to processor supplied types, like QUAD, that are not provided by a module or procedures?