Database HW link parameter string no longer optional in V7

[Till Straumann]

Until and including 3.15 -- (I didn't look at 3.16) the parameter string '@<parameter>' was an optional part of hardware link fields.

In V7 this no longer seems to be the case

dbStaticLib:2320:

       if (parm && pinfo->ltype != RF_IO) {
            /* move parm string to beginning of buffer */
            memmove(pinfo->target, parm, len + 1);
        } else if (!parm && pinfo->ltype == RF_IO) {
            /* RF_IO, the string isn't needed at all */
            free(pinfo->target);
            pinfo->target = NULL;
        }
        else goto fail;

        return 0;

i.e., RF_IO links must not have a '@<parameter>' part but all other link types do.

Since this has the potential for breaking many existing databases it would seem desirable to de-escalate, e.g:

if ( !parm || pinfo->ltype == RF_IO ) {
  if ( pinfo->target ) {
    free(pinfo->target);
    pinfo->target = NULL;
  }
} else {
  memmove(pinfo->target, parm, len + 1);
}

return 0;

[Andrew Johnson (anj)]

Core Group review at ESS:

Regression, "@" should be optional. MD to resolve whether parm should be NULL or an empty string based on previous behavior.

[mdavidsaver (mdavidsaver)]

Changed with 7c00cc8045b7684fdc3d6df0e2b24d4e4eb07daf to again allow trailing '@' to be omitted for HW links other than INST_IO. So "#C0 S0" is equivalent to "@C0 S0 @".

[Ralph Lange (ralph-lange)]

This just hit us in the 4-day CODAC Core training.
Not a single hardware link in the hands-on session was working.
Phew!

[Ralph Lange (ralph-lange)]

@Michael: this commit is not in the 3.15 branch. Where can I find it?

[Ralph Lange (ralph-lange)]

Ah, 3.16. Sorry.