Re: VAX/VMS 7.3 and build from copy of master.

[John E. Malmberg]

What I am proposing:

1. Rename config.h-vms.template to be config-h-vms.template (or similar) 
so that it is the same on all VMS volumes.  Include this in the release 
tarballs.

2. Put a #include <types.h> near the start of config-h-vms.template so 
that __CRTL_VER is properly defined.

3. Have a procedure that converts the config-h-vms.template to config.h. 
 Since the information needed for this is in the configure.ac file.

4. At this point we longer need the config.h_vms file at all so it no 
longer needs to be generated as part of making the release tarball.
This simplifies things.

5. makefile.com will now always generate the config.h from 
config-h-vms.template.

4. Remove the gmk-default.h dependency from makefile.vms until such time 
that code becomes usable on VMS.

This way the same procedures are used for the build on masters or on 
tarballs so that they get fully tested.

On 2/21/2014 6:40 AM, h.becker wrote:
> Btw. are all NFS clients for VAX/VMS doing the identical file name mapping?

Yes.

> This seems to indicate that handling all of the possible file names for
> config.h-vms.template is even more work.

Not really.  I only need to deal with NFS name mangling at this time.

> I don't understand your comment on NFS files and VMS utilities. A
> temporary file would go to SYS$SCRATCH, or? Anyway, I suggested to make
> it more obvious what is replaced. The script with reading the file and
> processing it line by line is less obvious to me, maybe because the to
> be replaced tokens didn't show.

It is some of the utilities reading from the NFS volume that have the 
problem.  They hit some sort of issue where the process hangs for 8 hours.

> My opinion here is that just for NFS-problems and only for building from
> a snapshot it isn't worth to add complexity to the scripts. Agreed, the
> pipe command isn't simple, either.

I use this also for building from tarballs, and this is the only the 
first step in the GNV merge where I am getting the existing master to 
build on all the VMS versions that I have installed.

Eventually I plan to have a single procedure that will build from source 
and produce a PCSI kit with one command, and I will have to deal with 
the NFS name issues at several points.

> How far in VMS versions do you want to go back that there is no pipe
> command available?

It is not the issue of PIPE.  It is the issue that some NFS read 
operations from VMS utilities randomly fail on VMS 8.3. alpha.  The DCL 
reading through the file always works.

Regards,
-John