J3/01-207r1 Date: 14-Jun-2001 To: J3 Members From: interp/Stan Whitlock Subj: F95 interp letter ballot #4 - interps ready for WG5 Letter ballot #4 for F95 interps was in 01-207.txt. The results of the ballot after J3 meeting #157 are in 01-224r1.txt. Below are the text of the interps that "passed by J3 letter baloot" and are ready for WG5. These are 2, 10, 11, 12, 18, 19, 20, 21, 22, 24, 25, 28, 29, 81, 85, 87, 88, 89, 90, 91, 92, 93, and JP-06. At the end of this paper is the new text for interp #23 that failed the letter ballot, has been amended, and will be submitted for J3 consideration at meeting #158. /Stan --------------------------------------------------------------------- NUMBER: 000002 TITLE: Free source form requirement for blank in PRINT statement KEYWORDS: Free source form, PRINT, blank DEFECT TYPE: Interpretation STATUS: Passed by J3 letter ballot QUESTION: Consider the following PRINT statement form Section 3.3.1 of the Fortran 95 standard states: A blank shall be used to separate names, constants, or labels from adjacent keywords, names, constants, or labels. NOTE 3.3 For example, the blank[s] after ... READ ... [is] required in the following: ... READ 10 ... Although the sentence preceding R420 is somewhat confusing in that it uses the phrase "a sequence of characters, DELIMITED by either apostrophes or quotation marks", the syntax rule itself is clear that the " or ' delimiting characters are part of the syntax. The first sentence on the top of page 36 then clarifies that although the delimiter characters are a part of the syntax, they are not a part of the value. Section 3.2 makes it clear that a character literal constant is a token. Section 3.2.5 describes delimiters but does not include either " or ' because it is describing delimiters for lists. Thus, it seems clear that in free source form a blank is required between the keyword PRINT and the character literal constant containing the format list but there is also some disagreement on this point among existing Fortran processors. (1) In free source form, is a blank required between the keyword PRINT and the character literal constant containing the format specification? (2) Also, for clarification, in free source form, is a blank required between the keyword PRINT and the asterisk that represents the list-directed output specifier? ANSWER: (1) Yes. The analysis in the QUESTION is correct. Since PRINT is a keyword, when the format specification is a character literal constant, the blank is required between the keyword PRINT and the format specifier. (2) No. Since PRINT is a keyword, according to the cited rule in 3.3.1, the asterisk would have to be a name, a constant, or a label in order for a blank to be required to separate PRINT from the asterisk. By 3.2.1, an asterisk is not a name. By 3.2.2, an asterisk is not a constant. And by 3.2.4, an asterisk is not a statement label. Yes, an asterisk may be used as a dummy argument alternate return specifier representing a statement label but that usage is irrelevant to the PRINT statement. Therefore, since an asterisk is none of the three items that must be separated from PRINT by a blank, the blank is not required. EDITS: None. SUBMITTED BY: Larry Rolison / Dick Hendrickson HISTORY: 97-238 m143 Submitted 01-151r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000010 TITLE: Meaning of embedded blanks in namelist input name KEYWORDS: NAMELIST, optional qualification, name, input DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot BACKGROUND: In running some older NAMELIST tests, a test case was encountered that has embedded blanks within the "optional qualification" following a namelist group object name: CHARACTER*10 string NAMELIST /junk/ string WRITE(9,*) ' $junk string( : ) = ''1234567890'' /' REWIND(9) READ(9,junk) ... END While studying the standard to determine if the above program was standard- conforming or not, some questions arose about the use of "name" and embedded blanks in the namelist section 10.9.1.1. Among them were: Since a blank is a delimiter, should it be allowed to be embedded in a variety of places in a namelist input record that are not currently described by the standard? Do we want to strain namelist input scanning more than a reasonable amount? In normal Fortran terminology, the term "name" does not include subobject qualification. In section 2.5.1 Name and designator, lines 2 through 5, the standard states: A name is used to identify a program constituent, such as a program unit, named variable, named constant, dummy argument, or derived type. The rules governing the construction of names are given in 3.2.1. A subobject designator is a name followed by one or more of the following: component selectors, array section, array element selectors, and substring selectors. In section 10.9.1 Namelist Input, lines 13 to 16, the standard states: In each name-value subsequence, the name shall be the name of a namelist group list item with an optional qualification and the name with the optional qualification shall not be a zero-sized array, a zero-sized array section, or a zero-length character string. The optional qualification, if any, shall not contain a vector subscript. In section 10.9.1.1 Namelist group object names, lines 32 and 33, the standard states: The name in the input record may be preceded and followed by one or more optional blanks but shall not contain embedded blanks. QUESTION: (1) Does Section 2.5.1 mean that the optional qualification described in the 10.9.1 is not part of the name for namelist input? If the optional qualification is not part of the name, there seem to be no rules for where blanks and end-of-record may occur within the optional qualification. Also note that size of the name and the optional qualification together may be longer than the size of an input record if there are multiple levels of structure component qualification in a structure component reference. (2) Are embedded blanks allowed within substring and subscript qualifications in the namelist input? May an END-OF-RECORD occur in these same places? (3) More specifically, are embedded blanks/END-OF-RECORD allowed in the following places in namelist input? a. Between the name and the left parenthesis of either a substring or subscript qualification? b. Between tokens within the parenthesized qualifier? c. Between the two sets of parentheses for array and substring notation? d. Within the two subscripts or start and end values themselves? (4) Are embedded blanks allowed within a reference to a structure? May an END-OF-RECORD occur within this reference? ANSWER: The answer to all of the questions above is no. In section 10.9.1, the standard says "In each name-value subsequence, the name shall be ...". This usage of "name" refers to the optionally qualified namelist group object list item; therefore, in this context, "name" refers to both the namelist group object name list item, and any following qualification. In section 10.9.1.1, the standard says "The name in the input record may be preceded and followed by one or more optional blanks but shall not contain embedded blanks." This usage of "name" also refers to optionally qualified namelist group object list item; therefore, no blanks are permitted anywhere within the "name" of a name-value subsequence. An edit clarifies this. Since the end of a record has the same effect as a blank in namelist input (10.9), an end of record may not appear anywhere within the "name" of a name-value subsequence. EDITS: Replace the last sentence of subclause 10.9.1.1 [179:32-33] by "In the input record, each object name or subobject designator may be preceded and followed by one or more optional blanks but shall not contain embedded blanks." SUBMITTED BY: Joanne Brixius / Larry Rolison HISTORY: 97-253 m143 submitted 01-152r1 m156 Passed unanimously at J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000011 TITLE: G editing typo KEYWORDS: G edit descriptor DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: In F95 section 10.5.4.1.2, the last line of the table shows an equivalent conversion of F(w-n).1,n('b'). This is different from the corresponding table in F90, which shows this conversion as F(w-n).0,n('b'). There is no obvious reason for this change; it appears to be a typo in f95. Should this conversion be F(w-n).0,n('b') as in Fortran 90? ANSWER: Yes, this was a typographical error in f95. EDITS: [171:23] Last line of the table in 10.5.4.1.2. Change ".1" to ".0" SUBMITTED BY: Richard Maine HISTORY: 97-225 m143 Submitted Aug. 12, 1997 01-151r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000012 TITLE: Evaluation of Intrinsic Procedures KEYWORDS: Algorithm, Mathematical, Computational DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: 1. When the standard specifies an algorithm for computing a mathematical procedure, must a processor use the specified algorithm? Example: ANINT is defined as INT (A + 0.5). On some processors ANINT (16 000 001.0) evaluates to 16 000 002.0 using this algorithm. 2. May a processor return a mathematically equivalent result in the evaluation of a numeric intrinsic procedure? Example: If ANINT (16 000 001.0) can be computed as the more expected 16 000 001.0, may a processor do so? ANSWER: 1. No, a processor is not required to use the algorithm from the standard. 2. Yes, a processor may return the mathematically equivalent result in the evaluation of any numeric intrinsic procedure. DISCUSSION: The use of an algorithm in the definitions for intrinsic procedures is intended to describe the end result. It is not intended to require that a processor use a particular method of achieving that result. Thus, an algorithm specifies the mathematical results, not the computational results, of the evaluation of intrinsic procedures. This standard does not require a processor to use a particular accuracy or precision. Section 1.4, "Exclusions", states in part: "This standard does not specify: . . . (6) The physical properties of the representation of quantities and the method of rounding, approximating, or computing numeric values on a particular processor." [1:26, 1:37-38] The edits correct the definitions of the results for the intrinsic functions ANINT and NINT to be expressed in mathematical terms. EDITS: On page 231, subclause 13.14.10, Result Value paragraph [231:16-17]. Replace by The result is the integer nearest A, or if there are two integers equally near A, the result is whichever such integer has the greater magnitude. On page 259, subclause 13.14.77, Result Value paragraph [259:9-10]. Replace by The result is the integer nearest A, or if there are two integers equally near A, the result is whichever such integer has the greater magnitude. SUBMITTED BY: Keith H. Bierman HISTORY: 95-247r1 m134 Submitted as F90 interp 000202 with proposed response, approved 9-4, subsumes F90 interp 000200. 95-256 m134 Ballot failed, 10-6 95-260 m135 Proposed response, no action 96-023 m136 Proposed response, no action 96-054 m137 Proposed response, no action 96-117 m138 Proposed response, no action 97-259r1 m143 Re-submitted as F95 interp 000012 with proposed response 00-214 m153 Passed unanimously 00-254 m154 Passed J3 letter ballot 00-268 m154 Failed WG5 ballot (N1403) 01-153 m155 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000018 TITLE: ELEMENTAL procedures with no arguments KEYWORDS: Elemental, procedure DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot Section 12.7.2 of ISO 1539-1:1997 seems to assume there is at least one argument to an elemental procedure. The semantics of the following examples do not seem to be clearly specified. Example 1: ELEMENTAL INTEGER FUNCTION f() f = 0 END Example 2: ELEMENTAL INTEGER FUNCTION g(i,j) INTEGER,INTENT(IN) :: i,j g = i + j END Example 3: ELEMENTAL INTEGER FUNCTION h(i) INTEGER,INTENT(IN),OPTIONAL :: i h = i END Example 4: ELEMENTAL SUBROUTINE s() REAL x,y,z x = 0 y = 0 z = x/y END QUESTION 1: What is the shape of a reference to function F in example 1? Is it scalar? QUESTION 2: If G() is referenced as G(iscalar,jarray(1:0)), the result is a zero-sized array. Is "iscalar" referenced (need it be defined)? Is the body of G() executed? QUESTION 3: What is the shape of a reference to H() if the argument is textually not present? QUESTION 4: When subroutine S is invoked, how many times is it executed? Is this example standard-conforming, or may a processor raise a divide-by-zero exception? ANSWER: (1) Yes, A reference to F() is intended to be scalar. A clarifying edit is supplied. (2) It is not possible to tell, in Fortran 95, whether the body of G() is executed, so that part of the question has no answer. [no edit required] Yes, ISCALAR must be defined; this is the same situation as writing ISCALAR + JARRAY(1:0) directly, with no user-defined ELEMENTAL procedure being used. [no edit required] (3) If the I argument to H is textually absent, a reference to H returns a scalar result (cf. answer to question 1). Note that execution of this procedure is not standard-conforming if the optional argument is not present. The "H = I" line should be replaced by IF (PRESENT(I)) THEN H = I ELSE H = 0 ! or some other desired scalar value END IF [no edit required] (4) The intention was that the subroutine is executed once per invocation. This is not clear from the standard; an edit is supplied to correct this oversight. However, the example is not standard-conforming on a processor whose arithmetic does not allow division by zero. [no edit required] EDITS: Replace the final sentence of subclause 12.4.3 [206:7-10] by: "A reference to an elemental subroutine (12.7) is an elemental reference if there is at least one actual argument corresponding to an INTENT(OUT) or INTENT(INOUT) dummy argument, all such actual arguments are arrays, and all actual arguments are conformable." [214:28] After "If" insert "there are no actual arguments or". SUBMITTED BY: Malcolm Cohen HISTORY: 98-117 m144 Submitted 00-123 m152 Revised by submitter (version above) but withdrawn for further study without vote 01-148r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000019 TITLE: Result of NULL intrinsic associated with CHARACTER(*) dummy argument KEYWORDS: NULL intrinsic, assumed character length parameter, argument association, pointer association DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Consider the following program. PROGRAM P CALL SUB(NULL()) CONTAINS SUBROUTINE SUB(C) CHARACTER(*), POINTER :: C PRINT *, LEN(C) END SUBROUTINE SUB END PROGRAM P According to 7.1.4.1 [91:27-30], "The data type, type parameters and rank of the result of the intrinsic function NULL when it appears without an argument are determined by the pointer that becomes associated with the result. See Table 7.2." In this case, the pointer that becomes associated with the result is the corresponding dummy argument when the result appears as an actual argument. However, according to 5.1.1.5 [51:14-18], a dummy argument of type character, with a character length parameter of *, "assumes the length of the associated actual argument when the procedure is invoked." Is this a standard conforming program? If so, how is the length of C determined, and what should be the output of this program? ANSWER: No, this program does not conform to the standard since the actual argument does not have a character length for the dummy argument to assume. The edit makes this clear. EDITS: Page 91, subclause 7.1.4.1, last line [91:41]. Add The optional argument shall also be present if the reference appears as an actual argument corresponding to a dummy argument with assumed character length. SUBMITTED BY: Henry Zongaro HISTORY: 98-148 m145 Submitted 01-154r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000020 TITLE: Execution of a WHERE statement within a WHERE construct KEYWORDS: WHERE statement, WHERE construct DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: 7.5.3.2 [113:17-19] indicates that: Upon execution of a WHERE statement or a WHERE construct statement that is part of a , the pending control mask is established to have the value m .AND.(.NOT. mask-expr). c [112:41-45] indicates that: Upon execution of an ENDWHERE statement, the control mask and pending control mask are established to have the values they had prior to the execution of the corresponding WHERE construct statement. Following the execution of a WHERE statement that appears as a , the control mask is established to have the value it had prior to the execution of the WHERE statement. Notice that after execution of an ENDWHERE statement, the pending control mask is established to have the value it had prior to the execution of the corresponding WHERE construct statement, but the same is not the case for the pending control mask established by the execution of a WHERE statement that appears as a . According to this, the output of the following program PROGRAM P INTEGER :: A(2) = 0, B(2) = 0 WHERE((/.TRUE.,.TRUE./)) WHERE((/.TRUE.,.FALSE./)) A = 1 ELSEWHERE A = 2 END WHERE PRINT *, A WHERE((/.TRUE.,.TRUE./)) WHERE((/.TRUE.,.FALSE./)) B = 1 END WHERE ELSEWHERE B = 2 END WHERE PRINT *, B END PROGRAM P will be 1 2 1 0 Was this difference in behaviour intended by the committee? ANSWER: No. It was intended that WHERE() should have the same interpretation as WHERE() END WHERE The output from the above program should be 1 0 1 0 An edit is supplied to correct the text. EDITS: Page 113, subclause 7.5.3.2. In the first line of the paragraph following NOTE 7.48 [113:17] delete 'a WHERE statement or'; after the paragraph [113:20+] add the new paragraph: Upon execution of a WHERE statement that is part of a where-body-construct, the control mask is established to have the value .AND.. The pending mask is not altered. SUBMITTED BY: Henry Zongaro HISTORY: 98-149 m145 Submitted 01-155 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000021 TITLE: Restrictions on on END INTERFACE KEYWORDS: INTERFACE block, END INTERFACE, generic procedures DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: According to section 12.3.2.1 [194:18-20] The may be included in the only if it was provided in the and, if included, shall be identical to the in the . It is not clear what the words "shall be identical" were intended to mean. In particular, were they intended to merely convey the idea that the two s shall specify the same entity, even though such an entity might be represented in more than one way, or were they intended to convey something more restrictive? Are either of the following examples standard conforming programs? Example 1: PROGRAM P INTERFACE OPERATOR(.NE.) LOGICAL FUNCTION F(I, C) INTEGER, INTENT(IN) :: I CHARACTER(*), INTENT(IN) :: C END FUNCTION F END INTERFACE OPERATOR(/=) END PROGRAM P Example 2: MODULE MOD INTERFACE GEN SUBROUTINE SUB1(I) INTEGER :: I END SUBROUTINE SUB1 END INTERFACE END MODULE MOD PROGRAM P USE MOD, G1=>GEN, G2=>GEN INTERFACE G1 SUBROUTINE SUB2(L) LOGICAL :: L END SUBROUTINE SUB2 END INTERFACE G2 END PROGRAM P ANSWER: The intention is that the in the END INTERFACE statement should be obviously the same as that in the INTERFACE statement, both to the human reader and to the compiler. The words "shall be identical" signify that they must be lexically identical, that is, identical apart from layout changes permitted by the source form and the interpretation of the operators .LT., .LE.,.GT., .GE., .EQ., and .NE. as always being identical to <, <=, >, >=, ==, and /=. Example 1 conforms to the standard, but example 2 does not. Edits are provided to make this clear. EDITS: On page 194, subclause 12.3.2.1, in the fourth constraint following R1207 [194:19-20] delete 'and, if included, ... ' and add: If the includes , the shall specify the same . If the includes ASSIGNMENT(=), the shall specify ASSIGNMENT(=). If the includes OPERATOR(), the shall specify the same . If one is .LT., .LE., .GT., .GE., .EQ., or .NE., the other is permitted to be the corresponding operator <, <=, >, >=, ==, or /=. SUBMITTED BY: Henry Zongaro HISTORY: 98-150 m145 Submitted 01-156 m156 1st answer passed unanimously at J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000022 TITLE: Use of NULL() as initialization KEYWORDS: NULL intrinsic, initialization, default initialization, DATA statement DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Rules R429, R505 and R540 respectively describe the syntax for default initialization of pointer components, pointer objects in type declaration statements and pointer objects in DATA statements. These rules seem to indicate that NULL is to be considered a keyword in those contexts. This has the surprising effect that the first example below is a standard conforming program, while the second is not. That is, in some contexts, something that looks very much like a reference to an intrinsic function, is in fact a keyword followed by parentheses. Example 1: MODULE MOD INTRINSIC NULL END MODULE MOD PROGRAM PROG USE MOD, DISASSOCIATE=>NULL INTEGER, PARAMETER :: NULL = 17 INTEGER :: J = NULL INTEGER :: I INTEGER, POINTER :: P => NULL(), P2 DATA P2, I/NULL(), NULL/ TYPE DT INTEGER, POINTER :: P => NULL() END TYPE DT TYPE DT2 INTEGER, POINTER :: I INTEGER :: J END TYPE DT2 TYPE(DT2) :: S = DT2(DISASSOCIATE(), NULL), S2 DATA S2/DT2(DISASSOCIATE(), NULL)/ END PROGRAM PROG Example 2: MODULE MOD INTRINSIC NULL END MODULE MOD PROGRAM PROG USE MOD, DISASSOCIATE=>NULL INTEGER, POINTER :: P => DISASSOCIATE(), P2 DATA P2/DISASSOCIATE()/ TYPE DT INTEGER, POINTER :: P => DISASSOCIATE() END TYPE DT END PROGRAM PROG In some places in the scoping unit of PROG example 1, NULL() is used to specify initialization or default initialization, while in others NULL is a named constant, and DISASSOCIATE() must be used to specify initialization. 1) Is example 1 a standard conforming program? 2) Is example 2 a standard conforming program? 3) Was it the intent of the committee that, within the contexts of R429, R505 and R540, NULL should be treated strictly as a keyword rather than as a reference to an intrinsic function? DISCUSSION: The intention is that the intrinsic function NULL be referenced with no arguments in initialization contexts. This is made clear in the middle of page 49 in the paragraph that begins 'If ' [49:20-21] and also in 7.1.4.1. If the intrinsic function has been renamed, the new name should be used. ANSWER: (1) No, the example is not conforming because it references two different entities with the same name (NULL the local variable, and NULL the intrinsic function). See 14.1.2. (2) No, the example does not conform to the standard as published. This is an error, corrected by the supplied edits which make this example conforming. (3) No, the intention was that "NULL()" should be a reference to the intrinsic function. Edits are supplied to remove the keyword status of this syntax. EDITS: Page 39, subclause 4.4.1, replace the second line of R429 [39:20] by <> => R429a <> Constraint: shall be a reference to the intrinsic function NULL with no arguments. Page 47, subclause 5.1, replace the second line of R505 [47:40] by <> => Page 49, subclause 5.1, in the paragraph that begins 'If ' [49:20-21], replace 'NULL()' by '' twice. Page 62, subclause 5.2.10, replace the fifth line of R540 [62:19] by <> Page 62, subclause 5.2.10, in the penultimate line of the page [62:48], replace 'NULL()' by 'pointer association status'. Page 63, subclause 5.2.10, in lines 1, 7 and 10 of the page [63:1,7,10], replace 'NULL()' by '' thrice. SUBMITTED BY: Henry Zongaro HISTORY: 98-151 m145 Submitted 01-157 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000024 TITLE: Termination of a partial record by a CLOSE, BACKSPACE, ENDFILE, or REWIND statement KEYWORDS: Record termination, CLOSE, BACKSPACE, ENDFILE, REWIND DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Suppose a partial record has been written as a result of a nonadvancing WRITE. If the file is closed, or a BACKSPACE, ENDFILE, or REWIND statement is performed on the file, must the partial record be terminated? ANSWER: After a nonadvancing WRITE there is always a current record in a file. Existing descriptions of file closure and of the BACKSPACE, ENDFILE and REWIND statements define subsequent operations on the file. Further clarification in the BACKSPACE, ENDFILE and REWIND cases is provided in C.6.1.5 [321:7-11]. An edit is provided to make the intention clear. EDITS: Page 136. At the end of the last paragraph of subclause 9.2.1.3.1 [136:33] add If a nonadvancing output statement leaves a file positioned within the current record and no further output statement is executed for the file before it is closed or a BACKSPACE, ENDFILE, or REWIND statement is executed for it, the file is positioned after the current record before the specified action is performed. SUBMITTED BY: Robert Corbett HISTORY: 98-155 m145 Submitted (part 2) WG5/N1410 Draft answer 00-260 m154 Passed by J3 meeting 00-329 m155 Failed J3 letter ballot 01-158r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000025 TITLE: List-directed input: types of variables corresponding to repeated values KEYWORDS: List-directed input, repeated values DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: When a repeat count is used in list-directed input, must all the variables into which the repeated value is read have the same type? I see no reason in the standard to assume that restriction, but at least one Fortran 90 implementation assumes it. ANSWER: No, the variables are not required to be of the same type, but the type of the repeated constant is either a literal constant, or a nondelimited character constant, but not both (10.8, "where is a literal constant or a nondelimited character constant"). Whether certain forms of a repeated constant are treated as a literal constant or a nondelimited character constant is determined by the type of the first variable which consumes this value. Indeed, how many characters from the input record are part of is dependent on the type of the first variable which consumes a value from that . For example, the input 2*(1.0, 0.0) conforms to the standard when consumed by three list items of type character or by two list items of type complex. If there are three character variables, the first two get the value "(1.0", while the third gets the value "0.0)". If there are two complex variables, they both are assigned the value (1.0,0.0). Similarly, the input 2*.TRUE. can be consumed by two logical variables, or two character variables, but not by one character and one logical variable. An edit is provided to clarify the situation. EDITS: In section 10.8.1, add the following as a new paragraph, just before Note 10.26 [175:33+]. "For the form of an input value, the constant is interpreted as a nondelimited character constant if the first list item corresponding to this value is of type default character, there is a nonblank character immediately after , and that character is not an apostrophe or a quotation mark; otherwise, is interpreted as a literal constant." SUBMITTED BY: Robert Corbett HISTORY: 98-155 m145 Submitted (part 3) WG5/N1410 Draft answer 01-159r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000028 TITLE: Implicitly Typed Statement Function Character Dummy KEYWORDS: statement function, constant expression, implicit typing DEFECT TYPE: Clarification STATUS: Passed by J3 letter ballot QUESTION: Is the following program-unit standard-conforming? SUBROUTINE sub(n) IMPLICIT CHARACTER*(n) (c) stmtfn(ch) = LEN(ch) END A quick survey of 6 compilers showed no consensus on whether this was legal, with 3 accepting it (as standard-conforming), 2 rejecting it, and 1 crashing with an internal error. The standard seems to say (middle of section 5.1.1.5, [51:32-33]) that if you explicitly specify the length then it must be a constant specification expression, but does not appear to constrain implicit lengths. ANSWER: No, the example is not standard-conforming. DISCUSSION: The quoted section of the standard says: "The length specified for a ... statement function dummy argument of type character shall be a constant specification expression." The first paragraph of 5.3 says "In a scoping unit, an IMPLICIT statement specifies a type, and possibly type parameters, for all implicitly typed data entities ..." Therefore the requirement that the character length of a statement function or statement function dummy argument be constant applies to implicitly typed entities as well as explicitly typed entities. EDITS: None SUBMITTED BY: Malcolm Cohen HISTORY: 98-164 m145 Submitted 01-149 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000029 TITLE: Nested Derived Types and Defined Assignment KEYWORDS: derived type, defined assignment DEFECT TYPE: Clarification STATUS: Passed by J3 letter ballot QUESTION: According to Fortran 90, intrinsic assignment of derived type entities does not invoke defined assignment for components of these entities, even if it exists. Should this situation have been corrected in Fortran 95 to avoid breaking encapsulation of derived types? That is, in the example program below, is the final assignment statement intended to invoke the user-defined assignment routine for the "name" components of "a" and "b", or should it do pointer assignment on the (PRIVATE) "value" components of the "name" components? MODULE lib_varying_string TYPE varying_string PRIVATE CHARACTER,POINTER :: value(:) END TYPE INTERFACE ASSIGNMENT(=) MODULE PROCEDURE assign_vs_to_vs END INTERFACE CONTAINS SUBROUTINE assign_vs_to_vs(var,expr) TYPE(varying_string),INTENT(OUT) :: var TYPE(varying_string),INTENT(IN) :: expr ALLOCATE(var%value(SIZE(expr%value,1))) var%value = expr%value END SUBROUTINE END PROGRAM programme USE lib_varying_string TYPE mytype TYPE(varying_string) name END TYPE TYPE(varying_string) x,y TYPE(mytype) a,b x = y ! invokes "assign_vs_to_vs(x,(y))". a%name = b%name ! invokes "assign_vs_to_vs(a%name,(b%name))" a = b ! in F90, pointer assigns the "value" components END ANSWER: No, there is no ambiguity or error in the standard. DISCUSSION: Intrinsic assignment is not intended to invoke user-written procedures. This would introduce anomalies with initialization by DATA statements. Library writers cannot in any case depend on defined assignment to enforce encapsulation automatically, because the user can deliberately (or accidentally) omit to import the defined assignment from the module. (Also, other facilities such as TRANSFER may break encapsulation). In the example given, a defined assignment should be provided for TYPE(mytype) by the programmer. EDITS: None. NOTE: This is MTE work item M16 for Fortran 2000. SUBMITTED BY: Malcolm Cohen HISTORY: 97-197 m142 Submitted 97-197 m145 Passed unanimously as amended {see 98-167 minutes} Failed WG5 letter ballot. 00-330 m155 Option 0 answer passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000081 TITLE: Definition status of derived-type objects with pointer components KEYWORDS: Definition status, Derived type, Pointer component DEFECT TYPE: Clarification STATUS: Passed by J3 letter ballot QUESTION: Consider: TYPE t REAL,POINTER :: x END TYPE TYPE(t) var1,var2 NULLIFY(var1%x) var2 = var1 ! Unfortunate. According to 14.7.1, "An object is defined if and only if all of its subobjects are defined." from which I conclude that the program fragment above is not conforming. Should the definition status of an object containing pointer components depend on the pointer association status of its pointer components and not their definition status? ANSWER: No, pointer components shall not affect the definition status of an object. The above program fragment was intended to be standard- conforming. Edits are supplied to clarify this situation. EDITS: [288:17] Replace with "(1) An array is defined if and only if all of its elements are defined. (2) A derived-type scalar object is defined if and only if all of its nonpointer components are defined. (3) A complex or character scalar object is defined if and only if all of its subobjects are defined." [288:18-19] Replace "(2)" with "(4)". SUBMITTED BY: Malcolm Cohen HISTORY: 99-266 m151 submitted 99-266r1 m151 approved uc 00-135 m152 revised but withdrawn (no vote) 01-165r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000085 TITLE: Public components of inaccessible types KEYWORDS: Accessibility, Derived Types, Components DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: F90 interpretation 32 indicated that components of a derived type may be accessible even in scopes where the derived type name is inaccessible. It added an edit to 12.1.2.2.1 of f90 (14.6.1.3 of f95) saying "If a derived type name of a host is inaccessible, data entities of that type or subobjects of such data entities can still be accessible." The discussions in the last 2 paragraphs of 4.4.1 of f95 appear to treat the accessibility of the derived type and its components as independent issues. Drafts of f2k added a note to make the independence clear. However, 14.1.2.5 of f95 says "If the type is accessible...and the definition of the type does not contain the PRIVATE statement (4.4.1), the component name is accessible...." This statement appears to make accessibility of the type a condition for accessibility of the component names, in contradiction with the above cited edit from f90 interpretation 32. The following code sample illustrates the issue. This sample works on several f90 compilers. module m type t integer :: i end type t type (t) :: a end module m program p use m, only: a a%i = 2 write (*,*) a%i end program p Question 1. Is it allowed to access public components of a derived type object when the type name is inaccessible as illustrated in the above sample code? Question 2. Should there be an edit in 14.1.2.5 of f95 to clarify this? ANSWER: (1) Yes. (2) Yes. DISCUSSION: The quoted text from 14.1.2.5 does not specify the accessibility of components when a type is not accessible. EDITS: [280:25-26] Replace "the type is accessible ... 14.6.1.3)" by "an entity of the type is accessible in another scoping unit" {Whether by use or host association, or - if the type is accessible - by declaring a local variable of the type.} SUBMITTED BY: Richard Maine HISTORY: 99-101 m148 Submitted WG5/N1404 Draft answer 00-260 m154 Passed by J3 meeting 00-329 m155 Failed J3 letter ballot 01-160 m156 Passed unanimously at J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000087 TITLE: MOD and MODULO intrinsic functions with zero divisor KEYWORDS: Intrinsic functions, MOD, MODULO DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Consider the programs: PROGRAM example1 PRINT *,MODULO(3,0) PRINT *,'Test succeeded' END PROGRAM example2 PRINT *,MOD(3,0) PRINT *,'Test succeeded' END Are these programs standard-conforming, and should "Test succeeded" be printed in each case? DISCUSSION: 13.14.74 says of MODULO: "If P=0, the result is processor dependent." 13.14.73 says of MOD: "If P=0, the result is processor dependent." It appears that example1 and example2 are equally conforming. INFORMATION: Out of 7 Fortran 90/95 compilers, with example1: 5 compiled it ok, 3 of these crashed it at runtime example2: 2 compiled it ok, both of these crashed it at runtime Out of 1 FORTRAN 77 compiler, example2 did not compile (error message). Note that section 7.1.7 [97:6-7] states "The execution of any numeric operation whose result is not defined by the arithmetic used by the processor is prohibited." This would seem to be contradictory, at least in spirit, to the definitions of MOD and MODULO. ANSWER: No, these programs were not intended to be standard-conforming. The processor should be allowed to optimise assuming that P is not zero. Edits are supplied to correct this mistake in the standard. EDITS: [3:32+] "(3) Earlier standards specified that if the second argument to MOD or MODULO was zero, the result was processor dependent. This standard specifies that the second argument shall not be zero." [257:30] Append "P shall not be zero." [257:32] Change "If P .ne. 0, the" to "The" [257:32-33] Delete "If P = 0, the result is processor dependent." [257:41] Append "P shall not be zero." [258:2] Delete "If P .ne 0," [258:4] Delete "If P = 0, the result is processor dependent." [258:5] Change "If P .ne. 0, the" to "The" [258:5-6] Delete "If P = 0, the result is processor dependent." SUBMITTED BY: Malcolm Cohen HISTORY: 00-325 m155 Passed unanimously as amended by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000088 TITLE: INTRINSIC statement and attribute KEYWORDS: intrinsic DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Is it possible for a variable or external procedure that has the same name as the specific name of an intrinsic function listed in section 13.13 of the Fortran 95 standard and not marked with a bullet to be used as an actual argument? For example, consider the following program: module SCREEN type :: COLOR integer :: RED, GREEN, BLUE end type COLOR contains subroutine SET_BACKGROUND ( WHAT ) type(color), intent(in) :: WHAT ! Call to set the background color end subroutine SET_BACKGROUND end module SCREEN program TEST_SCREEN use SCREEN real, external :: DIM character(len=15) :: SIGN real :: SIN type(color) :: TAN = color(210,180,140) call color_intensity ( dim ) call biblical ( sign ) call cardinal ( sin ) call set_background ( tan ) end program TEST_SCREEN Notice that the actual arguments have the names of specific intrinsic functions listed in section 13.13 of the Fortran 95 standard, but they do not appear in an INTRINSIC statement, nor are they given the INTRINSIC attribute in a type declaration statement, as required by the penultimate normative paragraph of 12.3.2.3 of the Fortran 95 standard [198:7-9]. ANSWER: The answer to the question in the first paragraph is yes. Fortran has no reserved words. In the scope of the above main program, none of the names DIM, SIGN, SIN and TAN are names of intrinsic procedures. The paragraph referenced reads: If the specific name of an intrinsic function (13.13) is used as an actual argument, the name shall either appear in an INTRINSIC statement or be given the INTRINSIC attribute in a type declaration statement in the scoping unit. and tells us that SIN does not refer to the intrinsic procedure. The names DIM, SIGN, and TAN can be identified from their declarations not to be intrinsic procedures. Since none of the names are those of intrinsic procedures, the paragraph does not affect their use as the names of actual arguments. The edit makes the meaning clearer. EDITS: Page 198, Subclause 12.3.2.3, replace lines 7-9 of page 198 [198:7-9] by 'If a specific intrinsic function (13.13) is used as an actual argument, it shall have been explicitly declared to have the INTRINSIC attribute.' SUBMITTED BY: Van Snyder HISTORY: 00-249 m154 Submitted 01-136r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000089 TITLE: Rules allowing duplicate names KEYWORDS: generic, duplicate names DEFECT TYPE: Clarification STATUS: Passed by J3 letter ballot QUESTION: Under what conditions can a name in class (1) as described in 14.1.2 be used in more than one declaration? The second complete paragraph after the list in 14.1.2, immediately before note 14.2 [276:5-7 in 97-007r2] begins: ``Within a scoping unit, a name that identifies a local entity of one class shall not be used to identify another local entity of the same class, except in the case of generic names (12.3.2.1).'' Is the intent here that a generic name may be used only to identify another generic interface, or that a generic name may be used to identify any other entity in class (1)? ANSWER: The rules are laid out in 12.3.2.1, which is referenced. The exceptional case is that a procedure is permitted to have the same specific and generic name. EDITS: None SUBMITTED BY: Van Snyder HISTORY: 00-246 m154 Submitted 01-137r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000090 TITLE: What do ``Prior Specification'' and ``defined previously'' mean? KEYWORDS: initialization, prior specification, defined previously DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION 1 What is a ``prior specification?'' The last normative paragraph of section 7.1.6.1 of the Fortran 95 standard, immediately before note 7.14 [94:38-41 in 97-007r2] states: ``If an initialization expression includes a reference to an inquiry function for a type parameter or an array bound of an object specified in the same , the type parameter or array bound shall be specified in a prior specification of the . The prior specification may be to the left of the inquiry function in the same statement.'' The first two sentences of the last normative paragraph of section 7.1.6.2 of the Fortran 95 standard, immediately before note 7.16 [96:32-35 in 97-007r2] are similar, but refer to specification expressions. Unfortunately, there is no definition of ``prior specification.'' Consider: 1. INTEGER :: P(complicated_expression_for_lower_bound_1: & & complicated_expression_for_upper_bound_1, & & complicated_expression_for_lower_bound_2: & & complicated_expression_for_upper_bound_2) = & & RESHAPE( (/ 11, 21, 12, 22 /), SHAPE(P) ) (Notice the reference to P in the .) Some processors reason ``the `specification' of P is only the specification of its name, kind, rank and array bounds, and therefore is `prior' by the time the is encountered,'' and accept this. Others reason ``the `specification' of P is the syntax term defined by syntax rule R504 on page 47, which includes the , and it is therefore not `prior' until the is completely specified,'' and do not accept this. By way of further examples, which of the following are standard conforming? 2. INTEGER(selected_int_kind(4)) :: A(KIND(A)) 3. INTEGER :: A(2,2*SIZE(A,1)+1) 4. CHARACTER :: C(10)*(SIZE(C,1)) 5. INTEGER :: P(10) = LBOUND(P,1) The paragraphs cited above are silent concerning inquiry functions that are not for type parameters or array bounds. These include BIT_SIZE, DIGITS, EPSILON, MAXEXPONENT, MINEXPONENT, PRECISION, RADIX, RANGE and TINY. The results of all of these are derived, however, from the types and kind type parameters of their arguments. Are the objects about which these functions inquire required to be specified in a prior specification of the , and is that specification allowed to be in the same statement so long as it is to the left of the inquiry function of which it is an argument? Which of the following are standard conforming? 6. INTEGER :: B = BIT_SIZE(B) 7. INTEGER :: B(BIT_SIZE(B)) 8. INTEGER :: D = DIGITS(D) 9. INTEGER :: D(DIGITS(D)) 10. REAL :: X = EPSILON(X) One could construct obvious similar examples for the remainder of inquiry functions that do not inquire about type parameters or array bounds. QUESTION 2 The second normative paragraph in subclause 5.1.2.1 [52:27-28 in 97-007r2] states "Any named constant that appears in the initialization expression shall have been defined previously in the same type declaration statement...." The third normative paragraph after syntax rule R531 in subclause 5.2.9 [61:4-7 in 97-007r2] is similar. Does "defined previously" mean that all of the properties of the named constant shall have been defined previously, or that a referenced property of it shall have been defined previously? E.g. if the KIND is needed, is it necessary for the value to have been defined? ANSWER 1 A prior specification refers to a specification in a previous or in a previous statement. None of the examples are legal. Edits are included to clarify this. ANSWER 2 In the referenced text (5.1.2.1), "Any named constant that appears ... shall have been defined previously" mean that all of the properties of the named constant shall have been defined previously; therefore, if the KIND of a named constant is needed, is it necessary for the value of that named constant to have been defined previously also. EDITS: Page 94, Subclause 7.1.6.1. In the first line of the last paragraph of page 94 [94:38], replace 'for a type parameter' by 'that depends on a type parameter'. Page 94, Subclause 7.1.6.1, replace the last sentence of page 94 [94:40-41] by 'The prior specification may be to the left of the inquiry function in the same statement, but shall not be within the same .' Page 96, Subclause 7.1.6.2. In the first line of the last paragraph of the subclause [96:32], replace 'for a type parameter' by 'that depends on a type parameter'. Page 96, Subclause 7.1.6.2, replace the second sentence of the last paragraph of the subclause [96:34-35] by 'The prior specification may be to the left of the inquiry function in the same statement, but shall not be within the same .' SUBMITTED BY: Van Snyder HISTORY: 00-229 m154 Submitted 00-324 m155 Amended 01-138r1 m156 Passed by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000091 TITLE: Definition of "present" is defective KEYWORDS: present, dummy argument DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: Subclause 12.4.1.5 begins with a definition of the term "present": "A dummy argument is present in an instance of a subprogram if it is associated with an actual argument and the actual argument either is a dummy argument that is present in the invoking subprogram or is not a dummy argument of the invoking subprogram." This seems to be defective in that if the actual argument is a dummy argument that is accessed by host association, it is not a dummy argument OF the invoking subprogram, and therefore although it may be "not present" in the procedure of which it is a dummy argument, it is nonetheless "present" in an internal procedure of that procedure. Consider: module M contains subroutine S1 ( A ) integer, optional :: A call S11 contains subroutine S11 call S12 ( A ) end subroutine S11 subroutine S12 ( B ) integer, optional :: B if ( present(B) ) print *, 'B is present' end subroutine S12 end subroutine S1 end module M program P use M, only: S1 call S1 end program P Should this print "B is present" as implied by the above analysis? ANSWER: No, the program should not print anything. DISCUSSION: The lack of consideration of host association in the cited text is accidental; an edit is supplied to correct this oversight. EDITS: [202:43-45] Replace the first sentence of 12.4.1.5 by "A dummy argument is not <> if it is (1) not associated with an actual argument, or (2) is associated with an actual argument that is (a) a dummy argument that is not present or (b) an entity that is host-associated with a dummy argument that is not present. Otherwise, it is present." [299:4-6] Delete the glossary entry for <> SUBMITTED BY: Van Snyder HISTORY: 01-134 m156 Submitted 01-134r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000092 TITLE: Values of the PAD= Specifier in the INQUIRE Statement KEYWORDS: PAD=, INQUIRE DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: The normative text of the Fortran 90 and Fortran 95 standards (9.6.1.22) specifies that the PAD= specifier in the INQUIRE statement returns the value YES if there is no connection or if the connection is for unformatted input/output. This behavior is different from the results of almost every other specifier for the INQUIRE statement if the property being inquired about is unknown or undefined. For those specifiers, the value returned in the specifier's variable is either UNKNOWN or UNDEFINED. Is the specified behavior of the PAD= specifier for an unformatted file or an unconnected unit intentional? ANSWER: No, this specification is in error. Edits are supplied to correct this defect. DISCUSSION: The concept of padding input records does not make any sense if the unit is connected for unformatted input/output or if the unit is not connected. A return value of YES does not reflect what is really going on. A return value of UNDEFINED is more appropriate in these situations. The normative text for this section in Fortran 95 is unchanged from Fortran 90, so correcting this defect is an incompatibility with Fortran 90. EDITS: [3:32+] Add the following list item to the list in section 1.5.1, "Fortran 90 compatibility". (3) The PAD= specifier in the INQUIRE statement in this standard returns the value 'UNDEFINED' if there is no connection or the connection is for unformatted input/output. The previous standard specified 'YES'. [159:40-41] Replace the second sentence of the paragraph in section 9.6.1.22 with the following. The in the PAD= specifier is assigned the value YES if the connection of the file to the unit included the PAD= specifier and its value was YES or if there was no PAD= specifier. If there is no connection or if the connection is not for formatted input/output, the is assigned the value UNDEFINED. [325:30] Change the entry in the table to: +--------+-----------+------------------------+-----------+ | PAD= | UNDEFINED | YES, NO, or UNDEFINED | UNDEFINED | +--------+-----------+------------------------+-----------+ SUBMITTED BY: Craig T. Dedo HISTORY: 01-172 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: 000093 TITLE: Allocatable arrays as actual arguments KEYWORDS: Allocatable array, actual argument DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: There does not seem to be any requirement that an allocatable array be allocated for it to be used as an actual argument. Consider PROGRAM P ALLOCATABLE A(:,:) CALL S(A) CONTAINS SUBROUTINE S(X) DIMENSION X(:,:) PRINT *,SHAPE(X) END SUBROUTINE END Is this a standard-conforming program? If so, what does it print? ANSWER: No, this is not intended to be a standard-conforming program. An allocatable array must be allocated for it to be used as an actual argument, except where otherwise specified in the standard (e.g. for the ALLOCATED intrinsic function. An edit is supplied to clarify this intent. EDITS: [80:34] In item (1) in the list in clause 6.3.1.2, change "; it" to ". It shall not be supplied as an actual argument except to certain intrinsic inquiry functions. It" SUBMITTED BY: Malcolm Cohen HISTORY: 01-184 m156 Passed unanimously at J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- NUMBER: JP-06 TITLE: Type declaration statements KEYWORDS: Object name, Type declaration DEFECT TYPE: Erratum STATUS: Passed by J3 letter ballot QUESTION: JP-6) 5.1 Type declaration statements As for the 16th and 19th constraint after R506: the 16th [48:25]: "Constraint: The function-name shall be the name of an external function, an intrinsic function, a function dummy procedure, or a statement function." Because the syntactic class `object-name' is only defined as a `name' in the standard, the following constraint should be added here: Constraint: The object-name shall be the name of a data object. [48:26+] After that, in the 19th [48:31]: "Constraint: initialization shall not appear if object-name is dummy argument, a dummy argument, a function result, an object in a named common block unless the type declaration is in a block data program unit, an object in blank common, an allocatable array, an external name, an intrinsic name, or an automatic object." In the above, `a function result,' should be removed. If we can not add the constraint above, `a statement function' should be added in the 19th constraint. DISCUSSION: There is a typographical error in the quoted 19th constraint in the question in that "dummy argument" appears twice. ANSWER: Yes, we agree that the constraint above should be added. However, the 19th constraint should not be modified since a function result is a data object that cannot be initialized. EDITS: [48:26+] Add: "Constraint: The object-name shall be the name of a data object." SUBMITTED BY: Japan HISTORY: 99-208 Submitted 99-221 Classed as Fortran 95 Interpretation. WG5-N1411 draft answer 01-151r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Passed by J3 letter ballot ---------------------------------------------------------------------- ********************************************************************** NUMBER: 000023 TITLE: Termination of the previous record by a WRITE statement KEYWORDS: Nonadvancing, Record termination, WRITE DEFECT TYPE: Clarification STATUS: J3 consideration in progress QUESTION: Suppose a sequential formatted file is positioned within a record as a result of a nonadvancing READ. 1. If the next operation performed on the file is a WRITE, can the current record be terminated at its current position before the next record is written? Section 9.2.1.3.2 appears to preclude the record from being terminated, but at least one Fortran 90 implementation does terminate the record in such an instance. 2. If the next operation is a nonadvancing WRITE and the number of characters written does not extend to the end of the current record, are the remaining characters at the end of the record preserved? ANSWER: 1. No. If the current record were to be terminated, the file position would be changed. Section 9.2.1.3.2 [136:37-39] states: "For sequential access on input, if there is a current record, the file position is not changed. Otherwise, the file is positioned at the beginning of the next record and this record becomes the current record." 2. Yes, the remaining characters in the record are preserved. The record read by the non-advancing READ is the current record and file position is within that current record. The WRITE overwrites characters starting at the current file position but all the characters in the current record are preserved. For example: WRITE (1, '(A3)') 'ABCDEFGHI' REWIND (1) READ (1, '(A3)', ADVANCE='NO') THREE_CHARS WRITE (1, '(A3)', ADVANCE='NO') 'XXX' CLOSE (1) would create a file with the single record 'ABCXXXGHI'. EDITS: None SUBMITTED BY: Robert Corbett HISTORY: 98-155 m145 Submitted (part 1) WG5/N1410 Draft answer 01-151r1 m156 Passed unanimously by J3 meeting 01-224r1 m157 Failed J3 letter ballot ----------------------------------------------------------------------