ID0001178: **This issue was created automatically from Mantis Issue 1178. Further discussion may take place here.** --- | **Id** | **Project** | **Category** | **View** | **Due Date** | **Updated** | | --- | --- | --- | --- | --- | --- | | ID0001178 | Frama-C | Kernel | public | 2012-05-04 | 2012-09-19 | | | | | | | | | --- | --- | --- | --- | --- | --- | | **Reporter** | Anne | **Assigned To** | virgile | **Resolution** | duplicate | | **Priority** | normal | **Severity** | major | **Reproducibility** | have not tried | | **Platform** | - | **OS** | - | **OS Version** | - | | **Product Version** | Frama-C Nitrogen-20111001 | **Target Version** | - | **Fixed in Version** | Frama-C Oxygen-20120901 | ### Description : On the following program : void f (void) { unsigned char c = 0xef; out (c); } void out (unsigned x) {} int main (void) { f(); return 0; } Frama-C alone seems happy, but the value analysis gives: toto.c:1:[value] warning: Pointed function type must match function pointer type when dereferenced: assert(Ook) which leads to : [value] Values for function f: NON TERMINATING FUNCTION. The problem disappears when out is declared before f. Frama-C should at least gives a warning (or fail ?) ### Additional Information : This is very strange since this warning occurs when processing the call to 'out' because [Kernel_function.get_return_type f] is different from the return type of the TFun type [Cil.typeOf funcexp]. It seems to me than both evaluation should go though the same variable, so there might be two different variables for the 'out' function ???