history.cgi is vulnerable to a buffer overflow due to the use of
sprintf with user supplied data that has not been restricted in size.
This vulnerability does not appear to be exploitable on the majority
of systems (due to stack cookies, the NX bit, etc).

In the process_cgivars function:
                /* do some basic length checking on the variable
identifier to prevent buffer overflows */
                if(strlen(variables[x]) >= MAX_INPUT_BUFFER - 1)
                        continue;

The above code is only applied to the parameter names, not their
values which allows the host_name and svc_description variables to be
of any length.

Later, in the get_history function:
                                if(history_type == HOST_HISTORY ||
history_type == SERVICE_HISTORY) {
                                        sprintf(match1, " HOST ALERT:
%s;", host_name);
                                        sprintf(match2, " SERVICE
ALERT: %s;", host_name);
                                        }

match1 and match2 are each MAX_INPUT_BUFFER (1024) bytes long. Similar
lines appear later in the same function. These should be changed to
snprintf calls (even host_name were restricted in size, these calls
could still result in an overflow).

The above code is only run under certain circumstances, eg when there
is a "SERVICE ALERT" line (for any host) in the log that is being
processed.

An example url that results in an overflow (and segfault):
http://nagiosserver/nagios/cgi-bin/history.cgi?host=aaaaaaa... (4000 'a's)

When the cgi is run from the command line (with setting environment
variables for the parameters), gdb shows:
Program received signal SIGSEGV, Segmentation fault.
0x61616161 in ?? ()

The above was tested on Slackware 14.0 i386.