Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Thu, 2 Mar 1995 20:51:09 GMT

The Environment Parameter system appears to be very poorly, if not actually 
broken in R3.12 (well, it was in Beta1).  Here's what I think it looks like:

base/include/envDefs.h		Declares parameters & defaults
base/src/misc/epicsEnvParams.h	Some IOC-only defaults
base/src/misc/epicsEnvParams.c	More IOC defaults, some IOC code
base/src/libCom/envSubr.c	IOC + host code
resource.def			Overrides for IOC
setenv ...			Overrides for Host tools

If you have different default settings in envDefs.h and epicsEnvParams.h, the 
IOC uses the ones in epicsEnvParams.h but the host tools the ones in envDefs.h. 
This is not much use for network stuff, where you need the same port numbers at 
both ends for it to work (has the log server port number been fixed?  In Beta1 
the envDefs.h and epicsEnvParams.c files gave different port numbers).  The 
epicsPrtEnvParams routine should either print *all* env parameters, or just the 
ones which affect the IOC in some way (why IOC_LOG_FILE_LIMIT ???).

Having this multiplicity of files also makes adding a new variable a bit of a 
pain. It would be nice to have *all* these defaults in one place, and I think 
they should go in the config directory so any site changes only need to happen 
there. Here's what I suggest:

1. Remove epicsEnvParams.c from miscLib, and put a revised version of 
epicsPrtEnvParams() into envSubr.c within a #ifdef vxWorks section.
2. Throw out epicsEnvParams.h completely.
3. Remove the default settings from envDefs.h
4. Add a file config/epicsEnvDefaults.h containing all defaults, which is 
#included by envDefs.h when ENV_PRIVATE_DATA is defined, ie in envSubr.c only.

This way, there is only one location for defaults, and only two places which 
need to be changed to add a new environment parameter.  I guess it would be 
possible to reduce that to one place, but I think it makes more sense to keep 
the declaration separate from the default.  I'd probably have the defaults as 
CPP symbols (DEFAULT_...), and maybe have a CPP macro to declare the ENV_PARAM 
structures to make the thing a bit cleaner and more readable.

Does anyone have any comments on this?  I'm happy to do the work if it is 
thought worthwhile.

- Andrew

          ,_}\_.        Andrew Johnson, Technology Division
         /      \       Royal Greenwich Observatory
 ~~~~~~~| ,--/;. )~~~~  Madingley Road, Cambridge, UK
    .,  |/  (/  U       Phone: +44 (0)1223 374823
  .`., /_\ . .          Email: [email protected]
     wwWWWww'           WWW: http://www.ast.cam.ac.uk/~anj/

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System