ifort does not notice array dimensions are wrong

[trobitaille]

I have run into a strange problem that was not diagnosed by ifort even with the -check all -warn all flags. The short code attached to this message compiles without complaning, but then produces wrong results (an array with intent(in) gets modified). There is a bug in the code, which is that the dimension of the c array is in this case (2,19) whereas the dimensions of y are (2,39). However, ifort doesn't pick this up, and instead produces wrong results.

I am using

Intel Fortran Intel 64 Compiler Professional for applications running on Intel 64, Version 11.1 Build 20090827 Package ID: m_cprof_p_11.1.067

Copyright (C) 1985-2009 Intel Corporation. All rights reserved.

[Steven_L_Intel1]

ifort can't pick this up because you prevented it from doing so. In routine b, you declared array y as (2*nx-1). The compiler takes your word for it and uses that as the dimension. If you had declared y as (:) instead not only would this avoid making a temporary copy for the non-contiguous array slice you pass, but now you'd get a run-time error like this:

forrtl: severe (408): fort: (2): Subscript #1 of the array Y has value 20 which is greater than the upper bound of 19

[joseph-krahn]

I think the OP was hoping for a compile-time warning. Often, the compiler will warn if the actual array argument is too small, but I think that only works when the actual and dummy arguments both have a static size.

For example, change the subroutine b dummy argument to "y(100)" and the actual argument in the call from routine a to "C(1,1:5)". Then, both sizes are static, and 5 is too small, giving this error:

ifort_bug.f90(35): error #7983: The storage extent of the dummy argument exceeds that of the actual argument.
call b(size(x), c(1, 1:4), x)
--------------------^

Not also that a larger actual argument does not produce a warning.

[Steven_L_Intel1]

In the original example, the compiler does not know, at compile time, what the extent of the actual argument is.