Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Wed, 25 Oct 1995 12:02:38 +0000 (GMT)

On Tue, 24 Oct 95 17:34:29 MDT Jeff Hill <[email protected]> wrote:
>Chris Timossi writes:
>
>> I still dont understand how the epics env. variables are now set
>> (when running under MS windows or NT).
...
>This problem will show up under VMS also. During the UNIX build 
>base/src/libCom/O.xxxx/envData.c is created from an ascii 
>configuration file. This file provides the defaults for the environment 
>variables. Unfortunately this file is not in with the rest
>of the source code. Worse yet it will not exist unless you have
>built the source distribution under UNIX. I suppose that we could include 
>an envData.c template for ms windows and VMS users in the libCom. Has 
>anyone got a better idea?

How do VMS and NT cope with blderrSymTbl, which does something very similar to make 
errSymTbl.c?

The base/src/libCom/O.*/envData.c files are identical for all target (and host) 
architectures at the same site, so provided you have one copy containing your site 
defaults you can use it for everything.

The reason for envData.c is that the defaults are site-specific, so either you have 
to put C code into the epics config directory and edit it directly (site-specific 
information should *not* go anywhere other than in config), or you build the C code 
from a CONFIG_* file.  A shell script seemed the simplest method when I wrote 
bldEnvData, which replaced a horrendous collection of files with different default 
values in several locations, resulting in different defaults for host vs target 
programs.  The problem with Jeff's template idea is that you're likely to end up 
going backwards on this unless you're careful.

I guess bldEnvData could be converted into a C program, but it seems rather a waste 
to have to use C for just simple text processing which can be done by a shell.  
Would this solve the VMS & NT issue?

I think this raises a more fundamental question about what tools we can use to 
build parts of EPICS base -- since NT and VMS are being supported for CA, and they 
don't allow the use of shell scripts, should we have a policy on this?  

How much of base do these CA-only clients actually use - could we have a top level 
Make target (to be run on a Unix system) which extracts the parts of source tree 
for a particular non-Unix host architecture and runs any shell scripts required?  
There are some notes at the top of the BUILD_VMS.COM script which indicate that 
some automation of this kind could be helpful for the VMS target at least.

- Andrew

  O o ._    Andrew Johnson, Royal Greenwich Observatory
     __\\_	Madingley Road, Cambridge, CB3 0EZ
  __|ooooo|__	Tel +44 (0)1223 374823  Fax 374700
~~\_________/~~	WWW http://www.ast.cam.ac.uk/~anj/

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System