The aim of this project is to produce an automatic bug finding tool for array use in Fortran code. Fortran is used a lot in natural and physical sciences for writing large scale numerical models, and remains popular. Arrays are heavily used in this context for representing data sets and modelling quantities that vary in space and time. Fortran compilers provide some static checking of array use, but there are a number of situations which are not checked.
For example, the following program declares an array of size 10, and then calls a subroutine (function that has no results) using that array. However, the subroutine has no requirements on the length of the array, and reads from position 11. The behaviour here is undefined: this may cause a memory violation or may just read from uninitialized memory.
program AssumedSize
integer, dimension(10) :: arr
call sub(arr)
contains
subroutine sub(arr)
integer, dimension(*) :: arr
print *, arr(11)
end subroutine sub
end program AssumedSize
Other problems can occur when an array is 'reshaped' but the reshaping information is not properly propagated. A more thorough treatment of various problems can be found here: https://github.com/camfort/project-ideas/blob/master/array-problems.pdf.
The aim of this project is to implement a static analysis for Fortran arrays, providing an automated bug finding tool. We have already developed an analysis framework for Fortran, called CamFort [0] which can be built upon. This will allow the project to move forward rapidly without having to spend time building basic components of parsers, lexers, and standard code analyses. We have a Fortran corpus comprising scientific Fortran packages and well exceeding one million lines of code. The project can be evaluted on this corpus.