rfg@MCC.COM (Ron Guilmette) (03/13/89)
When compiled with the G++ 1.34.0/Sun3 compiler, the following short
code gets no errors. It should elicit a compile-time error on the
indicated line. Note that an error *is* produced if the declaration
of "c_object_1" is removed from the code.
The fact that no error is issued causes further problems at run-time.
Specifically, execution of test() will result in two destructor calls,
but only one call to the constructor for the type "c".
/*
Description - check that "it is not possible to transfer control past a
declaration with an (implicit or explicit) initializer"
as called for by the "manual", section 9.11, page 296.
*/
class c {
public:
c (int i);
~c ();
};
void test ()
{
c c_object_1(1);
goto lbl; // ERROR
c c_object_2 (2);
lbl: ;
}
// Ron Guilmette - MCC - Experimental (parallel) Systems Kit Project
// 3500 West Balcones Center Drive, Austin, TX 78759 - (512)338-3740
// ARPA: rfg@mcc.com
// UUCP: {rutgers,uunet,gatech,ames,pyramid}!cs.utexas.edu!pp!rfgrfg@MCC.COM (Ron Guilmette) (03/13/89)
This error is related to (but different from) the last one I posted
regarding goto's jumping around object declarations.
In this case, G++ 1.34.0/Sun3 fails to issue an error when an
"unstructured" switch statement causes an apparently bogus transfer
of control past a declaration/initialization.
If an error message is ever issued for this condition, it seems like
it should be issued for the line indicated below.
/*
Description - check that "it is not possible to transfer control past a
declaration with an (implicit or explicit) initializer"
as called for by the "manual", section 9.11, page 296.
This test checks that legal but "unstructured" forms of
switch statements are not allowed to circumvent the transfer
rules for goto statements.
*/
class c {
public:
c (int i);
~c ();
};
void test ()
{
switch (1) {
case 0:
{
c c_object (1);
case 1: // ERROR - case in wrong binding contour
break;
}
}
}