Re: VMS progname and exit handling - Take 4

[John E. Malmberg]

On 3/31/2014 12:48 PM, h.becker wrote:
> On 03/31/2014 06:46 AM, John E. Malmberg wrote:
> > This has the makefile changes needed to build from master and on VAX/VMS.
> >
> > In addition, vms_progname.c that is compatible as a VMS specific
> > replacement for the progname.c that is used for several GNU projects.
> >
> > This variant of vms_progname.c has an additional feature that it
> > temporarily creates a foreign command that make uses for a macro to
> > represent the command that invoked make.  This foreign command has no
> > embedded spaces or dollar sign characters which could cause unexpected
> > results if further string processing is done.
> >
> > The vms_exit() is similar to the /DEFINE=_POSIX_EXIT, except that it
> > allows pass-through of VMS exit statuses of 8 or higher, and allows
> > setting the SEVERITY and message inhibit settings for the encoded Posix
> > status.  The message inhibit bits prevents the DCL shell from
> > re-displaying status messages that have already been displayed.
> >
> > Additional work needs to be done to convert POSIX encoded child exit
> > statuses back to the original Unix status internally.  This is usually
> > handled by the /DEFINE=_POSIX_EXIT.
> >
> > We are not simply using the /DEFINE=_POSIX_EXIT as it does not support
> > mixing the VMS exit codes and Unix exit codes, along with setting
> > special VMS status bits.
>
> It seems I need more than this patch to compile a test version. It seems
> this needs to be applied on top of a previous patch, or?

I was missing the changed main.c and makeint.h.  New patch includes them.

This patch was based on a branch made directly from master.

> Again, this looks like a lot of code to set the program name and to
> handle the exit code. And I'm still not convinced that defining a DCL
> symbol and deleting it at image exit is the right thing to do. As far as
> I can see, it can be simpler than that.

If you have the make variable be a MCR command, it has a space which can 
have side effects when if the resulting string is processed by other 
strings or macros.

With ODS-5 format filenames, the sys$filescan system service is the most 
reliable way detect and parse a VMS filename.  This adds to the 
complexity.  The other complexity is working around some bugs in how the 
C pre-main function renders the Unix format filenames.

Regards,
-John

0001-VMS-progname-and-exit-handling-take-5.patch.gz
Description: application/gzip