Thread: Of pointer typecasting and other matters...

Dear Salem,

At risk of abusing your generosity, I'm going to pester you with another question. Well actually, two.

The first one is a request for a good C book recommendation. Something not geared toward beginners, and that does not shy away from explaining the low level stuff and details of what is going on behind the scenes.
I own K&R's The C Programming language and although I admit I haven't read it from cover to cover but rather use it as a reference manual, consulting it when I want to learn about a topic, I find it's depth lacking in several subjects. One example being the topic of my other thread, variadic arguments. I had to learn about this by using the internet.

The other question relates to typecasting. This is one of those subjects I'd like to study in detail to really understand what is going on at a deeper level. Specifically pointer typecasting.
I can do this just fine:

Code:

void *vptr;

vptr = (int*)45;

Yet In other circumstances, I get a warning about a cast to pointer from integer of different size, which I "solve" by casting to long. Why not here? What is different in this particular case?
More importantly, why the above works and yet this doesn't:

Code:

Code:

void *vptr;

vptr = (float*)4.5f; // error: cannot convert to a pointer type

Finally this is OK:

Code:

void *ptr;
int    *iptr;
float *fptr;

*iptr = 45;
*fptr = 4.5f;
vptr = iptr;
vptr = fptr;

K&R is an excellent reference manual for the language itself. I trust you have a copy with a big red "ANSI-C" rubber stamp logo across it.

But K&R only goes as far as C90, there is also C99 and C11 to contend with (a new standard appears about once a decade).
C11 (C standard revision) - Wikipedia

N869 is the final draft of the ISO-C99 standard (being a draft version, it's available for $0). It's great for finding out what is really guaranteed by the language and not just what your current compiler will let you get away with. If you want to write code that has the best chance of compiling on any compiler (and doing what you want), it's all good to know stuff.

As for another book to read, then perhaps C Unleashed.

comp.lang.c Frequently Asked Questions is also worth a read through.

> I own K&R's The C Programming language and although I admit I haven't read it from cover to cover
My favourite exercise then is for you to read all the references to recursion then



> I can do this just fine:
> vptr = (int*)45;
Because 45 is a literal constant which the compiler will automatically widen and convert to a pointer.

Whereas vptr = (int*)f; involves converting one stored type into another, and gives the compiler more opportunity to gripe about things which seem wrong.

> vptr = (float*)4.5f; // error: cannot convert to a pointer type
Well no, it wouldn't. You can't have fractional addresses in memory.

> Finally this is OK:
The round trip assignment of void* = type*, then at some later time type* = void* is guaranteed by the standard.

What will cause you grief at some point is
iptr = fptr;
and then expecting to be able to do something meaningful with iptr.