1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
|
<!-- $Id$ -->
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="GENERATOR" CONTENT="Mozilla/4.04 [en] (X11; I; Linux 2.0.32 i486) [Netscape]">
<META NAME="Author" CONTENT="Billy Quinn">
<META NAME="Description" CONTENT="A first step towards using ACE productively">
<TITLE>ACE Tutorial 005</TITLE>
</HEAD>
<BODY TEXT="#000000" BGCOLOR="#FFFFFF" LINK="#000FFF" VLINK="#FF0F0F">
<CENTER><B><FONT SIZE=+2>ACE Tutorial 005</FONT></B></CENTER>
<CENTER><B><FONT SIZE=+2>On the road to a multithreaded server</FONT></B></CENTER>
<P>
<HR WIDTH="100%">
<P>Now we're finally at <I><A HREF="client_handler.cpp">client_handler.cpp</A></I>
where we have to write some code. This file has more code than the
rest of the application all together.
<P>
<HR WIDTH="100%">
<PRE>
<font color=red>// $Id$</font>
<font color=red>/* In client_handler.h I alluded to the fact that we'll mess around
with a Client_Acceptor pointer. To do so, we need the
Client_Acceptor object declaration.
We know that including client_handler.h is redundant because
client_acceptor.h includes it. Still, the sentry prevents
double-inclusion from causing problems and it's sometimes good to
be explicit about what we're using.
On the other hand, we don't directly include any ACE header files
here. */</font>
<font color=blue>#include</font> "<font color=green>client_acceptor.h</font>"
<font color=blue>#include</font> "<font color=green>client_handler.h</font>"
<font color=red>/* Our constructor doesn't do anything. That's generally a good idea.
Unless you want to start throwing exceptions, there isn't a really
good way to indicate that a constructor has failed. If I had my
way, I'd have a boolean return code from it that would cause new to
return 0 if I failed. Oh well... */</font>
<font color=#008888>Client_Handler::Client_Handler</font> (void)
{
}
<font color=red>/* Our destructor doesn't do anything either. That is also by design.
Remember, we really want folks to use destroy() to get rid of us.
If that's so, then there's nothing left to do when the destructor
gets invoked. */</font>
<font color=#008888>Client_Handler::~Client_Handler</font> (void)
{
<font color=red>// Make sure that our peer closes when we're deleted. This</font>
<font color=red>// will probably happened when the peer is deleted but it</font>
<font color=red>// doesn't hurt to be explicit.</font>
this->peer ().close ();
}
<font color=red>/* The much talked about destroy() method! The reason I keep going on
about this is because it's just a Bad Idea (TM) to do real work
inside of a destructor. Although this method is void, it really
should return int so that it can tell the caller there was a
problem. Even as void you could at least throw an exception which
you would never want to do in a destructor. */</font>
void
<font color=#008888>Client_Handler::destroy</font> (void)
{
<font color=red>/* Tell the reactor to forget all about us. Notice that we use the
same args here that we use in the open() method to register
ourselves. In addition, we use the DONT_CALL flag to prevent
handle_close() being called. Since we likely got here due to
handle_close(), that could cause a bit of nasty recursion! */</font>
this->reactor ()->remove_handler (this,
ACE_Event_Handler:: READ_MASK | <font color=#008888>ACE_Event_Handler::DONT_CALL</font>);
<font color=red>/* This is how we're able to tell folks not to use delete. By
deleting our own instance, we take care of memory leaks after
ensuring that the object is shut down correctly. */</font>
delete this;
}
<font color=red>/* As mentioned before, the open() method is called by the
Client_Acceptor when a new client connection has been accepted.
The Client_Acceptor instance pointer is cast to a void* and given
to us here. We'll use that to avoid some global data... */</font>
int
<font color=#008888>Client_Handler::open</font> (void *_acceptor)
{
<font color=red>/* Convert the void* to a Client_Acceptor*. You should probably use
those fancy ACE_*_cast macros but I can never remember how/when
to do so. Since you can cast just about anything around a void*
without compiler warnings be very sure of what you're doing when
you do this kind of thing. That's where the new-style cast
operators can save you. */</font>
Client_Acceptor *acceptor = (Client_Acceptor *) _acceptor;
<font color=red>/* Our reactor reference will be set when we register ourselves but
I decided to go ahead and set it here. No good reason really... */</font>
this->reactor (acceptor->reactor ());
<font color=red>/* We need this to store the address of the client that we are now
connected to. We'll use it later to display a debug message. */</font>
ACE_INET_Addr addr;
<font color=red>/* Our ACE_Svc_Handler baseclass gives us the peer() method as a way
to access our underlying ACE_SOCK_Stream. On that object, we can
invoke the get_remote_addr() method to get an ACE_INET_Addr
having our client's address information. As with most ACE
methods, we'll get back (and return) a -1 if there was any kind
of error. Once we have the ACE_INET_Addr, we can query it to
find out the clien's host name, TCP/IP address, TCP/IP port value
and so forth. One word of warning: the get_host_name() method of
ACE_INET_Addr may return you an empty string if your name server
can't resolve it. On the other hand, get_host_addr() will always
give you the dotted-decimal string representing the TCP/IP
address. */</font>
if (this->peer ().get_remote_addr (addr) == -1)
return -1;
<font color=red>/* If we managed to get the client's address then we're connected to
a real and valid client. I suppose that in some cases, the
client may connect and disconnect so quickly that it is invalid
by the time we get here. In any case, the test above should
always be done to ensure that the connection is worth keeping.
Now, register ourselves with a reactor and tell that reactor that
we want to be notified when there is something to read.
Remember, we took our reactor value from the acceptor which
created us in the first place. Since we're exploring a
single-threaded implementation, this is the correct thing to do. */</font>
if (this->reactor ()->register_handler (this,
<font color=#008888>ACE_Event_Handler::READ_MASK</font>) == -1)
ACE_ERROR_RETURN ((LM_ERROR,
"<font color=green>(%P|%t) can't register with reactor\n</font>"),
-1);
<font color=red>/* Here, we use the ACE_INET_Addr object to print a message with the
name of the client we're connected to. Again, it is possible
that you'll get an empty string for the host name if your DNS
isn't configured correctly or if there is some other reason that
a TCP/IP addreess cannot be converted into a host name. */</font>
ACE_DEBUG ((LM_DEBUG,
"<font color=green>(%P|%t) connected with %s\n</font>",
addr.get_host_name ()));
<font color=red>/* Always return zero on success. */</font>
return 0;
}
<font color=red>/* In the open() method, we registered with the reactor and requested
to be notified when there is data to be read. When the reactor
sees that activity it will invoke this handle_input() method on us.
As I mentioned, the _handle parameter isn't useful to us but it
narrows the list of methods the reactor has to worry about and the
list of possible virtual functions we would have to override. */</font>
int
<font color=#008888>Client_Handler::handle_input</font> (ACE_HANDLE handle)
{
<font color=red>/* Some compilers don't like it when you fail to use a parameter.
This macro will keep 'em quiet for you. */</font>
ACE_UNUSED_ARG (handle);
<font color=red>/* Now, we create and initialize a buffer for receiving the data.
Since this is just a simple test app, we'll use a small buffer
size. */</font>
char buf[BUFSIZ];
<font color=red>/* Invoke the process() method with a pointer to our data area.
We'll let that method worry about interfacing with the data. You
might choose to go ahead and read the data and then pass the
result to process(). However, application logic may require that
you read a few bytes to determine what else to read... It's best
if we push that all into the application-logic level. */</font>
return this->process (buf, sizeof (buf));
}
<font color=red>/* If we return -1 out of handle_input() or if the reactor sees other
problems with us then handle_close() will be called. The reactor
framework will take care of removing us (due to the -1), so we
don't need to use the destroy() method. Instead, we just delete
ourselves directly. */</font>
int
<font color=#008888>Client_Handler::handle_close</font> (ACE_HANDLE handle,
ACE_Reactor_Mask mask)
{
ACE_UNUSED_ARG (handle);
ACE_UNUSED_ARG (mask);
delete this;
return 0;
}
<font color=red>/* And, at last, we get to the application-logic level. Out of
everything we've done so far, this is the only thing that really
has anything to do with what your application will do. In this
method we will read and process the client's data. In a real
appliation, you will probably have a bit more in main() to deal
with command line options but after that point, all of the action
takes place here. */</font>
int
<font color=#008888>Client_Handler::process</font> (char *rdbuf,
int rdbuf_len)
{
ssize_t bytes_read = -1;
<font color=red>/* Using the buffer provided for us, we read the data from the
client. If there is a read error (eg -- recv() returns -1) then
it's a pretty good bet that the connection is gone. Likewise, if
we read zero bytes then something wrong has happened. The
reactor wouldn't have called us if there wasn't some kind of read
activity but there wouldn't be activity if there were no bytes to
read...
On the other hand, if we got some data then we can display it in
a debug message for everyone to see. */</font>
switch ( (bytes_read = this->peer ().recv (rdbuf, rdbuf_len)) )
{
case -1: <font color=red>// Complain and leave</font>
ACE_ERROR_RETURN ((LM_ERROR,
"<font color=green>(%P|%t) %p bad read\n</font>",
"<font color=green>client</font>"),
-1);
case 0: <font color=red>// Complain and leave</font>
ACE_ERROR_RETURN ((LM_ERROR,
"<font color=green>(%P|%t) closing daemon (fd = %d)\n</font>",
this->get_handle ()),
-1);
default: <font color=red>// Show the data</font>
<font color=red>// NULL-terminate the string before printing it.</font>
rdbuf[bytes_read] = 0;
ACE_DEBUG ((LM_DEBUG,
"<font color=green>(%P|%t) from client: %s</font>",
rdbuf));
}
<font color=red>/* It's also worth mentioning that recv() has a cousin: recv_n().
recv_n() will receive exactly the number of bytes you provide it.
This is very good when you know exactly how much you expect to
receive. For the application here, unfortunately, we don't have
any idea how much the client will be sending. recv() will read
up-to-but-not-more-than the number of bytes we specify (e.g. --
_rdbuf_len). That works well when we don't know how much the
client will provide. */</font>
return 0;
}
</PRE>
<P><HR WIDTH="100%">
<CENTER>[<A HREF="../online-tutorials.html">Tutorial Index</A>] [<A HREF="page06.html">Continue This Tutorial</A>]</CENTER>
|
