Trailing-Edge
-
PDP-10 Archives
-
FORTRAN-10_V7wLink_Feb83
-
fortra.doc
There are 10 other files named fortra.doc in the archive. Click here to see a list.
FORTRAN-10 -- Changes from V6 to V7
February 1983
COPYRIGHT (C) 1983 BY
DIGITAL EQUIPMENT CORPORATION, MAYNARD, MASS.
THIS SOFTWARE IS FURNISHED UNDER A LICENSE AND MAY BE USED AND COPIED
ONLY IN ACCORDANCE WITH THE TERMS OF SUCH LICENSE AND WITH THE
INCLUSION OF THE ABOVE COPYRIGHT NOTICE. THIS SOFTWARE OR ANY OTHER
COPIES THEREOF MAY NOT BE PROVIDED OR OTHERWISE MADE AVAILABLE TO ANY
OTHER PERSON. NO TITLE TO AND OWNERSHIP OF THE SOFTWARE IS HEREBY
TRANSFERRED.
THE INFORMATION IN THIS SOFTWARE IS SUBJECT TO CHANGE WITHOUT NOTICE
AND SHOULD NOT BE CONSTRUED AS A COMMITMENT BY DIGITAL EQUIPMENT
CORPORATION.
DIGITAL ASSUMES NO RESPONSIBILITY FOR THE USE OR RELIABILITY OF ITS
SOFTWARE ON EQUIPMENT WHICH IS NOT SUPPLIED BY DIGITAL.
FORTRA.DOC - FORTRAN-10 Version 7 Page 2
CONTENTS
1.0 SUMMARY . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Bibliography . . . . . . . . . . . . . . . . . . . 3
2.0 NEW FEATURES . . . . . . . . . . . . . . . . . . . . 4
2.1 Summary Of Enhancements To The Product . . . . . . 4
2.2 Character Data . . . . . . . . . . . . . . . . . . 6
2.2.1 Features Supported For Character Data . . . . . 6
2.2.2 Internal Representation Of Character Data . . . 9
2.2.3 FORDDT Support For Character Data . . . . . . . 9
2.2.3.1 Character Mode For TYPE And ACCEPT Commands . 9
2.2.3.2 Declaration Of Character Array Bounds . . . 10
2.3 IF THEN ELSE . . . . . . . . . . . . . . . . . . 10
2.4 DO WHILE And END DO Statements . . . . . . . . . 10
2.5 Expressions On Output Lists . . . . . . . . . . 11
2.6 Intrinsic And Generic Functions . . . . . . . . 11
2.7 Multi-record Internal Files . . . . . . . . . . 12
2.8 FORTRAN-77 DO Loop Semantics . . . . . . . . . . 12
2.9 Assumed-Size Array Declarators . . . . . . . . . 13
2.10 ASSIGN Of FORMAT Statement Numbers . . . . . . . 13
2.11 INTRINSIC And EXTERNAL Statements . . . . . . . 14
2.12 SAVE Statement . . . . . . . . . . . . . . . . . 14
2.13 Support Of Null Argument Lists For Functions . . 14
2.14 FORTRAN-77 Statement Ordering Rules . . . . . . 15
2.15 Minor Syntax Extensions . . . . . . . . . . . . 15
2.16 Constant Expressions . . . . . . . . . . . . . . 16
2.17 FORTRAN-77 PARAMETER Statements . . . . . . . . 16
2.18 Expressions As Specifier Values In OPEN And I/O
Statements . . . . . . . . . . . . . . . . . . . 17
2.19 Type Checking For Function And Subroutine
Arguments . . . . . . . . . . . . . . . . . . . 17
2.20 New Functionality In ERRSET . . . . . . . . . . 20
2.21 Utility Subroutine To Get A Free Unit Number . . 22
2.22 Additional Runtime Warnings . . . . . . . . . . 22
3.0 RESTRICTIONS . . . . . . . . . . . . . . . . . . . 22
3.1 FORTRAN-77 Features Not Supported . . . . . . . 22
3.2 Other Restrictions . . . . . . . . . . . . . . . 23
4.0 KNOWN BUGS AND DEFICIENCIES . . . . . . . . . . . 23
4.1 Execute-Only . . . . . . . . . . . . . . . . . . 23
4.2 Mandatory Patches For Related Products . . . . . 25
4.3 Known Deficiencies In Related Products . . . . . 25
5.0 COMPATIBILITY WITH PREVIOUS VERSIONS . . . . . . . 26
5.1 Incompatibilities Between Versions 6 And 7 . . . 26
5.2 Compiler "FORTRAN-77" Switch . . . . . . . . . . 32
5.3 FOROTS Interface . . . . . . . . . . . . . . . . 33
5.4 Subprogram Calling Conventions . . . . . . . . . 33
5.5 Using DBMS With FORTRAN V7 . . . . . . . . . . . 34
5.6 Incompatibilities Between Versions 5A And 6 . . 34
6.0 INTERNAL CHANGES . . . . . . . . . . . . . . . . . 38
FORTRA.DOC - FORTRAN-10 Version 7 Page 3
FORTRAN-10 -- Changes from V6 to V7
February 1983
1.0 SUMMARY
Version 7 of FORTRAN-10 is a development release of the FORTRAN
compiler, FOROTS, FORLIB, and FORDDT.
FORTRAN-10 version 7 supports the FORTRAN-77 standard at the subset
level. It also supports most of the features from the FORTRAN-77 full
language standard (see section 3.1 below). Character data is
supported in accordance with the FORTRAN-77 full language standard.
Note that you must use LINK version 5.1 (or later) for linking the REL
files generated by this version of FORTRAN.
1.1 Bibliography
The FORTRAN language and this implementation are described in the
reference manual:
FORTRAN-10/20 Language Manual - AA-N383A-TK
Two copies of this manual are enclosed with this release package.
The installation procedures for this tape are described in:
TOPS-10 FORTRAN Installation Guide - AA-P345A-TB
A copy of this installation is enclosed with this release package.
There is also a booklet that provides quick reference information
about FORTRAN-10/20 version 7:
TOPS-10/20 FORTRAN Reference Card - AV-P529A-TK
Two copies of this booklet are enclosed with this release package.
The algorithms and accuracy of the math routines in FORLIB will be
described in:
TOPS-10/20 Common Math Library Reference Manual - AA-M400A-TK
This manual will be available from SDC in May 1983.
FORTRAN-10/20 Version 7 is based on the FORTRAN-77 standard:
FORTRA.DOC - FORTRAN-10 Version 7 Page 4
American National Standard Programming Language FORTRAN - ANSI
X3.9-1978
Copies of this standard may be obtained from:
American National Standards Institute
1430 Broadway
New York, N.Y. 10018
1.2 Monitors
This version of Fortran will run on TOPS-10 versions 7.01, 7.01A, or
later
1.3 Related System Software
The following system software must be used with FORTRAN version 7:
LINK 5A(2027)
GALAXY 2 or 4.1
The following system software must be used to build FORTRAN Version 7
from sources:
BLIS10 7E(227)
MACRO 53A(1152)
MAKLIB 2B(104)
SCAN 7D(623)
HELPER 7D(47)
The following optional software may be used with FORTRAN Version 7:
SORT 4C(467)
DBMS 5A(422)
2.0 NEW FEATURES
2.1 Summary Of Enhancements To The Product
The following is a summary of the major features that in FORTRAN-10/20
version 7 that were not in version 6:
1. Character data at the FORTRAN-77 full language level.
Supported features include character assignments, character
relationals, substrings, concatenation, and character
functions and arguments, including functions and dummy
FORTRA.DOC - FORTRAN-10 Version 7 Page 5
arguments of length *. Character data is supported in DATA,
COMMON, and EQUIVALENCE statements, and in formatted, binary,
and image mode I/O. See section 2.2 below for more details.
2. IF THEN ELSE
3. DO WHILE and END DO statements
4. Expressions on output lists
5. Intrinsic and generic functions at the FORTRAN-77 full
language level
6. Internal files ( single-record and multi-record )
7. FORTRAN-77 DO loop semantics
8. Assumed-size array declarators
9. ASSIGN of FORMAT statement numbers
10. INTRINSIC statement; FORTRAN-77 semantics for the EXTERNAL
statement
11. SAVE statement
12. Support of null argument lists for functions
13. Support of FORTRAN-77 statement ordering rules.
14. Minor syntax extensions required by the FORTRAN-77 standard
15. Compiletime constant expressions in declarations, as array
bounds and string bounds
16. FORTRAN-77 PARAMETER statements
17. Expressions as specifier values in OPEN statements and in
control information lists in I/O statements
18. Optional linktime typechecking of subprogram arguments
19. New functionality in the ERRSET subroutine
20. Utility subroutine to get a free unit number
21. Additional runtime warnings
FORTRA.DOC - FORTRAN-10 Version 7 Page 6
2.2 Character Data
2.2.1 Features Supported For Character Data -
FORTRAN-10/20 version 7 supports character data as specified by the
full language FORTRAN-77 standard. This includes support for the
following features (please see the FORTRAN-10/20 Language Manual for
more information about these features):
1. Character constants and symbolic constants
2. Character variables and arrays
3. Character relational expressions
4. Character assignment statements
Note that in addition to meeting the requirements of the
FORTRAN-77 standard, version 7 supports the following
extensions:
o Version 7 supports the assignment of character constants
to non-character variables. This is an extension to the
FORTRAN-77 standard. It is supported for compatibility
with previous versions.
o Version 7 supports assignment statements in which there
is overlap between the left and right hand sides. The
results of such an assignment will be as if the
expression on the right hand side were assigned to a
temporary and then the value of the temporary were
assigned to the left hand side.
5. Substring reference and definition
Note that the user may optionally specify that substring
bounds checking be performed at execution time. The
/DEBUG:BOUNDS switch enables both array bounds and substring
bounds checking.
6. Character expressions, including the concatenation operator
Note that version 7 extends the FORTRAN-77 standard to allow
concatenation of formal parameters that are length *. This
extension is also provided by VAX FORTRAN.
7. Character variables and constants in DATA statements
FORTRAN version 7 supports the initialization of character
variables, arrays, and substrings to character constant
values as specified by the FORTRAN-77 standard. It also
supports the use of character constants to initialize
non-character variables. This is an extension to the
FORTRAN-77 standard, which is supported for compatibility
with previous versions.
FORTRA.DOC - FORTRAN-10 Version 7 Page 7
8. Character variables, array elements, arrays, and substrings
in input lists
According to the FORTRAN-77 standard, the A edit descriptor
is the only legal edit edit descriptor for input/output list
items if type character. Version 7 extends the standard to
allow the G edit descriptor as well. The G edit descriptor
behaves like the A edit descriptor for list items of type
character. Note that R edit descriptors (which are a DEC
extension to the standard) are NOT supported for character
data.
9. Character constants, variables, array elements, arrays,
substrings, and expressions in output lists
(Note: Character data is supported in formatted, binary, and
image mode files. It is not supported in dump mode files,
because character data does not necessarily begin on a word
boundary. )
10. Character functions (intrinsic functions, external functions,
and statement functions that return values of type character)
In the object code for a call to a character function, the
character descriptor for the return value is passed as what
is normally the first argument. If you write a function of
type character in assembly language, you can use
DMOVE AC,@0(16)
to load AC with a byte pointer to the location to which the
result should be returned, and AC+1 with the length of the
result expected by the caller. The first explicit argument
is obtained from @1(16).
11. Dummy and actual arguments of type character
FORTRAN-10/20 extends the standard to provide support of
character constants as actual arguments corresponding to
dummy arguments that are integer, real, double precision,
complex, or logical as well as character. This support is
provided by means of a limited type checking facility that is
supported by LINK version 5.1. Note that this feature only
works when both caller and callee have been compiled with
FORTRAN version 7. If a subprogram was compiled with an
earlier version of FORTRAN, it will not contain the REL file
blocks that LINK uses to recognize that a fix-up must be
performed on the character argument.
This feature does not work when the name of the function
called is itself a dummy argument, and therefore is not known
at compiletime or linktime.
If an actual argument is of type character and is not a
constant, then the corresponding dummy must be of type
FORTRA.DOC - FORTRAN-10 Version 7 Page 8
character. If a dummy argument is of type character, then
the corresponding actual must be of type character.
Actual arguments may be longer than corresponding dummy
arguments. Length * may be used for character dummy
arguments
12. FORTRAN-77 intrinsic functions that operate on character data
LEN, INDEX, CHAR, ICHAR, LGE, LGT, LLE, LLT
13. Character variable names, array element names, array names,
and substring names in EQUIVALENCE statements
It is illegal to equivalence a numeric variable to a
character variable. This is a restriction in both the
FORTRAN-77 standard and VAX FORTRAN. This helps to enforce
transportable coding practices.
The error message for EQUIVALENCE of a numeric variable to an
unaligned character variable is fatal. The error message for
EQUIVALENCE of a numeric variable to a word-aligned character
variable is non-fatal.
14. Character variables and arrays in COMMON blocks
If a character variable or character array is in a COMMON
block, all of the entities in that common block must be of
type character. This is a restriction in both the FORTRAN-77
standard and VAX FORTRAN. Specification of a COMMON block
that contains both character and numeric data will result in
a non-fatal warning message. All variables other than
character variables will always begin on a word boundary.
Thus if a COMMON block erroneously contains both character
and numeric data, there may be unused character positions.
15. Character variables and array names in namelists
Version 7 supports substrings in namelist input, but not in
namelist output.
16. Character variables and expressions as the values of the
specifiers in OPEN and CLOSE statements
17. Character constants, character arrays, and character
variables as format specifiers
Version 7 will not support global optimization of program units that
contain character data. If the user specifies the /OPTIMIZE switch
for such a program the following warning diagnostic will be issued:
%FTNCHO Global optimization not yet supported with character data - /OPT ignored
This message will be issued once for each subprogram that contains
character data.
FORTRA.DOC - FORTRAN-10 Version 7 Page 9
2.2.2 Internal Representation Of Character Data -
Character strings are represented in 7-bit ASCII. Therefore, there
are 5 characters per word and the collating sequence is the ASCII
collating sequence.
The object code generated by FORTRAN-10/20 version 7 includes
descriptors for character variables, primaries, and subprogram
arguments. The form of a character data descriptor is:
0 17 18 35
------------------------------------------
! byte pointer !
------------------------------------------
! character count !
------------------------------------------
This is the same format of descriptor as is used for COBOL subroutine
arguments.
The DDT/FORDDT symbol table entry for a character variable specifies
the location of its character data descriptor.
2.2.3 FORDDT Support For Character Data -
2.2.3.1 Character Mode For TYPE And ACCEPT Commands -
A mode to enable ACCEPT and TYPE for character data has been added to
FORDDT.
In FORDDT version 7, /C means character data mode, indicating to
FORDDT that the symbol table pointer is pointing to a character data
descriptor, rather than at the character variable itself. Note that
/C was used by earlier versions of FORDDT to indicate COMPLEX data
mode, which is now indicated by /X.
If the ACCEPT command has a bare array name followed by a character
constant, all the array elements will be filled with the character
constant, truncated or space-padded in accordance with the rules for
character constants as per the ANSI standard.
For example, if the FORTRAN source program contains the declaration:
CHARACTER*3 C1
You can type out the contents of C1 from FORDDT by saying:
TYPE C1/C
You can store 'ABC' in C1 from FORDDT by saying:
FORTRA.DOC - FORTRAN-10 Version 7 Page 10
ACCEPT C1/C 'ABC'
2.2.3.2 Declaration Of Character Array Bounds -
If a program has not been compiled with the /DEBUG switch, a FORDDT
user must declare his array dimensions at runtime. For numeric
arrays, the FORDDT "DIMENSION" and "DOUBLE" commands provide this
capability.
In version 7, there is a new FORDDT command "CHARACTER" which is used
to declare the dimensions of a character array. The syntax of this
command is identical to that of the "DIMENSION" command. Note that
the string length for the array elements is always determined by the
user's program. It cannot be declared in FORDDT.
2.3 IF THEN ELSE
Version 7 supports the Block IF, ELSE IF, ELSE, and END IF statements
as specified by the full FORTRAN-77 standard.
2.4 DO WHILE And END DO Statements
The DO WHILE and END DO statements are useful for writing well
structured programs.
DO WHILE / END DO support involves the following enhancements to the
FORTRAN 77 standard:
1. Optional statement label in the indexed (standard) DO
statement.
The syntax of the indexed DO statement is:
DO [s[,]] v=e1,e2[,e3]
where s is the label of the statement that terminates the
loop. If s is omitted, then the loop must be terminated by
an END DO statement as discussed below.
2. DO WHILE statement
The DO WHILE statement has the syntax:
DO [s[,]] WHILE (e)
FORTRA.DOC - FORTRAN-10 Version 7 Page 11
where s is the label of the statement that terminates the
loop. If s is omitted, then the loop must be terminated by
an END DO statement as discussed below.
e is a logical expression which is tested at the beginning of
each execution of the loop, including the first. If the
value of the expression is true, then the statements in the
body of the loop are executed; if the expression is false,
control transfers to the statement following the loop.
3. END DO statement
The END DO statement has the form:
END DO
An END DO statement terminates the range of a DO or DO WHILE
statement. The END DO statement must be used to terminate a
DO block if the DO or DO WHILE statement does not contain a
statement label. It may also be used as a labeled terminal
statement if the DO or DO WHILE statement does contain a
terminal statement label.
2.5 Expressions On Output Lists
An output list item can be any of the following: A variable name, an
array element name, a character substring name, an array name, any
other expression.
(Note: FORTRAN-10 supports output list expressions that include
concatenation of operands of length *. This is an extension of the
ANSI standard that is also supported by VAX FORTRAN.)
2.6 Intrinsic And Generic Functions
FORTRAN version 7 supports all intrinsic and generic functions
described in section 15.10 of the FORTRAN-77 standard.
The following intrinsic functions are new in version 7:
1. Truncation for double precision - DINT
2. Nearest whole number - ANINT, DNINT
3. Nearest integer - NINT,IDNINT
FORTRA.DOC - FORTRAN-10 Version 7 Page 12
4. Positive difference for double precision - DDIM
5. Double precision product (of real arguments) - DPROD
6. Character functions:
1. Conversion from character to integer - ICHAR
2. Conversion from integer to character - CHAR
3. Length of character entity - LEN
4. Location of substring - INDEX
5. Character comparisons - LGE, LGT, LLE, LLT
The following generic function names have been added:
ACOS, AINT, ANINT, ASIN, COSH, CMPLX, DBLE, DIM, LOG, LOG10, MAX, MIN,
NINT, REAL, SINH, TAN, TANH.
The second arument to CMPLX is now optional.
The generic function name INT has been extended to support arguments
that are COMPLEX and INTEGER (in V6 it only handled REAL and DOUBLE
PRECISION).
2.7 Multi-record Internal Files
Internal files provide a means of transferring and converting data
from internal storage to internal storage. The functionality that
they provide is similar to that provided by ENCODE/DECODE. An
internal file is a character variable, character array element,
character array, or character substring. If a character array is
used, each array element is a record in the file. Reading and writing
records is accomplished by sequential access formatted input/output
statements.
2.8 FORTRAN-77 DO Loop Semantics
The behavior of DO loops in the FORTRAN-77 standard is incompatible
with that of the earlier FORTRAN-66 standard. Version 6 (and earlier
versions) handled DO loops in accordance with the old standard.
Version 7 supports the new standard.
There are two differences between FORTRAN-77 DO loops and FORTRAN-66
DO loops:
FORTRA.DOC - FORTRAN-10 Version 7 Page 13
1. In FORTRAN-77, zero trip loops are possible; previously all
loops executed at least once
2. In FORTRAN-77, the loop index is always available after loop
exit; previously it was not
Implied DO loops in I/O statements also follow the FORTRAN-77 DO loop
rules. Therefore, the implied DO loop index will be available after
loop exit. The only exception to this is when an error or end-of-file
condition occurs during execution of the I/O statement. In that case,
the value of the implied DO variable is indeterminate.
FORTRAN version 7 also provides support of the old-style DO loop
semantics. The compiletime switch /F66 (or /NOF77) may be used to
specify this. Specification of /F66 results in the following:
1. All DO loops execute at least once. (If /F66 is not
specified, zero-trip semantics are used.)
2. There is no guarantee that the loop index will be available
after loop exit. (If /F66 is not specified, the loop index
will always be available after loop exit.)
(Note that the /F66 switch also affects the behavior of the EXTERNAL
statement.)
2.9 Assumed-Size Array Declarators
In an array declarator for a dummy array, the upper bound of the last
dimension may be specified by an *. Note that an assumed-size array
cannot be used as an I/O list element, or in any other context where
the length of the array must be known. This is in accordance with the
FORTRAN-77 standard.
2.10 ASSIGN Of FORMAT Statement Numbers
Scalar variables that have been ASSIGNed FORMAT statement numbers can
be used as format specifiers in I/O statements.
FORTRAN version 7 does not allow the assignment of FORMAT statement
numbers to subscripted variables; nor does it allow the use of a
subscripted variable as a format specifier. This restriction is in
the FORTRAN-77 standard.
FORTRA.DOC - FORTRAN-10 Version 7 Page 14
2.11 INTRINSIC And EXTERNAL Statements
The INTRINSIC statement is used to identify a symbolic name as
representing an intrinsic function. It permits the name to be used as
an actual argument.
In version 7, if the name of an intrinsic function appears in an
EXTERNAL statement, that name is subsequently treated as the name of a
user defined function. This is accordance with the FORTRAN-77
standard. (See section 8.7 of the standard.) This is incompatible
with previous versions of FORTRAN-10
If the /F66 switch is specified at compiletime, then if an intrinsic
function name appears in an EXTERNAL statement (and is not preceded by
* or &) that name will still refer to the intrinsic function.
2.12 SAVE Statement
A SAVE statement is used to retain the definition of a variable, an
array, or a named common block after the execution of a RETURN or END
statement.
In fact, FORTRAN-10 usually retains the definitions of all variables,
arrays, and common blocks. The only exception to this is in the case
of overlays.
In version 7, if a FORTRAN overlay contains any local variables that
are SAVEd, all writable storage in that overlay will be preserved. If
a named COMMON block is SAVEd, the overlay that contains that common
block will be preserved. Blank COMMON is always preserved.
LINK support of writable overlays is required for support of this
feature. This is provided by LINK version 5.1.
2.13 Support Of Null Argument Lists For Functions
In version 6 it was possible to define an external function with a
null argument list, but there was no way to reference such a function.
Furthermore, the form of the declaration was incompatible with the
FORTRAN-77 standard. In version 6 there was no way to declare a
statement function that had a null argument list.
In version 7 we support the following:
1. The form:
fun()
FORTRA.DOC - FORTRAN-10 Version 7 Page 15
may be used to reference a function with a null argument
list.
2. Statement functions with null argument lists may be declared.
A null, parenthesized argument list appears in the
declaration.
3. Parentheses may optionally be used for null argument lists in
FUNCTION statements.
The standard requires the parentheses. Version 6 supports
FUNCTION statements with null argument lists, but does not
allow parentheses. Version 7 supports both forms.
2.14 FORTRAN-77 Statement Ordering Rules
Version 7 has relaxed some of the statement ordering restrictions that
were imposed by previous versions. Version 7 supports the statement
ordering rules specified by section 3.6 of the full language
FORTRAN-77 standard.
In addition, DATA statements, PARAMETER statements, and type
declarations may appear in any order as long as there are no
interdependencies. This extension to the standard is also supported
by VAX FORTRAN.
2.15 Minor Syntax Extensions
Version 7 includes the following minor extensions for compatibility
with the FORTRAN-77 standard:
1. The comma is optional in the following: DATA statements,
COMMON statements, assigned GOTO, after statement number in
DO statements.
2. Parentheses may optionally be used for null argument lists in
SUBROUTINE and CALL statements.
Null subroutine argument lists were supported in version 6,
but parentheses were not allowed.
3. Statement numbers are legal on non-executable statements.
These statement numbers are ignored and are not available at
execution time.
FORTRA.DOC - FORTRAN-10 Version 7 Page 16
4. Exponentiation to an integer power is allowed in the
subscript expressions in data statements.
2.16 Constant Expressions
FORTRAN Version 7 supports constant expressions in the following
constructs: Array declarators, PARAMETER statements, implied-DO lists
in DATA statements.
2.17 FORTRAN-77 PARAMETER Statements
FORTRAN version 7 supports PARAMETER statements in accordance with the
FORTRAN-77 standard, with the exception that we do NOT support
compiletime expressions that involve multiplication, division, or
exponentiation of COMPLEX data.
FORTRAN version 7 PARAMETER statement support differs from FORTRAN
Version 6 in the following ways:
1. In FORTRAN-77, the data type of a PARAMETER is determined by
its symbolic name. This is incompatible with the
implementation of PARAMETER that was supported under version
6. Under version 6 the type was determined by the type of
the constant.
2. In FORTRAN-77, the list of parameters must be enclosed in
parentheses. Under version 6, the list of parameters is
never enclosed in parenthees.
3. In FORTRAN-77, the parameter may be set to a constant
expression. FORTRAN version 6 only support simple constants.
If the list of parameters is not enclosed in parentheses, the compiler
assumes that it is not a FORTRAN-77 PARAMETER statement. In this
case, it handles it compatibly with FORTRAN version 6, by using the
data type of the constant rather than that of the symbolic name.
(Note: If the list of parameters is not enclosed in parentheses, and
if the data type of the constant and the symbolic name differ, the
compiler will issue a warning message. This warning message is
suppressed when the user compiles with the /F66 switch.)
In version 7 (as in FORTRAN-77), PARAMETER statements may precede type
declaration statements except for those statements that specify the
type of the parameter.
FORTRA.DOC - FORTRAN-10 Version 7 Page 17
FORTRAN-77 PARAMETER statements are implemented as specified by
section 8.6 of the FORTRAN-77 standard, with the exception that the
compiletime expression cannot contain multiplication, division, or
exponentiation of COMPLEX values.
2.18 Expressions As Specifier Values In OPEN And I/O Statements
Version 7 supports expressions as the values of specifiers in OPEN
statements and control information lists.
Integer expressions are supported as external unit identifiers, record
numbers, and the values of all numeric OPEN statement specifiers.
Character expressions are supported as format identifiers and as the
values of all character OPEN statement specifiers.
2.19 Type Checking For Function And Subroutine Arguments
LINK version 5.1 provides a limited type checking capability. FORTRAN
version 7 will generate REL file blocks to take advantage of this
capability.
The primary reason for implementing type checking in FORTRAN version 7
and LINK version 5.1 is to provide support for character constants
which are passed as actual arguments that correspond to numeric dummy
arguments. This feature worked in FORTRAN version 6 because the
argument passing mechanism for quoted strings was by the word address
of the string. This argument passing mechanism was changed in FORTRAN
version 7 in order to support FORTRAN 77 character data; the new
argument passing mechanism involves passing the address of a
descriptor for the string. If the formal is numeric, then the pointer
to the descriptor will be changed to a pointer to the string at
LINKtime. Note that this feature only works when both caller and
callee have been compiled with FORTRAN version 7. If a subprogram was
compiled with an earlier version of FORTRAN, it will not contain the
REL file blocks that LINK uses to recognize that a fix-up must be
performed on the character argument.
FORTRAN version 7 will also support a new option to the DEBUG switch.
This option will have the form /DEBUG:ARGUMENTS. If this option is
specified, FORTRAN will generate REL file blocks that specify that
illegal argument type mismatches should result in non-fatal error
messages at load time.
With FORTRAN version 7 and LINK version 5.1, the following type
checking will be performed:
FORTRA.DOC - FORTRAN-10 Version 7 Page 18
1. If either the caller or the callee does not have LINK
argument descriptor blocks, no type checking will be
performed. Therefore, no type checking will be performed on
calls the involve old REL files.
2. If an actual argument is "passed by descriptor" and the
corresponding formal is "passed by address" then LINK will
transform the actual argument into a "passed by address"
argument in some cases. This argument fixup was implemented
in LINK in order to provide FORTRAN support of character
constants passed as actuals to numeric dummies. Before doing
the fixup, LINK checks that all of the following conditions
are true; if any is not, it gives a non-fatal error message
and does not do the fixup:
- The argument must be a constant.
- The byte pointer word in the descriptor in the user's
image must be word aligned.
- The string must be in the same section as the argument
block.
If all of the above are true, LINK puts the word address of
the string into the argument block in place of the address of
the descriptor for the string. LINK also changes the
argument type code in the argument block from 15 (string
descriptor) to 17 (ASCIZ string).
The fixup of character constants passed as actuals to numeric
formals will be done whether or not the user specified
/DEBUG:ARGUMENTS when compiling his FORTRAN program.
3. If the user specified /DEBUG:ARGUMENTS when compiling his
FORTRAN program, non-fatal error messages will be issued at
LINK time for the following cases:
1. If the argument counts in the caller and callee disagree.
2. If a subroutine is called as a function or vice versa
3. If an actual argument is a constant and the value of the
corresponding dummy argument is potentially modified by
the subprogram. (Note that if the dummy argument is
passed as an actual to another subprogram, it is
potentially modified.)
4. If the length of a character actual is less than that of
the corresponding character dummy. This is only checked
if the length of the actual is known at compile time.
5. If the following associations of actual argument data
types with dummy argument data types occur:
FORTRA.DOC - FORTRAN-10 Version 7 Page 19
Actual Associated Dummy
Argument Type Argument Type
------------ -------------
Alt return label Logical
Alt return label Integer
Alt return label Real
Alt return label Double Precision
Alt return label G-floating
Alt return label Complex
Alt return label Character
Logical Alt return label
Logical Integer
Logical Real
Logical Double Precision
Logical G-floating
Logical Complex
Logical Character
Integer Alt return label
Integer Logical
Integer Real
Integer Double Precision
Integer G-floating
Integer Complex
Integer Character
Real Alt return label
Real Logical
Real Integer
Real Double precision
Real G-floating
Real Complex
Real Character
Double Precision Alt return label
Double Precision Logical
Double Precision Integer
Double Precision Real
Double Precision G-floating
Double Precision Complex
Double Precision Character
G-Floating Alt return label
G-Floating Logical
G-Floating Integer
G-Floating Real
G-Floating Double Precision
G-Floating Complex
G-Floating Character
Complex Alt return label
Complex Logical
FORTRA.DOC - FORTRAN-10 Version 7 Page 20
Complex Integer
Complex Real
Complex Double Precision
Complex G-floating
Complex Character
Character Alt return label
Character Logical
Character Integer
Character Real
Character Double Precision
Character G-floating
Character Complex
Octal Alt return label
Octal Double Precision
Octal G-floating
Octal Complex
Octal Character
Hollerith Character
Double Octal Alt return label
Double Octal Logical
Double Octal Integer
Double Octal Real
Double Octal Character
Note that in FORTRAN programs, all arguments that have
data type "character" will be "pass by descriptor", while
all other argument types will be "pass by address".
Therefore if the actual argument is a character constant
and the dummy is any other type, the FORTRAN special case
action described in item 2 above will be performed.
All combinations of data type (e.g. octal, double octal)
not listed in the above table are ignored.
2.20 New Functionality In ERRSET
The functionality of ERRSET has been enhanced in version 7. Version 7
provides ERRSET trapping for additional classes of errors in addition
to those trapping in version 6. Also, in version 7, the user can
write his own fixup routines for arithmetic exceptions.
The calling sequence for ERRSET is:
Call:
CALL ERRSET (N)
FORTRA.DOC - FORTRAN-10 Version 7 Page 21
or CALL ERRSET (N, I)
or CALL ERRSET (N, I, SUBR)
where N = max number of error messages to type
I = which error this call applies to. One of:
-1 any of the following
0 integer overflow
1 integer divide check
2 input integer overflow
3 input floating overflow
4 floating overflow
5 floating divide check
6 floating underflow
7 input floating underflow
8 library routine error
9 output field width too small
21 FORLIB warnings
22 non-standard usage warnings
( These warnings are described in
section 2.22 below.)
23 Array and substring bounds violations
If I is not specified, -1 is assumed
SUBR = routine to call on the trap
The effect is as if
CALL SUBR (I, IPC, N2, ITYPE, UNFIXED, FIXED)
were placed in the program just after the instruction causing
the trap.
I = error number of trap, same as above
IPC = PC of trap instruction for
error numbers 0, 1, 4, 5, and 6
( Note: Error 0 - integer overflow - can
occur in a jump instruction such as AOJA;
in that case, the value of IPC will
be incorrect. This is due to a permanent
restriction in the hardware. )
Otherwise, PC of FOROTS call
N2 = 2nd error number (reserved for DEC)
ITYPE = data type of value
UNFIXED = value returned by the processor
FIXED = value after fixup
SUBR can change this value.
If SUBR is not specified, no routine is called on the APR trap.
FORTRA.DOC - FORTRAN-10 Version 7 Page 22
2.21 Utility Subroutine To Get A Free Unit Number
FORTRAN V7 provides an additional FORTRAN-supplied subroutine which
can be used to get an unused unit number. The routine FFUNIT ("First
free unit") is called by:
CALL FFUNIT(IUNIT)
where IUNIT is an integer variable which will be set by FFUNIT to the
first available unit number.
If there are no available unit numbers, IUNIT is set to -1.
2.22 Additional Runtime Warnings
FOROTS version 7 issues warning messages for a number of non-standard
usages for which the results may be unexpected. These warnings may be
disabled by using the ERRSET function with error number 22.
FOROTS version 7 issues warnings for the following:
1. Data type mismatches between format specifiers and
corresponding I/O list elements. ( Bad data will be stored
in the I/O list elements. )
2. Writing beyond the end of a fixed length record. (The data
is truncated.)
3. Input to Hollerith fields of FORMAT statements. (This works
in V7, but we may de-support it in the future.)
3.0 RESTRICTIONS
3.1 FORTRAN-77 Features Not Supported
FORTRAN-10/20 version 7 supports the subset level FORTRAN-77 standard.
It also supports most of the features in the the full level FORTRAN-77
standard.
The following features from the full level standard are not supported
by FORTRAN version 7:
1. The INQUIRE statement is not supported by version 7.
2. In version 7, comment lines and blank lines may not appear
between an initial line and its first continuation line, nor
may they appear between two continuation lines.
FORTRA.DOC - FORTRAN-10 Version 7 Page 23
3. In version 7, the compiletime expression in a PARAMETER
statement cannot contain multiplication, division, or
exponentiation of COMPLEX data.
3.2 Other Restrictions
Version 7 will not support global optimization of subprograms that
contain character data. If the user specifies the /OPTIMIZE switch
for such a subprogram the warning diagnostic "/OPT ignored" will be
issued.
4.0 KNOWN BUGS AND DEFICIENCIES
4.1 Execute-Only
FORTRAN and FOROTS do not work when protected EXECUTE-ONLY. This
problem will be fixed in the first autopatch of version 7.
o Patch to make the compiler execute-only
If you want to protect the compiler execute-only, you can do
so if you install the following DDT patch to FORTRA.EXE:
443/ JSP 10,400031 JSP 10,143
454/ JRST 400062 JRST 174
455/ JSP 10,400031 JSP 10,143
463/ JRST 400062 JRST 174
If you prefer to make a source patch to the compiler, the
source patch is as follows. Note that you should save the
original source for autopatch.
File 1) V7AC:MAIN.BLI[4,504] created: 1849 12-Jan-1983
File 2) DSK:MAIN.BLI[4,714] created: 1141 03-Feb-1983
1)1 GLOBAL BIND MAINV = 7^24 + 0^18 + #1710; ! Version Date: 5-Jan-83
1)
****
2)1 GLOBAL BIND MAINV = 7^24 + 0^18 + #1725; ! Version Date: 3-Feb-83
2)
**************
1)1 ***** End Revision History *****
****
2)1 ***** End V7 Development *****
2)
2) 1725 TFV 3-Feb-83
FORTRA.DOC - FORTRAN-10 Version 7 Page 24
2) Add FORLINK linkage to statistics routines PHASBEGIN, PHASEND,
2) and RNTIME. This cures the execute only problem on the -10.
2)
2) ***** End Revision History *****
**************
1)2 ROUTINE PHASBEGIN=
1) BEGIN
****
2)2 ROUTINE FORLINK PHASBEGIN= ![1725] Use FORLINK linkage
2) BEGIN
**************
1)3 ROUTINE PHASEND=
1) BEGIN
****
2)3 ROUTINE FORLINK PHASEND= ![1725] Use FORLINK linkage
2) BEGIN
**************
1)4 ROUTINE RNTIME=
1) BEGIN
****
2)4 ROUTINE FORLINK RNTIME= ![1725] Use FORLINK linkage
2) BEGIN
**************
o Patch to make FOROTS execute-only
If you want to protect FOROT7.EXE execute-only, you must
install edit 3266 to FORFMT.MAC. Note that you should save
the original source for autopatch.
File 1) V7AO:FORFMT.MAC[4,707] created: 1817 13-Jan-1983
File 2) DSK:FORFMT.MAC[4,704] created: 1030 11-Feb-1983
1)1 TV FORFMT FORMAT PROCESSOR,7(3250)
1)
****
2)1 TV FORFMT FORMAT PROCESSOR,7(3266)
2)
**************
1)1 ***** End Revision History *****
****
2)1 ***** End V7 Development *****
2)
2) 3266 JLC 11-Feb-83
2) Allows FOROTS to be protected execute-only on TOPS-10.
2)
2) ***** End Revision History *****
**************
1)8 %IFSET: SKIPA T1,[JRST DATIN] ;SETUP FOR INPUT
1) %OFSET: MOVE T1,[JRST DATOUT] ;SETUP FOR OUTPUT
1) MOVEM T1,DATENT
1) MOVEM P,OTHERP ;SAVE THE CURRENT PDP
****
2)8 %IFSET: XMOVEI T1,DATIN ;[3266] SETUP FOR INPUT
2) JRST COMSET ;[3266]
2) %OFSET: XMOVEI T1,DATOUT ;[3266] SETUP FOR OUTPUT
FORTRA.DOC - FORTRAN-10 Version 7 Page 25
2) COMSET: MOVEM T1,DATENT ;[3266]
2) MOVEM P,OTHERP ;SAVE THE CURRENT PDP
**************
1)21 PUSHJ P,DATENT ;READ OR WRITE A RECORD
1) MOVE T1,ENC.AD ;GET ENCODED BLOCK ADDR
****
;**;[3266] @ EXELP + 6L, change 1 line, JLC, 11-Feb-83
2)21 PUSHJ P,@DATENT ;[3266] READ OR WRITE A RECORD
2) MOVE T1,ENC.AD ;GET ENCODED BLOCK ADDR
**************
1)23 DATLP1: PUSHJ P,DATENT ;GO DO INPUT OR OUTPUT
1) SKIPN T1,IO.INC ;ADD OFFSET TO I/O ADDR
****
2)23 DATLP1: PUSHJ P,@DATENT ;[3266] GO DO INPUT OR OUTPUT
2) SKIPN T1,IO.INC ;ADD OFFSET TO I/O ADDR
**************
1)23 NODATA: PUSHJ P,DATENT ;DO INPUT OR OUTPUT
1) MOVEI T1,2 ;ON TO NEXT FORMAT
****
2)23 NODATA: PUSHJ P,@DATENT ;[3266] DO INPUT OR OUTPUT
2) MOVEI T1,2 ;ON TO NEXT FORMAT
**************
4.2 Mandatory Patches For Related Products
None
4.3 Known Deficiencies In Related Products
1. Do NOT install DBMS edit 10500.
This is an edit to the module DBGATE in DBSFOR.REL. It was
required in order to build DBMS with FORTRAN Version 6 and
was described in the FORTRAN Version 6 Installation Guide.
This patch to DBGATE is also done by the FORTRAN version 6
autopatch tape 5; therefore you should be careful to use a
version of DBGATE that has not been autopatched with this
tape.
2. FOROTS version 7 cannot handle magtapes with blocksize larger
than 4096 words per block because of a bug in versions 7.01
and 7.01A of the TOPS-10 operating system. TOPS-10 MCO 10563
fixes this problem. This will be included in the next
autopatch of the TOPS-10 operating system.
FORTRA.DOC - FORTRAN-10 Version 7 Page 26
5.0 COMPATIBILITY WITH PREVIOUS VERSIONS
5.1 Incompatibilities Between Versions 6 And 7
o Changes in the behavior of OPEN statements
1. OPEN statements with no FORM or MODE specifier behave
differently in version 7 than they did in version 6.
This change only affects programs that do file
positioning operations (BACKSPACE or SKIPRECORD) of files
opened with no FORM or MODE specifier.
If an OPEN statement does not specify FORM or MODE,
version 6 allows the FORM= attribute to be determined by
the first IO statement executed. Version 7 always
determines whether a file is FORMATTED or UNFORMATTED
when an OPEN statement is executed. If the OPEN
statement does not include a FORM= or a MODE= specifier,
it defaults in accordance with the FORTRAN-77 standard.
The default is:
o If the OPEN statement specifies ACCESS='DIRECT' or
ACCESS='RANDOM' or ACCESS='RANDIN', then UNFORMATTED
(BINARY)
o Otherwise, FORMATTED (ASCII)
In order to minimize the impact of this incompatibility,
FORTRAN version 7 will support formatted I/O to
unformatted files and unformatted I/O to formatted files.
After such an I/O operation has occurred to a file, that
file will be considered to be a "mixed form" file and all
file positioning operations ( BACKSPACE and SKIPRECORD )
will be illegal.
For example, version 7 will issue a fatal error at
runtime for the following:
OPEN(UNIT=1, FILE='FOO') ! Formatted OPEN
READ (UNIT=1) X ! Unformattted READ; this makes the
! the file have "mixed form"
SKIPRECORD (UNIT=1) ! It is illegal to do SKIPRECORD
! to a mixed form file
The user can fix his program by modifying the OPEN
statement to include the FORM specifier.
This functionality is NOT affected by the /F66 switch.
FORTRA.DOC - FORTRAN-10 Version 7 Page 27
2. The behavior of OPEN on a connected unit has been changed
to satisfy the requirements of the FORTRAN-77 standard.
If the OPEN statement contains a "STATUS=OLD" specifier,
then its behavior depends on whether the file specified
by the OPEN is the same file that is currently connected
to the unit. If it is a different file, the connected
file is closed and the new file is opened. If it is the
same file, the file is not closed, and the file pointer
is not moved. This is a requirement of the FORTRAN-77
standard. This behavior is not affected by the /F66
switch.
Version 6 did not check for the case where the second
OPEN refers to the same file as the first OPEN. It
always closes the file and reopens it.
3. In FORTRAN version 6, if the RECORDSIZE specification in
an OPEN statement for a BINARY file conflicted with the
record size actually found in the file, no error was
indicated. In version 7 this is a fatal error. In
addition, if a type 2, or continuation-type LSCW (logical
segment control word) is found in a file for which a
RECORDSIZE specification has been given, this too will
yield a fatal error, as files written with a RECORDSIZE
specification should not have type 2 LSCWs. This error
was also not detected by version 6.
o Changes in the behavior of I/O statements
1. Changes in formatted I/O
The treatment of special characters in formatted files
has been changed in order to enable users to read and
write character variables that contain these characters,
and to improve the performance of formatted I/O.
o Formatted input for files OPENed with fixed length
records
In version 7, when formatted input is done from a
file with fixed length records, all characters are
read exactly as they appear in the file. The number
of characters read is always equal to the RECORDSIZE
(RECL) specified in the OPEN statement. Thus, in
version 7, A-format input from a fixed length record
will read NUL, CR, LF, FF, or VT if those characters
are within the record length specified in the OPEN
statement.
This differs from version 6 in two ways:
FORTRA.DOC - FORTRAN-10 Version 7 Page 28
1. Version 6 always terminates record input when it
encounters CR, LF, FF, or VT, regardless of the
RECORDSIZE.
2. Version 6 ignores NULs.
o Formatted input for files OPENed with variable length
records
In version 7, A-format input from a variable length
record reads all NULs that are interior to the
record. Version 6 ignores them. (Note: Since line
terminators terminate the record, version 7 does not
read them. )
As in version 6, leading NULs in variable length
records are still ignored. This is necessary in
order to support variable length input from a file
that was written with fixed length output. This is
because fixed length records are word-aligned and
inter-record gaps contain NULs.
o A-format output (for both fixed and variable length
records)
In version 7, A-format output writes out all
characters exactly as they appear in the
corresponding I/O list variables. In version 6,
blanks were substituted for NULs. This is no longer
done.
o Padding of fixed length records
In version 7, fixed length records in formatted files
are padded with blanks. In previous versions, they
were padded with NULs. (The user can specify the pad
character desired by using PADCHAR= in his OPEN
statement. Space is the default.)
2. Change in multi-record ENCODE/DECODE
In previous versions of FORTRAN, the character count in
ENCODE and DECODE statements was interpreted as a total
string count, and a "/" format or indefinite repeat,
which for normal files causes a new record to be read or
written, aligned the record pointer to the next word
boundary. In V7, the count specified is interpreted as
an internal record size in characters, and "/" format or
indefinite repeat cause the record pointer to be advanced
to the next internal record.
FORTRA.DOC - FORTRAN-10 Version 7 Page 29
o New Runtime Warning Messages
1. In FORTRAN V6 and all previous versions, Hollerith and
quoted string input to FORMAT statements was allowed and
not flagged with any warning. FORTRAN V7 will give a
warning message, as it is illegal according to the ANSI
Standard, and some future version of the compiler may
compile or pool FORMATs, thus disallowing Hollerith and
quoted string input altogether.
This warning may be disabled with the ERRSET function.
2. In FORTRAN V6, data type mismatches between format
specifiers and I/O list items were ignored. FOROTS
Version 7 issues a warning for this. This warning may be
disabled with the ERRSET function.
o Changes that affect assembly language subprograms
1. In FORTRAN V6, if a quoted string is an actual argument
to a MACRO subprogram, the argument block that is pointed
to by AC 16 contains the address of the ASCIZ string. In
FORTRAN V7, the argument block contains the address of a
two-word descriptor block, in which the first word is a
byte pointer to the string (which will in fact be word
aligned and ASCIZ) and the second word is the length of
the string.
2. In FORTRAN V6 and previous versions, calls to ALCHN.,
DECHN., FUNCT., ALCOR., and DECOR. could have indexed
arguments and ACs specified as arguments. For
performance reasons, this capability has been removed.
Users were warned about this upcoming change in the V6
manual.
3. In versions 6 and before (all the way back to LIB40), an
instruction followed by JFCL works as follows: If the
instruction is "JFCL 0,addr", do not type an error
message, fix up the result of the instruction, and jump
to addr. If the instruction is "JOV addr", do type an
error message, do fix up the result, and jump to addr.
The behavior of JOV is different under version 7: do NOT
type an error message, do NOT fix up the result, and jump
to addr. This makes JOV behave the same under the trap
handler as it does under the hardware. "JFCL 0," will
not change.
o Changes in the behavior of declarations
1. In version 7, if the name of an intrinsic function is
specified in an EXTERNAL statement that name is treated
as the name of a user-defined function. In version 6,
FORTRA.DOC - FORTRAN-10 Version 7 Page 30
the function name is treated as the name of an intrinsic
function. (In version 6, this kind of EXTERNAL statement
was used to allow the intrinsic function name to be
passed as a subprogram argument.)
If the /F66 switch is specified, EXTERNAL statements are
processed compatibly with version 6.
2. FORTRAN version 7 supports the FORTRAN-77 syntax and
semantics for PARAMETER statements. This is different
from the version 6 implementation; however, the old
syntax/semantics also works under version 7 except that a
warning is issued when it is used. (This message is
suppressed in the /F66 compilation switch is specified.)
In particular:
o In version 7, the list of parameters must be enclosed
in parentheses. If it is not enclosed in
parentheses, a warning message is issued.
o In version 7, the data type of the PARAMETER is
determined by its symbolic name, unless the
parentheses around the parameter list have been
omitted. If the parentheses have been omitted, then
the data type is determined in the same way as it was
in version 6, which is by the type of the constant.
3. In Version 7, a type declaration of a generic function is
ignored. This is in accordance with the FORTRAN-77
standard. This functionality is not affected by the /F66
switch.
The interpretation of type declarations of generic
functions by FORTRAN Version 6 was not consistent. For
example, if SIN was declared to be double precision it
was subsequently treated as a user function name. The
compiler did no type checking on its arguments and a call
to "SIN" (not "SIN.") was generated. However, if SIN was
declared to be real it was treated as a non-generic
library function name. The compiler checked that its
argument in all subsequent calls was real, and a call to
"SIN." was generated.
o DO loops
Under FORTRAN version 7, the iteration count for a DO loop
may be zero. Under version 6, all loops were executed at
least once. The /F66 switch may be specified at compiletime
to cause DO loops to be executed compatibly with version 6.
o FORDDT
1. In version 7 of FORDDT, /C is used for CHARACTER mode.
In version 6, /C was used for COMPLEX mode. In version
7, /X is used for COMPLEX mode.
FORTRA.DOC - FORTRAN-10 Version 7 Page 31
o Old version 5A constructs that generate warnings under
version 6
FORTRAN-10/20 version 6 issues warning messages for a number
of constructs that are fatal in version 7.
1. In FORTRAN version 5A the character '/' could be used
instead of ':' to separate the upper and lower bounds in
a dimension declarator. In version 6, use of '/' causes
a non-fatal warning message to be issued. Version 7
supports constant expressions in array declarations, and
therefore in version 7 the '/' will be treated as a
division operator.
2. In version 5A the character '#' could be used to separate
the unit and record numbers in random I/O statements. In
version 6, this use of '#' causes a non-fatal warning
message to be issued. In version 7 the '#' is parsed as
a .NE. operator in the expression for the unit number;
this then results in the fatal error "Relational
expression illegal as UNIT specifier".
3. In version 5A, a subscripted reference could be used in
an assigned GOTO statement. This feature was never
documented. In version 6, use of this feature causes a
non-fatal warning message to be issued. In version 7
this will be illegal. (This is because in FORTRAN-77 the
comma in an assigned GOTO is optional. Without the
comma, use of a subscripted reference in an assigned GOTO
is ambiguous.)
o Other
1. The error numbers associated with certain fatal runtime
errors have changed, and some errors which were
previously warnings are now fatal. In particular, all
errors whose first error number was 799 in version 6 now
have 97 as the first error number; all errors whose
first number was 899 now have 98. All errors in these
two error classes are now fatal.
2. In FORTRAN V6, quoted literals could be passed as
arguments to numeric formals. In FORTRAN V7, those
quoted literals are treated as character constants, and
are passed by descriptor. In most cases, V7 will
continue to support passing these constants to numeric
formals. A LINKtime fixup will be done to support this.
Note that this feature only works when both caller and
callee have been compiled with FORTRAN version 7. If a
version 7 program calls a subprogram that was compiled
with an earlier version of FORTRAN, the subprogram will
not contain the REL file blocks that LINK uses to
recognize that a fix-up must be performed on the
character argument.
FORTRA.DOC - FORTRAN-10 Version 7 Page 32
If the name of the function being called is a dummy
argument, then the LINKtime fixup cannot be done.
5.2 Compiler "FORTRAN-77" Switch
There will be a compiletime switch which will specify whether certain
constructs should be interpreted compatibly with the FORTRAN-77
standard or compatibly with previous versions of FORTRAN. The two
values of this switch are:
/F77 - FORTRAN-77 compatible
/F66 (or /NOF77) - Version 6 compatible
The default value of this switch is /F77.
The following constructs will behave differently depending on the
value of this switch:
1. DO loops
When /F77 is specified, 0-trip loops are possible. When /F66
is specified, all DO loops execute at least once.
When /F77 is specified, the loop index will always be
available after loop exit. When /F66 is specified, there is
no guarantee that the loop index will be available after loop
exit.
2. Use of intrinsic function names in EXTERNAL statements
If the name of an instrinsic function appears in an EXTERNAL
statement, but is not preceded by * or &:
1. When /F77 is specified, the name is subsequently treated
as the name of a user defined function.
2. When /F66 is specified, the name still refers to the
intrinsic function.
3. PARAMETER statements
The warning message for non-parenthesized PARAMETER
statements is suppressed when /F66 is specified.
FORTRA.DOC - FORTRAN-10 Version 7 Page 33
5.3 FOROTS Interface
Programs that are compiled with FORTRAN version 7 must be executed
with FOROTS version 7 and FORLIB version 7. If you inadvertantly load
version 6 FORLIB/FOROTS with a .REL file that was generated by the
version 7 compiler the you may get the runtime error:
?Data in IO list but not in format
Other errors may also occur.
In most cases, programs that were compiled with FORTRAN version 6 will
work with FORLIB/FOROTS version 7; however, we cannot guarantee that
this is always true. In you encounter a problem running a V6 .REL
file with the new FORLIIB/FOROTS, you must recompile.
MACRO programs that search FORPRM.UNV should be reassmebled with the
version 7 FORPRM in order to use version 7 FORLIB and FOROTS.
5.4 Subprogram Calling Conventions
In most cases subprograms compiled under version 6 can be called by
subprograms compiled under version 7 and vice-versa. The exceptions
to this are:
1. If a subprogram is compiled with version 7, it cannot pass
character or quoted string arguments to a subprogram compiled
with an earlier version of FORTRAN.
The LINKtime fixup of actual arguments that are quoted
strings only works when both caller and callee have been
compiled with FORTRAN version 7. If a subprogram was
compiled with an earlier version of FORTRAN, it will not
contain the REL file blocks that LINK uses to recognize that
a fix-up must be performed on the character argument.
2. If a version 7 subprogram has dummy arguments of type
character, it cannot be called by a subprogram compiled with
an earlier version of FORTRAN.
3. Subprograms that do I/O must be compiled with the same
version of FORTRAN. We cannot guarantee that FOROTS version
7 will always work when called by subprograms compiled with
the version 6 compiler.
FORTRA.DOC - FORTRAN-10 Version 7 Page 34
5.5 Using DBMS With FORTRAN V7
The FORTRAN/DBMS interface has been thoroughly reworked for FORTRAN
V7. Instead of the DBMS files being merged into FORLIB, they are
placed in a separate file called SYS:DBMSF.REL, which is searched by
LINK after FORLIB.REL if the program contains DBMS calls. Under
normal circumstances, a FORTRAN/DBMS program will use a combined
FOROTS/DBMS file called FDBOT7.EXE for its sharable high segment. If
the user desires to use /OTS:NONSHARE to load FOROTS with the program,
DBMS will also be loaded with the program. Additionally, there is a
new switch provided - %DBSNS - which, if it is defined before FORLIB
is searched, will load DBMS with the user's program regardless of
whether the sharable FOROTS is used.
Examples:
1. Using sharable FOROTS/DBMS (the default action):
.LOAD program
causes the user program to be loaded and uses the sharable FDBOT7.EXE
for calls to FOROTS and DBMS.
2. Using sharable FOROTS with DBMS loaded with the user's program:
.R LINK
*program/DEFINE:%DBSNS/G
causes the user program to be loaded with DBMS to use the sharable
FOROT7.EXE for calls to FOROTS.
3. Using nonsharable FOROTS and DBMS:
.R LINK
*program/OTS:NONSHARE/G
causes the user program to be loaded with FOROTS and DBMS.
5.6 Incompatibilities Between Versions 5A And 6
There were a number of changes introduced between versions 5A and 6 of
FORTRAN. If your site is converting directly from version 5A to
version 7, you should also be aware of the following:
1. The constructs '' and 0H which were legal in version 5A are
illegal in versions 6 and 7.
In version 5A, '' produced a string of 5 spaces. In versions
6 and 7, a string of one space (' ') should be used for this
purpose.
FORTRA.DOC - FORTRAN-10 Version 7 Page 35
In version 5A, 0H produced a word of nulls. In versions 6
and 7, a numeric 0 should be used to initialize a Hollerith
string with nulls. In version 7, CHAR(0) can be used to
represent the character NUL.
2. Numeric differences
There were a number of changes made between versions 5A and 6
to improve the numeric accuracy of results. These changes
can propagate through a calculation to produce larger
differences in final results.
o The mathematical functions in FORLIB were rewritten for
increased accuracy.
o The floating point input and output routines of FOROTS
were made more accurate, so that the numbers they produce
(both internally and externally) are more accurate (and,
therefore different) than those produced by version 5A.
o The maximum number of digits encoded by the floating
point output routine was increased in version 6 over that
encoded by version 5A. Thus printing with format, say,
E30.20 will produce different results with version 6 than
with version 5A: for both real and double precision
numbers the number of non-zero digits has increased to
20.
o Integer division by 0 now returns the largest possible
integer value having the sign of the numerator. In
version 5A, the value returned was the numerator.
3. In FORTRAN versions 6 and 7 the default interpretation of
blanks in formatted files is BLANK=NULL, unless no OPEN
statement has been executed. This is a requirement of the
FORTRAN-77 standard. It is incompatible with version 5A in
which blanks are always treated as 0's.
4. A string of consecutive linefeeds is treated as one linefeed
by 5A; version 6 treats them as multiple records.
5. The functions DFIX and IDFIX have been replaced by the
equivalent IDINT.
6. DIALOG input may be terminated by a carriage return. Altmode
invokes command recognition and so does not terminate dialog
input. If it is necessary to type extremely long dialog
responses, either type over the line boundary without hitting
carriage return, or use - (hyphen) at the end of the line as
a continuation character.
7. In versions 6 and 7, when a READ takes the ERR= branch, the
file is positioned after the erroneous record. A BACKSPACE
may be used to position the file before the erroneous record;
FORTRA.DOC - FORTRAN-10 Version 7 Page 36
otherwise the next READ will get the record after the one
with the error. This is different than version 5A, which
leaves the file positioned before the erroneous record.
If it is desired to reread the erroneous record and specify
an ERR= branch, you can now use REREAD with an ERR= branch.
8. The BLOCKSIZE parameter of the OPEN statement specifies tape
block size in bytes. In version 5A, the block size is
specified in words.
9. If RECORDSIZE is specified when opening a file, and the file
contains records that are not of the specified size, behavior
will differ between version 5A and version 6. In formatted
mode 5A reads until it finds a CRLF, 6 reads until the
specified number of characters have been read. In
unformatted mode 5A reads until the I/O list is satisfied
(and no farther), 6 reads until the specified number of words
have been read. Reading BINARY-mode files with the wrong
record size is fatal with both versions.
10. When reading list-directed I/O (* format), 5A reads '4* 5' as
'5,5,5,5'. 6 reads it as ',,,,5' (4 null items followed by a
5). Files written with 5A can contain repeat counts followed
by spaces, and so will be interpreted incorrectly by 6.
Files written by 6 can be read by 5A.
11. FOROTS V6 does not ignore nulls encountered in a DECODE
statement, whereas V5A did. Null is an illegal character in
many of the conversion routines.
12. The action of the CLOSE statement has changed in version 6,
such that specification of any file parameters (e.g. FILE=)
will cause a rename to be performed. Version 5A only did the
rename if it detected an actual change.
13. With version 6, recordsize is enforced for input files. The
actual record size used is the record size specified, plus 2
for the carriage return, line feed sequence, plus the padding
necessary to bring the record to a word boundary. Thus the
formula is ((RECSIZ+6)/5)*5. Records read using a RECORDSIZE
specification in the OPEN statement must have precisely this
calculated number of characters. FOROTS no longer stops upon
reading an end-of-line character, nor scans for them, in
advancing to the next record. However, a premature
end-of-line character within a record will terminate the
characters read into the FOROTS internal record buffer.
14. CALL ERRSET (n) now tells FOROTS to print out n warnings of
each type. ERRSET extensions... CALL ERRSET(n,i) to set a
limit for a specific error, CALL ERRSET(n,i,x) to call
subroutine x when the error occurs, after FOROTS error
actions have occurred.
FORTRA.DOC - FORTRAN-10 Version 7 Page 37
15. Sequential files written with version 6 without a record size
specified are not padded with nulls to a word boundary,
whereas version 5A padded to a word boundary to be compatible
with FORSE files. This means that files written sequentially
without a record size specified in the OPEN statement will,
in general, not be readable as random-access files.
16. With Version 5A, BLOCKSIZE for magtape files always meant
words, and the default was 128 words for TOPS-10 and 512
words for TOPS-20. With version 6, BLOCKSIZE is in bytes,
36-bit bytes for binary and image mode files, and 7-bit bytes
for formatted (ASCII) files. The default is still 128 words
for TOPS-10 (640 bytes for formatted files), but is 1000
bytes for TOPS-20 (i.e., different blocksizes for formatted
and unformatted files).
17. The definitions of X and T formats have changed from 66 to 77
for output statements. With -77, T and X do not actually
write blanks, they merely position the record pointer. Only
the actual data sets the record size, so that (A5,100X)
writes a 5-character record whereas (A5,100X,' ') writes a
106 character record.
18. With list-directed and namelist I/O, logical input data is
strictly defined as an optional period, followed by a T or F,
followed by any set of non-terminator characters. There will
no longer be any diagnostic messages given for such trashy
input.
19. The default field width for F (with no field width supplied)
for output is dependent on the type of the variable in
version 6, and expands for extremely large numbers, rather
than producing asterisks. In version 5A, the default field
width was invariant for data type.
20. As an extension to the ANSI standard, list-directed input
will accept any data type for any variable type, and do the
kind of conversion done by the compiler. This is also true
for data preceded with a repetition count (e.g. 2*5.6 input
to a real variable and an integer variable will provide 5.6
to the real variable and 5 to the integer variable). The
exception is that if complex data is specified for a double
precision variable, the DOUBLE PRECISION value of the real
part is transmitted to the double precision variable (the
compiler truncates it to a single precision value and sets
the 2nd word of the double precision variable to 0).
21. With list-directed and NAMELIST I/O, certain types of data
with no delimiters are interpreted differently by versions 5A
and 6. For instance, with version 5A, -1-1 are read
differently depending on whether the target variables are
integer or real: it reads two values of minus 1 for integer,
and -1E-1 for real. Version 6 always reads -1-1 as a single
value (-1E-1) and converts it appropriately.
FORTRA.DOC - FORTRAN-10 Version 7 Page 38
22. The ANSI standard specifies that random-access files have no
end. Version 6 obeys this rule, which is a change from
version 5A. Thus the END= branch will never be taken, and
should not be used, for random-access data transfer
statements.
23. For formatted data and binary unformatted data, the ERR=
branch will be taken for any unwritten record, whether in the
middle or at the end of the file. Since there is no way to
know whether data has been written in an image mode file, the
ERR= branch cannot be used for this purpose for image mode
files.
24. In version 6, dollar format is defined as suppressing the
carriage return and one line feed (if any) that would
normally appear if the next data transfer were a WRITE
statement, or the automatic <carriage return - line feed>
sequence that would normally be inserted by FOROTS if the
next data transfer were a READ statement. Version 5A did not
suppress the line feed for the succeeding WRITE.
25. There is no support in version 6 for KA10s or for
F40-generated code. Also, support for the FOROTS version 1
calling sequence has been removed.
6.0 INTERNAL CHANGES
Both FOROTS and the compiler have been substantially modified to
support the new features that are in this release. Complete revision
histories are contained in REVHST.MAC (for the compiler), FORHST.MAC
(for FOROTS), MTHHST.MAC (for the math library routines), and
FORDDT.MAC (for FORDDT).
[End of FORTRA.DOC]