[apt4ssx8] Re: APT4 documentation
- From: dave <dengvall@xxxxxxxxxxx>
- To: apt4ssx8@xxxxxxxxxxxxx
- Date: Sun, 3 Jun 2012 13:37:02 -0700
On Sun, 03 Jun 2012 15:40:32 -0400
"Kent A. Reed" <kentallanreed@xxxxxxxxx> wrote:
> On 6/3/2012 2:36 PM, Dave Caroline wrote:
> >> PS - Among its many wonderful options, the tool ftnchek has the
> >> ability to generate skeletal FORTRAN module descriptions in the
> >> form of html pages. I had been thinking these pages could then be
> >> amended to include functional descriptions, etc. Now I'm wondering
> >> if at some point we can mash together the output from f77_diagram
> >> and ftnchek.
> >>
> > Had a play with ftnchek
> > It has a bug with no-prune set when making a tree, seems to get lost
> > in the APT recursion
> > and when you look at the APT recursion one feels ill.
> > It created a 5gb output file!
>
> Interesting. I hadn't pushed the tool that direction and so hadn't
> run into that particular bug. It reminds me of...oh, well, never
> mind. It was a long time ago and I've buried the mistake.
>
> > I have created a ASECT0.FOR to replace the assembler initialization
> > and first CALL ACNTRL
> >
> > http://www.archivist.info/apt/aptos/apt360/orig_source/ftnchek/
> >
> > call tree
> > http://www.archivist.info/apt/aptos/apt360/orig_source/ftnchek/CallTree.html
> >
> > ftnchek output with errors,warnings etc it finds
> > http://www.archivist.info/apt/aptos/apt360/orig_source/ftnchek/calls.txt
>
> Looking good, Dave.
>
> Last week, as I looked through my own ftnchek outputs and referred
> back to the sources, I began to wonder just who wrote this APT360
> code. I'm supposing it's a combination of the work of IBM coders and
> some baseline APT4 code set from the CAM-I. In turn, the CAM-i code
> set was probably a conglomeration of original and edited routines
> from CAM-I members and contractors that evolved through various
> releases. The end result is truly ugly; not even close to the coding
> standards of major physics projects in the same time period that I
> crossed paths with. It's only redeeming feature is that it worked,
> mostly. I tip my hat to Brent for managing to transliterate it to C
> and get it to work again.
>
> > I cannot understand why they do part of the html creation in C
> > then use a shell script to add the html header and footer.
> > makes shoehorning in other stuff a little harder
>
> I confess I've done this kind of thing myself when I thought it was
> expedient but I don't usually impose the result on others.
>
> I turned to ftnchek because of its "semantic" checking capability,
> hoping in particular to get a handle on the (in)consistency of
> APT360's common block usage across its gazillion modules, and only
> discovered the html option when later I looked at the full list of
> options.
>
> > Dave Caroline
> >
>
> Regards,
> Kent
Since I don't have a coding reputation to uphold I'm going to ask dumb
questions.
a. is there a tool to scan the code base and make a global
common?
b. is there a max size for a common, either in number of
variables or its compiled size.
I ask this because of the passing comment on creating a #include for a
common.
I realize that one could cut and paste to make a massive file with all
the common blocks in it and then parse, sort and eliminate duplicates
but there must be an easier way.
Dave
Other related posts:
