The product of two integers can overflow in L-BFGS

[junzi-taotao]

What happened?

Hi, i'm trying to use L-BFGS in v2.6.2. But it return failure when the dimension of the problem is 12852225.

How to reproduce the issue?

set *nf to 12852225 for function plis_
static void plis_(int *nf, int *nb, double *x, ...)

Version

v2.6.2

Operating System

Linux

Installation media

unknown

Additional Context

"kit = -(ires1 * *nf + ires2);" this line is in function pils_, line 236, variable kit / ires1 / *nf / ires2 are all int, ires1 is 999, ires2 is 0, *nf is 12852225, then the actual result of -(ires1 * *nf + ires2) is -12,839,372,775, it is less than the lower boundary of int, and this value will cause the following flow incorrect.
Should change the type of kit from int to long (long long)?

[stevengj]

arguably all array lengths should be size_t and all indices should be either size_t or ptrdiff_t.

[junzi-taotao]

yes, and it will change a lot, including struct, interface parameters and so on.

arguably all array lengths should be size_t and all indices should be either size_t or ptrdiff_t.