Can anyone sol ve this casting issue?

[Loctan]

We have a structure defined something like (it does a lot more than this, but this is all that's necessary for my problem):

typdef struct
{

some_data_type source;

void* val;

int default

}blah_st;

Now I populate an array of these structures as follows: first choose data from source and if for some reason it isn't valid, we default val to default. Now the problem I have is that sometimes val isn't an int - sometimes it's a float. I have been solving this by putting in the IEEE floating point representation of the floating point default (ie, 0x428C0000 for 70.0) but we know it's wrong to using the underlying floating point representation this way. We would like to do something and be able to type in 70.0 and let the compiler figure it out.

Now we have a solution for this that involves fairly significant rework. Can you see any way to do this without using things like void* to default, etc that would cause me to have to rework this data structure?

[Dino]

What you are saying about field "val" makes absolutely no sense to me, because you have field "val" defined as a "void *" but are talking about using it as an int. And, it's not clear what "we default val to default" means. Does this mean you put a default value of some sort in field "default", or, or you saying the field "default" gets its default value from field "val"?

Regardless, why not use a union? Declare a union of either an int or a float, and then use

myunion * val ;

[Loctan]

What you are saying about field "val" makes absolutely no sense to me, because you have field "val" defined as a "void *" but are talking about using it as an int. And, it's not clear what "we default val to default" means. Does this mean you put a default value of some sort in field "default", or, or you saying the field "default" gets its default value from field "val"?

Regardless, why not use a union? Declare a union of either an int or a float, and then use

myunion * val ;

"val" is just a pointer to some variable that we either populate with "source" if it's valid or populate with "default" if "source" is not valid. The variable pointed to by "val" is sometimes a float and other times an int.

Sorry for the confusion.

[Dino]

Thanks. Then my suggestion should work fine, and no casting will be required.