[Codescan] Release or reuse EDA-FF parameter arrays before re-entering the analysis

[Stardust0831]

Global scan provenance

This issue was found by the Codex global repository scan of every non-unit-test file in Stardust0831/Multiwfn, pinned to commit d2b770a531eb7039153198d04520b5992f15c49d.

Problem

parmA and parmB are allocatable variables in the saved EDA_FF_mod module state:

Multiwfn/EDA.f90

Lines 44 to 48 in d2b770a

	!------ Module for EDA-FF
	module EDA_FF_mod
	integer :: ielemode=1,ivdwmode=2
	real*8,allocatable :: parmA(:),parmB(:)
	end module

Every entry into EDA_forcefield() allocates both arrays unconditionally:

Multiwfn/EDA.f90

Lines 50 to 66 in d2b770a

	subroutine EDA_forcefield
	use defvar
	use util
	use EDA_FF_mod
	implicit real*8 (a-h,o-z)
	integer :: nfrag=0,ishowatmpair=0,ioutatmpqr=0
	character c200tmp*200,c200tmp2*200,c2000tmp*2000
	integer,allocatable :: frag(:,:),fragnatm(:)
	real*8,allocatable :: elemat(:,:),repmat(:,:),dispmat(:,:),totmat(:,:) !Interfragment interaction matrix of electrostatic, repulsive, dispersion and total interaction
	real*8 eleatm(ncenter),repatm(ncenter),dispatm(ncenter),totatm(ncenter)
	!vdW parameter of each atom, for UFF they are well depths and vdW distance; For AMBER/GAFF they are well depths and vdW radii
	character(len=2) c2tmp,FFtype(ncenter) !Force field atom types, only needed by AMBER and GAFF
	FFtype="?"
	if (ifiletype/=4) a%charge=0 !If the input file is not chg or pqr format, assume atomic charges to be zero
	allocate(parmA(ncenter),parmB(ncenter))

Selecting 0 Return exits the subroutine without deallocating them:

Multiwfn/EDA.f90

Lines 81 to 122 in d2b770a

	write(*,*) "0 Return"
	write(*,*) "1 Start analysis"
	if (allocated(frag)) then
	write(*,"(a,i4,a)") " 2 Redefine fragments, current:",nfrag," fragments"
	else
	write(*,"(a)") " 2 Define fragments (undefined currently)"
	end if
	if (ivdwmode==1) then !UFF
	write(*,*) "3 Load atomic charges for current system"
	else !AMBER&GAFF
	write(*,*) "3 Load atomic types and charges for current system"
	end if
	if (ivdwmode==1) then !UFF
	write(*,*) "4 Show current atomic charges"
	else !AMBER&GAFF
	write(*,*) "4 Show current atomic types and charges"
	end if
	read(*,*) isel
	if (isel==-4) then
	if (ioutatmpqr==1) then
	ioutatmpqr=0
	else
	ioutatmpqr=1
	end if
	else if (isel==-3) then
	if (ishowatmpair==1) then
	ishowatmpair=0
	else
	ishowatmpair=1
	end if
	else if (isel==-2) then
	write(*,*) "1: 1/r interaction potential"
	write(*,*) "2: 1/r^2 interaction potential"
	read(*,*) ielemode
	else if (isel==-1) then
	write(*,*) "1: UFF van der Waals model"
	write(*,*) "2: AMBER99 & GAFF van der Waals model"
	read(*,*) ivdwmode
	else if (isel==0) then
	return

The parent EDA menu remains active and allows option 1 to call EDA_forcefield() again in the same process:

Multiwfn/EDA.f90

Lines 1 to 30 in d2b770a

	!-------- Main interface of various energy decomposition analyses
	subroutine EDA_main
	use defvar
	implicit real*8 (a-h,o-z)
	do while(.true.)
	write(*,*)
	write(*,*) " ============ Energy decomposition analysis ============ "
	write(*,*) "0 Return"
	write(*,*) "1 Energy decomposition analysis based on molecular forcefield (EDA-FF)"
	write(*,*) "2 Shubin Liu's energy decomposition analysis (Gaussian is needed)"
	write(*,*) "3 sobEDA and sobEDAw energy decomposition analyses (J. Phys. Chem. A, 127, 7023 (2023))"
	write(*,*) "4 Analysis of atomic contribution to dispersion energy"
	! write(*,*) "2 Mayer energy decomposition analysis"
	! write(*,*) "3 Fuzzy space based energy decomposition analysis"
	read(*,*) infuncsel2
	if (infuncsel2==0) then
	return
	else if (infuncsel2==1) then
	call EDA_forcefield
	else if (infuncsel2==2) then
	call EDA_SBL
	else if (infuncsel2==3) then
	write(*,"(a)") " This kind of analysis needs using shell script. Please check detailed sobEDA/sobEDAw tutorial: http://sobereva.com/soft/sobEDA_tutorial.zip"
	write(*,*) "Also see original paper: J. Phys. Chem. A, 127, 7023 (2023) DOI: 10.1021/acs.jpca.3c04374"
	write(*,*) "Press ENTER to return"
	read(*,*)
	else if (infuncsel2==4) then
	call atomdispcontri
	end if
	end do

The second entry therefore executes allocate(parmA(...), parmB(...)) while both allocatables are already allocated, which is a Fortran runtime error.

Impact

A normal interactive sequence, EDA -> EDA-FF -> Return -> EDA-FF, can terminate the program instead of reopening the analysis. Reloading a structure with a different number of atoms also leaves arrays with stale dimensions until the failing allocation is addressed.

Suggested direction

• Prefer local allocatables if these parameters do not need to survive the subroutine.
• Otherwise, deallocate them on every return path or guard/reallocate when their size differs from ncenter.
• Add a cleanup routine for EDA-FF module state and call it from whole-system cleanup.
• Add a functional test that enters and leaves EDA-FF twice in one process.