#72 Printing hidden errors to terminal

[Bohdan Tkhir]

Hi!
I've had a problem when I've made little mistake using cm_event_queue().
I've called model in time == TIME and time < TIME.
There was no warning or errors in terminal.
So,
How about printing errors from file cmevt.c ?

change this :

    /* If breakpoint time <= current event time, return error */
    if(time <= g_mif_info.circuit.evt_step) {
        g_mif_info.errmsg =
        "ERROR - cm_event_queue() - Event time cannot be <= current time\n";
        return(MIF_ERROR);
    }

to this:

    /* If breakpoint time <= current event time, return error */
    if(time <= g_mif_info.circuit.evt_step) {
        g_mif_info.errmsg =
        "ERROR - cm_event_queue() - Event time cannot be <= current time\n;
        printf("ERROR - cm_event_queue() - Event time cannot be <= current time\nInstance: %s\n", g_mif_info.instance->MIFname);
        return(MIF_ERROR);
    }

Maybe, it makes sense to print other hidden errors? Especially in debug mode.

---

Thanks,
Bohdan

[Bohdan Tkhir]

By the way.
Generic xspice adc_bridge. It switches between ANALOG and EVENT by making cm_event_queue(current_time).
Isn't it an error?

V_SRC vdd 0 DC 0 SIN (0 1 10)
A_ADC [vdd] [events] ADC
    .model ADC adc_bridge(in_low=0.5 in_high=0.5)
.control
tran 1 1
run
plot events
.endc

Console output after printing errors:

Initial Transient Solution

Node                                   Voltage
----                                   -------
vdd                                          0
v_src#branch                                 0

ERROR - cm_event_queue() - Event time cannot be <= current time
Instance: a_adc
ERROR - cm_event_queue() - Event time cannot be <= current time
Instance: a_adc
ERROR - cm_event_queue() - Event time cannot be <= current time
Instance: a_adc
......

[marcel hendrix]

I don't see anything out of the order (with ngspice-29).
There are two events for each positive sinewave half, as expected.
(Making in_low and in_high slightly different doesn't change this).

ngspice 1 -> eprint events

**** Results Data ****

Time or Step
events


0.000000000e+00     0s
1.280000100e-02     1s
4.560000100e-02     0s
1.256000010e-01     1s
1.456000010e-01     0s
2.256000010e-01     1s
2.456000010e-01     0s
3.256000010e-01     1s
3.456000010e-01     0s
4.256000010e-01     1s
4.456000010e-01     0s
5.256000010e-01     1s
5.456000010e-01     0s
6.256000010e-01     1s
6.456000010e-01     0s
7.256000010e-01     1s
7.456000010e-01     0s
8.256000010e-01     1s
8.456000010e-01     0s
9.256000010e-01     1s
9.456000010e-01     0s



**** Messages ****


**** Statistics ****

Operating point analog/event alternations:  1
Operating point load calls:                 2
Operating point event passes:               2
Transient analysis load calls:              58
Transient analysis timestep backups:        0

-marcel

[Bohdan Tkhir]

Hi, Marcel!
It will work , but there will be hidden errors in g_mif_info structure.
It's not about in_low or in_high.
Did you change cmevt.c as described in main message?( It requires rebuild of ngspice )
Just now checked on 31+ (pre-master ):

ERROR - cm_event_queue() - Event time cannot be <= current time
Instance: a_adc

It is caused by next lines of generic adc_bridge:

 /* call for event breakpoint... */
 current_time = TIME;
 cm_event_queue(current_time);

It means that we are calling model in time == TIME , which is an error in order to cmevt.c:244-249

[marcel hendrix]

On 2020-01-10 09:46, Bohdan Tkhir wrote:

By the way.
Generic xspice adc_bridge. It switches between ANALOG and EVENT by
making cm_event_queue(current_time).
Isn't it an error?
~~~
V_SRC vdd 0 DC 0 SIN (0 1 10)
A_ADC [vdd] [events] ADC
.model ADC adc_bridge(in_low=0.5 in_high=0.5)
.control
tran 1 1
run
plot events
.endc
~~~

Did you intentionally make in_low == in_high to provoke this error?

-marcel

[Bohdan Tkhir]

No, It's not about in_low == in_high, you can set anything you want
(Replied to this question extensively in comment above).

[Giles Atkinson]

Here is an error reporting mechanism that never reports an error.

[Giles Atkinson]

cm_event_queue(current_time);
... is almost certainly correct, as far as I understand it. So that probably should be '<', not '<=".

[Giles Atkinson]

Raising priority as it has just cost me hours of debugging time. Needs fixing, but is it limited to XSPICE? Check for use of macros in inpmacs.h.