This runs via IP as:
==========================================================================
program test;
type pnode = ^integer;
function myfun : pnode;
begin
new(myfun);
myfun:=nil;
end;
begin
end.
C:\TEST>parse test
Pascal parser vs. 1.13.05 Copyright (C) 2005 S. A. Moore
Processing: C:\TEST\test.pas
new(myfun);
^
*** C:\TEST\test.pas [9:14] 'myfun' must be a variable type ***
Errors this compilation: 1
No output file was generated
Function complete
==========================================================================
The function definitely isn't defined as a variable in it's body. The
only valid reference to it is the last statement, myfun := nil, which
is not, in fact, an assignment, but rather the establishment of a
function result (for myfun). Although logically treating the function
result as a "write only" variable is one implementation solution,
there are many others. In fact, reduction to a single register is a
reasonable optimization, and that, in and of itself implies that
promoting function results to variables in attempts to achieve
artificial orthogonality is unwise.
In any case, no, this does not comply with ISO 7185.
John Reagan wrote:
> Waldek Hebisch wrote:
>> steelman22 <steelman22@[EMAIL PROTECTED]
> wrote:
>>
>>> I have a function
>>>
>>> function myfun(....) : pnode {pnode is pointer}
>>> ..
>>> is it ok to do this thing:
>>>
>>> function myfun(....) : pnode
>>> begin
>>> ...
>>> new(myfun);
>>
>> ^^^^^
>> According to ANSI/ISO this is a recursive call.
>
> Not in my ANSI/ISO standard it isn't.
>
> The parameter for NEW isn't an expression but a variable-access.
>
> Variable-access is defined as
>
> variable-access = entire-variable | component-variable |
> identified-variable | buffer-variable.
>
> So you can't get to function-call.
>
>
> However, doing "new(myfun)" still isn't legal since function-idenitifer
> isn't part of variable-access. The standard says there must be at least
> one assignment-statement to the function-identifier in the
> closest-containing block.
>
> Any compiler that accepts "new(myfun)" is doing so as an exception. My
> compiler signals an error:
>
> $ pascal xxx
>
> new(p);
> .....^
> %PASCAL-E-IVUSEFNID, Illegal use of function identifier P
> at line number 7 in file HIYALL$:[REAGAN]XXX.PAS;1
> %PASCAL-E-ENDDIAGS, PASCAL completed with 1 diagnostic
>
>
>
|
