#128 Timestep too small could happen if insert a breakpoint close to tran stop time

[qingran zheng]

This is a kind of related to another issue: https://sourceforge.net/p/ngspice/patches/127/, According to the explaination there, CKTbreaks will have a minial size of 2, with both value to be CKTfinalTime.

    if(ckt->CKTbreakSize >2) {
        ...
    } else {
        ckt->CKTbreaks[0] = ckt->CKTbreaks[1];
        ckt->CKTbreaks[1] = ckt->CKTfinalTime;
    }

In this case, if a new breakpoint is inserted and it is very close to CKTfinalTime, e.g. CKTfinalTime - a few ULPs, then in CKTsetBreak func, it will be inserted successfully, and the breakpoints will be {CKTfinalTime - a few ULPs, CKTfinalTime}.

    for(i=0;i<ckt->CKTbreakSize;i++) {
        if(ckt->CKTbreaks[i]>time) { /* passed */
            if((ckt->CKTbreaks[i]-time) <= ckt->CKTminBreak) {
                /* very close together - take earlier point */
                ckt->CKTbreaks[i] = time;
                return(OK);
            }

If so, ckt->CKTbreaks[1] - ckt->CKTbreaks[0] = a few ULPs, and jt could cause ckt->CKTdelta to be set to 0.1 * the ULPs delta in Dctran and then leads simulation abort (see more details in https://sourceforge.net/p/ngspice/patches/127/).

A simple fix is to add a check to see if ckt->CKTbreaks[i] == ckt->CKTfinalTime. Please have a look on the patch.