HP Fortran v2.9 for HP-UX 11i v2 Release Note*5991-0697*

HP Fortran for HP-UX v2.9
Previous Fixes
Chapter 1 45
• The handling of A(:) versus A(numeric:numeric) was reevaluated.
• Wrong answers were occasionally received from an array assignment with dependence.
• Wrong answers were received at +O2.
• Calling a generic intrinsic with an INTEGER*2 argument now calls the INTEGER*2 version
of the intrinsic and if necessary, interprets other arguments that are constants as
INTEGER*2 constants.
• There was a syntax error when the imaginary part of a complex constant ended with a
space.
• Ucode-2 from int2 + iand(int2,int4) was incorrect.
•TheC$PAR directive is now allowed as comments in +Oopenmp mode.
• There were incorrect bounds of character string arrays.
• When using shift count equal to the number of bits in the integer value, rshift should
have returned “0” but did not.
• There was a compile-time assert.
PHSS_30520
• Wrong answers were received from pointer assignments at +O2 and higher.
• Wrong answers were received from FORALL containing loop-varying shapes at all
optimization levels.
• An error was incorrectly detected on this OpenMP example: The SINGLE directive should
be bound to the nested PARALLEL region and not conflict with the CRITICAL directive.
• There was a carry over problem from MODULE.
• +oversubscriptcommon was enhanced in order to extend the size of local arrays
equivalenced to arrays in a common block.
• Substring comparisons were enhanced to generate a simple length load of “1” rather than
generating the length computation of e=(length-start+1).
• The options FAILSAFESIG{BUS,SEGV}{HARD,SOFT} occasionally aborted.
• The f90 style of declaration -Idir+1 enabled f66 behavior for loops.
• +i8 and +U77 were causing the wrong libU77 etime library to be pulled in.
• Copyin/copyout should be used on Cray pointer arguments so that debug information
works properly.
• The message on promotion of Cray pointers to integers was clarified.