coleman@bert.dg.com (Kim Coleman) (03/29/89)
// Test that the compiler catches uninitialized references. Given that
// references are conceptually an object rename, they must be initialized
// to something (see Stroustrap, p. 56). At best, g++ gives a warning,
// at worst, it says nothing at all.
void some_function()
{
int& a_ref; // g++ gives a warning
}
class erroneous {
int& a_ref; // g++ says nothing
};
Personally, I think both these cases should be an error, but a warning is
better than nothing. Sorry if this has been posted before.
---------------------------
Kim Coleman
Data General Corp., Research Triangle Park, NC
{the world}!mcnc!rti!dg-rtp!colemantiemann@YAHI.STANFORD.EDU (Michael Tiemann) (04/08/89)
Date: 28 Mar 89 17:47:34 GMT
From: rti!xyzzy!bert!coleman@mcnc.org (Kim Coleman)
Organization: Data General Corporation, Research Triangle Park, NC.
Sender: bug-g++-request@prep.ai.mit.edu
// Test that the compiler catches uninitialized references. Given that
// references are conceptually an object rename, they must be initialized
// to something (see Stroustrap, p. 56). At best, g++ gives a warning,
// at worst, it says nothing at all.
void some_function()
{
int& a_ref; // g++ gives a warning
}
It should give an error here.
class erroneous {
int& a_ref; // g++ says nothing
};
This is no cause for alarm. If you try to use this class without
specifying an initialzer for the reference, they you should get an
error. But here is how you won't run aground:
int savior;
erroneous not_so_erroneous = { savior };
In this case, savior is specified as an initializer for the reference
in not_so_erroneous.
Personally, I think both these cases should be an error, but a warning is
better than nothing. Sorry if this has been posted before.
Error for references are better than warnings, but the compiler should
not be too eager to slap you.
By the way, in working out this example, I found that it does not work
if not_so_erroneous is in global scope. I have therefore fixed it for
the next release.
---------------------------
Kim Coleman
Data General Corp., Research Triangle Park, NC
{the world}!mcnc!rti!dg-rtp!coleman