Re: CVF bug ?
From: B52B (B52B_at_pl.pl.pl)
Date: Thu, 07 Oct 2004 18:04:10 +0200
Jugoslav Dujic wrote:
> B52B wrote:
> | Steve Lionel wrote:
> || On Wed, 06 Oct 2004 11:42:03 +0200, B52B <B52B@pl.pl.pl> wrote:
> ||| There is a bug introduced in the latest version of CVF 6.6C3 patch.
> ||| program test
> ||| implicit none
> ||| real, dimension(:), pointer :: a_ptr => null(), b_ptr => null()
> ||| allocate (a_ptr(1:10))
> ||| a_ptr = 1.0
> ||| b_ptr => a(::2)
> || This program won't compile. I assume you meant to write a_ptr instead of a
> || in the last quoted line.
> | Yes of course. I was too tired when I wrote this mail. BTW
> | I The same bug exists for ordinary allocatable arrays.
> | program test
> | implicit none
> | real, allocatable :: x(:)
> | allocate (x(10))
> | x = 1.0
> | print *, size(x(::2))
> | end program test
> FWIW, I applied most (three?) of 6.6.C updates and I'm not sure which
> one was applied last [I *think* it's actual 6.6.C3 but a) I'm not positive
> b) I'm not sure if the order of application mattered]. Thus, my
> CVF gets all your examples (with appropriate syntax fixes) right.
> FWIW, it says:
OK. Let me summarize problems:
There were three 6.6C patches and HP withdrawn the first
and the second one. There is only one (6.6C3) official and available
on the HP site.
Since this is the last patch (according to Steve Lionel
or I missed something) and it breaks compiler support for the
very important part of Fortran 95 features (allocatable arrays
and pointer aliases), CVF users were left orphaned and
without obvious choice.
a) apply patch C3 and lost support for Fortran 95 certain features
b) stay at 6.6B and lost support for allocatable components, bug fixes
c) leave updates out and upgrade to the Intel 8.1 compiler
I think HP should release the next patch (C4?) which will remove
the serious bug (I insist) introduced by 6.6C3.
> D:\Users\Duja>df /what
> Compaq Visual Fortran Optimizing Compiler Version 6.6 (Update C)
> Copyright 2003 Compaq Computer Corp. All rights reserved.
> Compaq Visual Fortran 6.6-3780-47D3B
> C:\Program Files\Microsoft Visual Studio\DF98\BIN\decfor90.exe
> | so CVF 6.6 C3 is almost completely USELESS.
> I think it's too bold statement; myself, I would never have encountered
> such bug because I never omit leading strides in a triplet, and would
> certainly never omit both leading and ending (which appears to be only
> troublesome combination). What would you say about e.g. 6.6B which allowed
> pointer-association with non-TARGETs then?