Re: socket communication: send & receive doesn't work right

That sounds great! But can I actually send and receive strings? I thought
that I only can send and receive bytes. How do I send and receive strings?

Thanks in advance for your answer!


"Michael K. O'Neill" wrote:

You can eliminate all issues of endian-ness by converting the doubles to
strings (sprintf-style) and sending the string. On the receiving side, use
an atof() function to convert back.

Note that this approach also resolves all issues concerning whether the
sending and receiving side are both using the same implementation of
IEEE-754 for storing double precision numbers. Since string equivalents are
sent, the precise implementation simply doesn't matter.

Finally, it's often easier to debug, since you can "see" the number that's
being sent.

Mike

"Scherbina Vladimir" <v_scherbina@xxxxxxxxxxxxxxx> wrote in message
news:EB69C218-6DAE-4776-ABC0-F601C7878116@xxxxxxxxxxxxxxxx
In Windows everything is little endian,
http://support.microsoft.com/kb/q102025/. I suggest you to send simple
data
from java program, and analyze what you obtain on c++ side. For example,
send an integer value 10, and tell us here what you recieve in C++
application.

--
--Vladimir, Windows SDK MVP
"Ananya" <Ananya@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:7E0E8354-A53A-4B8F-A0B4-33A51785FEE4@xxxxxxxxxxxxxxxx
Well, I tried not to reverse bytes by saying:
for (i = 0; i < j; i++)
{
ptr[i] = result[i];
}

instead of:
for (i = 0; i < num; i++)
{
for (j=0; j<sizeof(double); j++)
{
ptr[i*sizeof(double)+j] = (char)result[(i+1)*sizeof(double)-j-1];
}
}
in my C++ receiving method.

But now the doubles:
1.23 & 4.5
which I send from my Java program always become:
1.8584604523406555e+038 & 5.910042899492e-318#DEN
in my C++ program.

Please help! By the way, how can I test if my server needs reversing
bytes
or not?

"Scherbina Vladimir" wrote:

I am saying that if your server is big endian, then you _do_not_ need
to
reverse bytes.

--
--Vladimir, Windows SDK MVP
"Ananya" <Ananya@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:8B29C2D6-060A-41BA-8180-F4AA146F49A6@xxxxxxxxxxxxxxxx
Thanks! Are you saying that in my C++ receiving method the code for
reversing the byte order is incorrect? Where can I find the correct
code?


"Scherbina Vladimir" wrote:

The problem with this approach is that Java stores the binary data
as
big
endians only (no matter what CPU architecture is), if your C++
client
is
litttle endian then unpredictable results may be obtained. Check
this
issue.

--
--Vladimir, Windows SDK MVP
"Ananya" <Ananya@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote in message
news:2951DED5-4DDF-40BF-A581-1A13A95EA631@xxxxxxxxxxxxxxxx
I am trying to establish socket communication between my Java and
C++
program.

I called my Java program from my C++ program with ShellExecuteEx.
I created a C++ Server and a Java Client, which is accepted by the
Server.

I did a test of sending two doubles:
1.23 & 4.5
from my Java program to my C++ program, however I always received
the
following 2 different doubles:
1.1648250968930678e-302 & -6.4627233651951511e-086.

Here is my Java sending method:
public void send_doubles(double vals[], int len) throws
IOException
{
// convert our array of doubles into an array of bytes
ByteArrayOutputStream bytestream;
bytestream = new ByteArrayOutputStream(len*8);

DataOutputStream out;
out = new DataOutputStream(bytestream);

for (int i=0; i<len; i++)
{
out.writeDouble(vals[i]);
}

output.write(bytestream.toByteArray(), 0, bytestream.size());
output.flush();

recv_ack();
send_ack();
}

and my Java acknowledgement methods:
// send a short acknowledgement to the server
private void send_ack() throws IOException
{
int ack;

ack = 0;

output.write(ack);
output.flush();
}

// recv a short acknowledgment from the server
private void recv_ack() throws IOException
{
int ack;

ack = (int)input.read();
}

And here is my C++ receiving method:
int Server::recv_doubles(double *val, int maxlen) throw (string)
{
int i, j;
int numbytes = 0;
int end = 0;
int total_bytes = 0;
char *temp;
char *result;

temp = (char *)buffer;
result = (char *)buffer2;

j = 0;

// we are receiving the incoming doubles one byte at a time
while (!end)
{
if ((numbytes=recv(new_fd, temp, BUFFSIZE, 0))==-1)
{
throw string("help!");
}

for (i=0; i<numbytes; i++)
{
result[j] = temp[i];
j++;
}

total_bytes = total_bytes + numbytes;
if (total_bytes==maxlen*sizeof(double) + 1)
{
end = 1;
}
}

// now we need to put the array of bytes into the array of doubles
char *ptr;
int num = (j - 1)/sizeof(double);

ptr = (char *)val;

// going from Java to C++, we need to reverse the order of each
set
of
bytes
for (i = 0; i < num; i++)
{
for (j=0; j<sizeof(double); j++)
{
ptr[i*sizeof(double)+j] = (char)result[(i+1)*sizeof(double)-j-1];
}
}

send_ack();
recv_ack();

return num;
}

and my C++ acknowledgement methods:
// receive a short acknowledgement from the client
void Server::recv_ack()
{
char temp[1];
int total = 0;

while (total<1)
{
total += recv(new_fd, temp, 1, 0);
}
}

// send a short acknowledgement to the client
void Server::send_ack()
{
char temp[1];
temp[0] = 42;

send(new_fd, temp, 1, 0);
}

Why does my C++ program receive incorrect doubles?

Thanks for looking at my code!








.

Relevant Pages

  • Re: socket communication: send & receive doesnt work right
    ... Maybe start from sending simple byte array to check what happen with byte ... I did a test of sending two doubles: ... public void send_doubles(double vals, int len) throws IOException ... char *result; ...
    (microsoft.public.win32.programmer.networks)
  • Re: socket communication: send & receive doesnt work right
    ... I did a test of sending two doubles: ... public void send_doubles(double vals, int len) throws IOException ... And here is my C++ receiving method: ... char *result; ...
    (microsoft.public.win32.programmer.networks)
  • Re: socket communication: send & receive doesnt work right
    ... You can eliminate all issues of endian-ness by converting the doubles to ... On the receiving side, use ... public void send_doubles(double vals, int len) throws ... char *result; ...
    (microsoft.public.win32.programmer.networks)
  • Re: [PATCH] markers: modpost
    ... by merging the name and format strings. ... as long as the marker structs are always defined by this macro. ... +static void read_markers(const char *fname) ... int main ...
    (Linux-Kernel)
  • Re: [PATCH] markers: modpost
    ... by merging the name and format strings. ... as long as the marker structs are always defined by this macro. ... +static void read_markers(const char *fname) ... int main ...
    (Linux-Kernel)