I have two structures as below
When I used the above structures as a function argument as belowCode:struct prefix { u_char a; u_char b; u_char c; u_char d; u_char e[4]; }; struct in_addr { union{ u_int8_t addr8[16]; u_int16_t addr16[8]; u_int32_t addr32[4]; } in6; }
.....
.....
log(.......,(struct in_addr*)p->addr32,......); // where p is a pointer to structure prefix
.....
.....
at that point, I am getting the warning as "cast increases required alignment of target type".
Please help me
Googled for solution but very confused among "_attribute_ ((packed));" and "_attribute_ ((alligned) (x)) ;", which one to use and why?
I suspect you're asking the wrong question.
A "struct prefix" is significantly smaller than a "struct in_addr", any attempt to use a struct prefix as a struct in_addr will almost certainly tromp areas of process memory that you don't intend.
On the face of it, it appears you introduced that typecast to stop the compiler complaining about something else (eg you are trying to supply the address of a "struct prefix" to a function that expects a pointer to "struct addr32", and you've tried to use a typecast to force the compiler to permit that). Even if you do get such code to compile, all that will probably happen is that your program will crash when it is executed.
Rather than trying to hack code, and then beat the compiler into submission with typecasts and pragmas or other non-standard settings to stop it complaining, it would be beneficial if you tried writing clean code.
My question is correct grumpy. Because same kind of questions are there when googled, and they said about using "_attribute_ ((packed)); to the structure. But the problem here is the structure in_addr belongs to kernel. so, dont want to change that. and moreover the function "log" accepts variable arguments. And the argument I have shown you is one of the variable arguments I am passing.
Where is your prototype for your log function?
Alignment cast warnings work like this
Now whilst v has enough bytes to store a double, it may be aligned on a byte boundary, and not a double boundary.Code:void foo ( double *p ) { *p = 0; } int main ( ) { unsigned char v[sizeof(double)]; foo( (double*)v ); }
As such, when you try to dereference the pointer in foo, it will be an invalid address and the program will just crash.
The correct way to convert a struct prefix into a struct in_addr would be to do something like
Code:struct in_addr temp = { 0 }; memcpy( temp.in6.addr32, p, sizeof(*p) ); log(..., temp.in6.addr32 ... );
If you dance barefoot on the broken glass of undefined behaviour, you've got to expect the occasional cut.
If at first you don't succeed, try writing your phone number on the exam paper.
Guys, Thanks a lot for your replies.
could you please look in to this link which is most similart to my problem and requesting you to respond
Warning "cast increases required alignment of target type" - EmbDev.net
The fact of others having asked similar questions doesn't mean your question is the right one.Originally Posted by karthik537
Lemmings follow, regardless of the sense of following.
Yes, but if you use a typecast to convince the compiler to permit passing a different data type to a function than the function expects, it is your responsibility to ensure that what you pass is acceptable to the called function. One aspect of compatibility is that the data structure you pass cannot be smaller than what the function expects. And that is exactly what you are doing.
The fact you can bludgeon the compiler into accepting that does not make the data you pass suddenly valid. However, such tricks do not make invalid data become valid - they simply stop the compiler from complaining about the fact you are passing invalid data. The question you need to be asking is how to ensure the data you pass is valid, not about how to force the compiler to stop complaining about passing invalid data.
