9 replies to this topic

[hsine]

client code like this :

 

</p><div>   typedef struct _T{</div>
<div>        int id;</div>
<div>        char* testdata;</div>
<div>    }T, *PT;</div>
<div>    PT pt;</div>
<div>    memset(pt,0,sizeof(&pt));</div>
<div>    pt->id=1;</div>
<div>    pt->testdata="test datya";</div>
<div>    char pack[128];</div>
<div>    memcpy(&pack,pt,sizeof(&pt));    ////////////////////////////  pack got nothing？？？</div>
<div>    ENetPacket *packet=enet_packet_create((const void*)&pack,128,ENET_PACKET_FLAG_RELIABLE);</div>
<div>    if(NULL == packet){</div>
<div>        fprintf(stderr,"create packet with error\n");</div>
<div>        exit(EXIT_FAILURE);</div>
<div>    }</div>
<div> </div>
<div>    enet_peer_send(peer,0,packet);</div>
<div> </div>
<div>    enet_host_flush(client);</div>
<div>

[Bacterius]

Your use of pointers is wrong. "pack" is already a pointer, being an array. Also, writing "PT pt" and later on "pt->id = 1" indicates you typedef'd a pointer or array type to a non-pointer type, which is bad because it makes it hard to see what is a pointer and what isn't. I strongly recommend you rewrite the code without writing any * (asterisk) in your typedef's. Then, fix the code to properly pass to memcpy and memset what they need. In any case, assuming there are no other bugs, your immediate fix is:

 

 

 

memcpy(pack,pt,sizeof(&pt));

 

 

But the code is not easy to follow for the reason I mentioned above, and you should rewrite it if you wish to keep your sanity.

[RulerOfNothing]

Actually, sizeof(&pt) takes the size of a pointer, so that code will only copy 4 or 8 bytes depending on the platform. You want memcpy(pack,pt,sizeof(*pt)); or memcpy(pack,pt,sizeof(T));

[hsine]

after i  fix the code ,but  i still can't send struct to remote server ? why ?

[rip-off]

First note, the intermediate buffer is unnecessary. You could just pass the address of the structure to enet to initialise the packet (obviously adjusting the size parameter you are passing too.

 

However, your structure contains a pointer. A pointer sent over the network is meaningless. One option is to change to a fixed length array, something like this:

const int MaxPacketStringLength = 100;
 
</p><div>struct Packet {</div>
<div>        int id;</div>
<div>        char data[MaxPacketStringLength];</div>
<div>};</div>
 

This has two major drawbacks, a naive client will send too much data for short strings, and it also limits the maximum size of the data.

 

Another way is to serialise the structure into a contiguous byte array. This is possibly what a ENetPacket is for.  Enet might provide functions to help you do this. Another point is that you need to also handle endianness. Again, ENet might provide helpers, or simply expose the usual hton*() and ntoh*() functions.

[rip-off]

Sometimes it can help if you post your current code after making changes. As I mentioned before, use [code] tags , or the WYSIWYG functionality in the editor to insert pretty code blocks, like this:

[code]

int main() {

    printf("Hello, world");

}

[/code]

[Servant of the Lord]

My experience is limited in this area, but wouldn't memcpy()ing a struct into bytes cause problems because of alignment? Couldn't the struct by packed different (variables in the same order, but extra padding between them for performance purposes) on the server and the client, possibly writing and reading too little or too many bytes?

 

Is the packing of a struct standardized by C++ for POD types?

[rip-off]

Good point, packing must be taken into consideration too. I actually meant memberwise serialisation of the structure, with "smart" handling of dynamically sized members such as strings and containers.

 

Is the packing of a struct standardized by C++ for POD types?

The OP appears to be using C, (potentially via a C++ compiler). That said, I believe neither language guarantees any sort of standard packing behaviour.

Edited by rip-off, 11 January 2013 - 04:05 AM.

[Ravyne]

Packing and also endianess can be a problem. Explicit serialization of some kind is the most robust solution, IMO.

[Zaoshi Kaba]

Either I'm retarded when it comes to this post, or code in OP is wrong on way too many levels:

line 05.: variable is a pointer pointing to invalid memory address;

line 06.: memset'ing inaccessible memory and invalid size (or at least illogical);

line 07.: dereferencing invalid pointer;

line 08.: dereferencing invalid pointer and you cannot assign string like that, did this code even compile?;
line 10.: memcpy will change pointer (the one on the stack), that's not what OP wants, also size is incorrect;

line 11.: sending invalid data (stack pointer) and line 10 didn't actually copy any string, again not what OP expects;

 

I hope I'm not wrong anywhere, seems way too fishy.