| Newsgroups |
gmane.comp.emulators.hercules390.mvs |
| Message-ID |
<[email protected]> |
> -----Original Message-----
> From: [email protected] <[email protected]>
> Sent: 11 September 2019 18:50
> To: [email protected]
> Subject: Re: [H390-MVS] MVT COBOL under MVS3.8j.
>
> On Wed, 11 Sep 2019 at 04:17, Giuseppe Vitillaro giuseppe-yuD/ahkh7LvrZ44/[email protected] wrote:
>
> > It looks like the compiler, when time to create the symbol table
> > comes, needs to allocate memory for performing the task, but there is
> > not any large enough contiguos area available in the job region.
>
> A common technique for compilers that don't know in advance how much work
> space they'll need is to issue a variable-length GETMAIN for a very large
> amount, and then FREEMAIN part of it to allow space for buffers and other
> things that either the OS or other compiler components will need. One wonders
> if the compiler is grabbing most of the available space, and not returning
> enough of it.
>
> > Mmm ... this is a COBOL compiler designed for MVT, isn't it?
> >
> > Is possible GETMAIN calls are different between MVS and MVT?
>
> MVS has many more GETMAIN options, but the MVT GETMAIN calls should
> behave entirely compatibly. In MVS the notion of REGION is rather different
> from that in MVT. In the absence of other controls (SMF,
> etc.) a GETMAIN on MVT cannot return a piece of storage larger than the
> REGION= value. On MVS, the REGION= value limits the size of any
> *single* GETMAIN, but if there is room in the private area, GETMAIN can be
> issued again and again until the private area is used up.
> (Well, in chunks the size of the GETMAIN.)
>
> But I'm not sure this is relevant to the COBOL problem. And BTW, this seems to
> be a problem in the compiler rather than a true "not enough storage" issue.
> There is no reported abend or GETMAIN failure - just a compiler message.
>
Oddly the level of OS implemented in VM/CMS is that of MVT, I can't remember the exact level.
So it could be an issue with the way GETMAIN works, but since every other MVT compilers work I can't see it...
I believe one of the main drivers of implementing this was to allow the MVT compliers to work in VM
> Tony H.
>
Dave
>
> ------------------------------------
> Posted by: Tony Harminc <[email protected]>
> ------------------------------------