NMODL code no longer works in version 8.2

[pascal]

Many moons ago at a NEURON workshop, Michael Hines helped me write the following NMODL code, which I used to stochastically stimulate various sets of synapses:

Code: Select all

:this code uses Random123, which requires NEURON 7.3 or higher
:uses nrnran123.c and nrnran123.h from http://www.neuron.yale.edu/hg/neuron/nrn/file/9d4ab20927bc/src/oc/
VERBATIM
#define VOIDCAST void** vp = (void**)(&(_p_ptr))
extern void * nrnran123_newstream(int,int);
extern void nrnran123_deletestream(void *);
extern double nrnran123_dblpick(void *);
ENDVERBATIM

PROCEDURE setrand(id1,id2) {
	VERBATIM
	VOIDCAST;
	if(*vp) {
		nrnran123_deletestream(*vp);
	} 
	*vp = nrnran123_newstream((int) _lid1,(int) _lid2);
	ENDVERBATIM
} 

FUNCTION pick() {
	VERBATIM
	VOIDCAST;
	_lpick = nrnran123_dblpick(*vp);
	ENDVERBATIM
}

This worked fine in NEURON 8.0, but after upgrading to 8.2 I get the following compilation error:

Compiling ampa_D2.c
ampa_D2.c:499:15: error: conflicting types for ‘nrnran123_newstream’; have ‘void *(int, int)’
499 | extern void * nrnran123_newstream(int,int);
| ^~~~~~~~~~~~~~~~~~~
In file included from /home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/mech_api.h:15,
from ampa_D2.c:7:
/home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/nrnran123.h:45:25: note: previous declaration of ‘nrnran123_newstream’ with type ‘nrnran123_State *(uint32_t, uint32_t)’ {aka ‘nrnran123_State *(unsigned int, unsigned int)’}
45 | extern nrnran123_State* nrnran123_newstream(uint32_t id1, uint32_t id2);
| ^~~~~~~~~~~~~~~~~~~
ampa_D2.c:500:13: error: conflicting types for ‘nrnran123_deletestream’; have ‘void(void *)’
500 | extern void nrnran123_deletestream(void *);
| ^~~~~~~~~~~~~~~~~~~~~~
In file included from /home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/mech_api.h:15,
from ampa_D2.c:7:
/home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/nrnran123.h:47:13: note: previous declaration of ‘nrnran123_deletestream’ with type ‘void(nrnran123_State *)’
47 | extern void nrnran123_deletestream(nrnran123_State*);
| ^~~~~~~~~~~~~~~~~~~~~~
ampa_D2.c:501:15: error: conflicting types for ‘nrnran123_dblpick’; have ‘double(void *)’
501 | extern double nrnran123_dblpick(void *);
| ^~~~~~~~~~~~~~~~~
In file included from /home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/mech_api.h:15,
from ampa_D2.c:7:
/home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/include/nrnran123.h:55:15: note: previous declaration of ‘nrnran123_dblpick’ with type ‘double(nrnran123_State *)’
55 | extern double nrnran123_dblpick(nrnran123_State*); /* uniform open interval (0,1)*/
| ^~~~~~~~~~~~~~~~~
make: *** [makemod2c_inc:31: ampa_D2.o] Error 1
make: *** Waiting for unfinished jobs....
Traceback (most recent call last):
File "/home/tcgfink/.local/bin/nrnivmodl", line 94, in <module>
subprocess.check_call([exe, *sys.argv[1:]])
File "/usr/lib/python3.10/subprocess.py", line 369, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['/home/tcgfink/.local/lib/python3.10/site-packages/neuron/.data/bin/nrnivmodl']' returned non-zero exit status 2.

Reading the documentation for 8.2, I'm guessing the error is due to Random123 now using C++ rather than C? Whatever the reason, my C and/or C++ skills are too rusty for me to be sure how to fix this. Thank you for the help.

[ramcdougal]

I think the main issue you're running into here is that NEURON is trying to be too helpful. You're getting a bunch of "conflicting types" errors because some things are now available directly to NMODL without needing a VERBATIM declaration.

What happens if you simply remove the three extern lines that it's complaining about?

[hines]

try replacing the three lines

Code: Select all

#define VOIDCAST void** vp = (void**)(&(_p_ptr))
extern void * nrnran123_newstream(int,int);
extern void nrnran123_deletestream(void *);

with the single line

Code: Select all

#define VOIDCAST nrnran123_State** vp = (nrnran123_State**)(&(_p_ptr))

and also

Code: Select all

*vp = nrnran123_newstream((int) _lid1,(int) _lid2);

with the line

Code: Select all

*vp = nrnran123_newstream((uint32_t) _lid1,(uint32_t) _lid2);

[pascal]

Thank you both for the help. I could use a little more advice, though, because both approaches seemed to work. Simply removing all three extern lines resulted in successful compilation. But your approach also resulted in successful compilation, Michael (although I did have to also remove the third extern line). Which approach should be preferred?

[hines]

successful compilation

No warnings?
I would prefer accurate use of types. I.e.

Code: Select all

nrnran123_State*

instead of

Code: Select all

void*

and

Code: Select all

uint32_t

instead of

Code: Select all

int

(in that particular case)

[pascal]

Thanks, that makes sense. Also, I stumbled on one more issue: my mod files are now compiling fine using the Linux distribution, but in Windows I receive the following error message:

Ih.c:7:10: fatal error: scoplib_ansi.h: No such file or directory
7 | #include "scoplib_ansi.h"
| ^~~~~~~~~~~~~~~~
compilation terminated.
make: *** [/cygdrive/c/nrn/bin//../lib/mknrndll.mak:26: Ih.o] Error 1

There was an error in the process of creating nrnmech.dll

Here is the code for Ih.mod:

Code: Select all

TITLE anomalous rectifier channel
COMMENT
:
: Anomalous Rectifier Ih - cation (Na/K) channel in thalamocortical neurons
:
: Kinetic model of calcium-induced shift in the activation of Ih channels.
: Model of Destexhe et al., Biophys J. 65: 1538-1552, 1993, based on the
: voltage-clamp data on the calcium dependence of If in heart cells
: (Harigawa & Irisawa, J. Physiol. 409: 121, 1989)
:
: The voltage-dependence is derived from Huguenard & McCormick, 
: J Neurophysiol. 68: 1373-1383, 1992, based on voltage-clamp data of 
: McCormick & Pape, J. Physiol. 431: 291, 1990. 
:
: Modified model of the binding of calcium through a calcium-binding (CB)
: protein, which in turn acts on Ih channels.  This model was described in
: detail in the following reference:
:    Destexhe, A., Bal, T., McCormick, D.A. and Sejnowski, T.J.  Ionic 
:    mechanisms underlying synchronized oscillations and propagating waves
:    in a model of ferret thalamic slices. Journal of Neurophysiology 76:
:    2049-2070, 1996.
: See also http://www.cnl.salk.edu/~alain , http://cns.fmed.ulaval.ca
:
:   KINETIC MODEL:
:
:	  Normal voltage-dependent opening of Ih channels:
:
:		c1 (closed) <-> o1 (open)	; rate cst alpha(V),beta(V)
:
:	  Ca++ binding on CB protein
:
:		p0 (inactive) + nca Ca <-> p1 (active)	; rate cst k1,k2
:
:	  Binding of active CB protein on the open form (nexp binding sites) :
:
:		o1 (open) + nexp p1 <-> o2 (open)	; rate cst k3,k4
:
:
:   PARAMETERS:
:	It is more useful to reformulate the parameters k1,k2 into
:	k2 and cac = (k2/k1)^(1/nca) = half activation calcium dependence, 
:	and idem for k3,k4 into k4 and Pc = (k4/k3)^(1/nexp) = half activation
:	of Ih binding (this is like dealing with tau_m and m_inf instead of
:	alpha and beta in Hodgkin-Huxley equations)
:	- k2:	this rate constant is the inverse of the real time constant of 
:             	the binding of Ca to the CB protein
:	- cac:	the half activation (affinity) of the CB protein;
:		around 1 to 10 microM.  
:	- k4:	this rate constant is the inverse of the real time constant of 
:             	the binding of the CB protein to Ih channels
:		very low: it basically governs the interspindle period
:	- Pc:	the half activation (affinity) of the Ih channels for the
:		CB protein;
:	- nca:	number of binding sites of calcium on CB protein; usually 4
:	- nexp:	number of binding sites on Ih channels
:       - ginc: augmentation of conductance associated with the Ca bound state
:	  (about 2-3; see Harigawa & Hirisawa, 1989)
:
:
:   IMPORTANT REMARKS:
:       - This simple model for the binding of Ca++ on the open channel 
:	  suffies to account for the shift in the voltage-dependence of Ih
:	  activation with calcium (see details in Destexhe et al, 1993).
:	- It may be that calcium just binds to the Ih channel, preventing the 
:	  conformational change between open and closed; in this case one
:	  should take into account binding on the closed state, which is 
:	  neglected here.
:
:   MODIFICATIONS
:	- this file also contains a procedure ("activation") to estimate
:	  the steady-state activation of the current; callable from outside
:	- the time constant now contains a changeable minimal value (taum)
:	- shift: new local variable to displace the voltage-dependence
:	  (shift>0 -> depolarizing shift)
:
:
: Alain Destexhe, Salk Institute and Laval University, 1995
:
: modified by Chris Fink to remove dependence on temperature
: also replaced 'shift' with 'fac_gh_TC'
ENDCOMMENT

NEURON {
	SUFFIX iar
	USEION h READ eh WRITE ih VALENCE 1
	USEION ca READ cai
    RANGE ghbar, h_inf, tau_s, m, fac_gh_TC
	GLOBAL k2, cac, k4, Pc, nca, nexp, ginc, taum
}

UNITS {
	(molar)	= (1/liter)
	(mM)	= (millimolar)
	(mA) 	= (milliamp)
	(mV) 	= (millivolt)
	(msM)	= (ms mM)
}


PARAMETER {
	eh	= -40	(mV) :Bazhenov C++ line 290
	:celsius = 36	(degC)
	ghbar	= 0.000017 (mho/cm2)    :Bazhenov C++ line 1506
	cac	= 0.0015 (mM)		: half-activation of calcium dependence (Bazhenov C++ line 272)
	k2	= 0.0004 (1/ms)		: inverse of time constant (Bazhenov C++ line 292)
	Pc	= 0.007			: half-activation of CB protein dependence (Bazhenov C++ line 1497)
	k4	= 0.001	(1/ms)		: backward binding on Ih (Bazhenov C++ line 274 & 1498)
	nca	= 4			: number of binding sites of ca++ (Bazhenov C++ line 293)
	nexp	= 1			: number of binding sites on Ih channels (Bazhenov C++ line 293)
	ginc	= 2.0			: augmentation of conductance with Ca++ (Bazhenov C++ line 1493)
	taum	= 20	(ms)		: min value of tau (Bazhenov C++ line 293)
	fac_gh_TC	= 0	(mV)		: shift of Ih voltage-dependence
}


STATE {
	c1	: closed state of channel
	o1	: open state
	o2	: CB-bound open state
	p0	: resting CB
	p1	: Ca++-bound CB
}


ASSIGNED {
	v	(mV)
	cai	(mM)
	ih	(mA/cm2)
    gh	(mho/cm2)
	h_inf
	tau_s	(ms)
	alpha	(1/ms)
	beta	(1/ms)
	k1ca	(1/ms)
	k3p	(1/ms)
	m
	tadj
}


BREAKPOINT {
	SOLVE ihkin METHOD sparse

	m = o1 + ginc * o2

	ih = ghbar * m * (v - eh)
}

KINETIC ihkin {
:
:  Here k1ca and k3p are recalculated at each call to evaluate_fct
:  because Ca or p1 have to be taken at some power and this does
:  not work with the KINETIC block.
:  So the kinetics is actually equivalent to
:	c1 <-> o1
:	p0 + nca Cai <-> p1
:	o1 + nexp p1 <-> o2

	evaluate_fct(v,cai)

	~ c1 <-> o1		(alpha,beta)

	~ p0 <-> p1		(k1ca,k2)

	~ o1 <-> o2		(k3p,k4)

	CONSERVE p0 + p1 = 1
	CONSERVE c1 + o1 + o2 = 1
}





INITIAL {
:
:  Experiments of McCormick & Pape were at 36 deg.C
:  Q10 is assumed equal to 3
:
        tadj = 1.0 :3.0 ^ ((celsius-36 (degC) )/10 (degC) ) :Krishnan currents.h line 379

	evaluate_fct(v,cai)

	: commented-out intializations are what was used in 2002 Bazhenov C++ code (see 6/29/19 journal entry)
	p1 = 0 :1/(1 + (cac/cai) ^ nca)
	o1 = 0  :1/(1 + beta/alpha + (p1/Pc)^nexp )
	o2 = 0 :((p1/Pc)^nexp) * o1
	c1 = 1 :1-o1-o2
	p0 = 1 :1-p1
	
}


UNITSOFF
PROCEDURE evaluate_fct(v (mV), cai (mM)) {
	: CF: I replaced 'shift' with 'fac_gh_TC' and *added* fac_gh_TC (rather than subtracting shift), in order to replicate Krishnan's approach in currents.cpp
	h_inf = 1 / ( 1 + exp((v+75+fac_gh_TC)/5.5) )

	tau_s = (taum + 1000 / ( exp((v+71.5+fac_gh_TC)/14.2) + exp(-(v+89+fac_gh_TC)/11.6) ) ) / tadj

	alpha = h_inf / tau_s
	beta  = ( 1 - h_inf ) / tau_s

	k1ca = k2 * (cai/cac)^nca :Krishnan currents.cpp line 178

	k3p = k4 * (p1/Pc)^nexp :Krishnan currents.cpp line 179

}



:
:  procedure for evaluating the activation curve of Ih
:
PROCEDURE activation(v (mV), cai (mM)) { LOCAL cc

	evaluate_fct(v,cai)

	cc = 1 / (1 + (cac/cai)^nca ) 		: equil conc of CB-protein (Bazhenov C++ line 281)

	m = 1 / ( 1 + beta/alpha + (cc/Pc)^nexp ) :Bazhenov C++ line 282

	m = ( 1 + ginc * (cc/Pc)^nexp ) * m :Bazhenov C++ line 283
}

UNITSON

[hines]

scoplib_ansi.h was removed from the system back in Jan 2022. I expect you need a clean folder where you type nrnivmodl. Just remove all the *.o, *.c, *.cpp, *.dll files from your folder. (and remove the x86_64 folder if it exists in your folder)

[pascal]

That did it...thank you very much for all your help!