Re: RW on Working-Storage (was: J4 - presentation/discussion on

On Mar 14, 6:37 pm, Robert <n...@[EMAIL PROTECTED]
> wrote:
> On Thu, 13 Mar 2008 11:11:08 -0700 (PDT), Richard <rip...@[EMAIL PROTECTED]
>
wrote:
> >On Mar 14, 3:14 am, Robert <n...@[EMAIL PROTECTED]
> wrote:
> >> On Thu, 13 Mar 2008 06:28:30 GMT, "William M. Klein"
<wmkl...@[EMAIL PROTECTED]
> wrote:
>
> >> >I can certainly understand the confussion but calling something
"static data"
> >> >(and defining what that means does NOT mean that the storage is
allocated
> >> >statically (which was your statement).
>
> >> The problem with abstract sounding language is that peoples' eyes
glaze over and they
> >> forget the original question. Let's try to stay focussed on one
question -- plain vanilla
> >> working-storage. Of the three choices below, it is the one called
static. Forget automatic
> >> and initial, concentrate on what the standard says about static.
>
> >> >The section that you quoted from is
> >> >under
>
> >> >  "8.6.3 Automatic, initial, and static internal items"
>
> >> >The first paragraph explains what all of this is about, i.e.
>
> >> >"Each internal item has one of the three persistence attributes:
automatic,
> >> >initial, or static. The designation of automatic, initial, and
static items
> >> >relates to their persistence and the persistence of their contents
during the
> >> >execution of the run unit."
>
> >> >Notice inparticular that for "automatic" data, it says,
> >> >  "Their storage is allocated and set to initial state each time the
runtime
> >> >element containing them is activated."
>
> >> >For "static" items, the statement is quite different.  It says,
>
> >> >"They are allocated no later than immediately before initialization"
>
> >> >Which I think (to me) is NOT what you said, i.e.
>
> >> >" The standard says working-storage is allocated statically."
>
> >> Initialization means at program load time. Statically also means at
program load time. In
> >> plain English, the standard says 'working-storage is allocated at
program load time.'  I
>
> >NO, Robert, that is your _INTERPRETATION_ of what you think it means.
>
> No, it's what the standard SAYS. "static items ..  are allocated no
later than
> immediately before initialization .."
>
> >> said "working-storage is allocated at program load time." There is no
confusion.
>
> >NO, Robert, that is simply the same _INTERPRETATION_ of what you think
> >it means.
>
> >When you make a mistake you merely repeat it until everyone gives up.
> >(and then later claim that you were right anyway).
>
> >If my program, loaded and running, CALLs a subroutine for the first
> >then it will create a new working-storage for that. When I CANCEL it
> >will destroy that area and make it available for reuse by another.
>
> If you're talking about dynamic load, the first call is program load
time. The first call
> after cancel is another load.
>
> >This could happen even if the program was linked so that the program
> >code loaded with the run unit.
>
> No it doesn't. We went over this a month ago, when we discussed
ELF-style linking (loaded
> with the run unit) of Fujitsu programs. In that case, CANCEL did
nothing, in violation of
> the standard. The CANCELled program was NOT reinitialized. Now you are
claiming it is.

No, I am claiming that it _could_, just as my words _actually_ say.

If the dll or .so is linked so that it loads at the same time as the
run-unit then the W-S area for the programs that are in the dll or .so
don't need to be created until the program is first CALLed. This is
because the W-S is part of the run-unit but the code is part of
the .so.

That is, the initialisation is only required at the time when the
program in the library is CALLed for the first time. It is at this
point that the W-S is allocated and initialised. There may be several
or dozens of programs in a dll or .so, only a few of them actually
being called in a particular run. As the .so is _shared_ with other
run-units the W-S for those programs cannot be within the .so, they
are created as required in the run-unit itself.


> >In OO COBOL a new object will dynamically create a new working-storage
> >for itself.
>
> Not if it is a FACTORY object.

That is because you can't create a new factory object, there is and
can be only one factory object for a class and that will create the
new objects which will have  a new Working-Storage.

Actually, looking at Fujitsu, it says that the factory object will be
created "before it is referenced" which could be anytime between the
program loading and the first  NEW for that class.