coredump_on_error · execerror · nrn_feenableexcept · show_errmess_always

Error Handling

coredump_on_error()

Python

Syntax:

n.coredump_on_error(True or False)

Description:

On unix machines, sets a flag which requests (True) a coredump in case of memory or bus errors — Or floating exceptions if nrn_fenableexcept() has been turned on. 1 or 0 may be used as synonyms for True and False, respectively.

HOC

Syntax:

coredump_on_error(1 or 0)

Description:

On unix machines, sets a flag which requests (1) a coredump in case of memory or bus errors — Or floating exceptions if nrn_fenableexcept() has been turned on.

---

nrn_feenableexcept()

Python

Syntax:

previous_floating_point_mask = n.nrn_feenableexcept(boolean)

Description:

Sets or turns off a flag which, if on, causes a SIGFPE when a floating error occurs which consist of divide by zero, overflow, or invalid result. Known to work on linux. Turning on the flag is very helpful in finding the code location at which a variable is assigned a value of NaN or Inf. For a serial model, this is most easily done when running under gdb (or lldb on macOS). For a parallel model, one can combine with coredump_on_error and, to force a coredump on abort(), use the bash command ‘ulimit -c unlimited’.

Return is the previous value of the floating-point mask. (or -1 on failure to set the floating-point flags; or -2 if feenableexcept does not exist).

Without an arg, SIGFPE is turned on.

Note:

The normal trap for exp(x) for x > 700 in mod files becomes a floating exception when x is out of range.

HOC

Syntax:

previous_floating_point_mask = nrn_feenableexcept(boolean)

Description:

Sets or turns off a flag which, if on, causes a SIGFPE when a floating error occurs which consist of divide by zero, overflow, or invalid result. Known to work on linux. Turning on the flag is very helpful in finding the code location at which a variable is assigned a value of NaN or Inf. For a serial model, this is most easily done when running under gdb. For a parallel model, one can combine with coredump_on_error and, to force a coredump on abort(), use the bash command ‘ulimit -c unlimited’.

Return is the previous value of the floating-point mask. (or -1 on failure to set the floating-point flags; or -2 if feenableexcept does not exist).

Without an arg, SIGFPE is turned on.

Note:

The normal trap for exp(x) for x > 700 in mod files becomes a floating exception when x is out of range.

---

show_errmess_always()

Python

Syntax:

n.show_errmess_always(boolean)

Description:

Sets or turns off a flag which, if on, always prints the error message even if normally turned off by an execute1() statement or other call to the interpreter.

HOC

Syntax:

show_errmess_always(boolean)

Description:

Sets or turns off a flag which, if on, always prints the error message even if normally turned off by an execute1() statement or other call to the interpreter.

---

execerror()

Python

Syntax:

n.execerror("message1", "message2")

Description:

Raise an error and print the messages along with a NEURON interpreter stack trace. If there are no arguments, then nothing is printed.

The error may be caught in Python as a RuntimeError exception, however this happens after NEURON prints out its internal stack trace.

For example, consider the code:

from neuron import n
try:
    n.execerror("Unable to initialize model", "not enough parameters")
except RuntimeError as e:
    print(f"Caught an error: {e}")

Running this displays:

NEURON: Unable to initialize model not enough parameters
 near line 0
 objref hoc_obj_[2]
                   ^
        execerror("Unable to ...", "not enough...")
Caught an error: hocobj_call error: hoc_execerror: Unable to initialize model not enough parameters

For pure Python models, consider if using Python’s raise keyword will work instead.

HOC

Syntax:

execerror("message1", "message2")

Description:

Raise an error and print the messages along with an interpreter stack trace. If there are no arguments, then nothing is printed.