Kay Ramme - Sun Germany - Hamburg wrote:
Hi,
please correct me, if I am wrong. I understand this as a 'C' inherited
C++ oddity (no constructors for integral values), which leads to a
warning if the first operation on the integral value is not classified
as "assignment". Obviously 'operator >>=' has not been classified as
"assignment" at least not for the right operand.
No, the compiler does not assume the user-supplied operator >>= has any
assignment semantics. Rather, as the operator >>= is inline, GCC tries
an analysis whether or not b is definitely assigned at point (1) in
com::sun::star::uno::Any a;
T b;
if (a >>= b) {
// (1)
}
Depending on T (i.e., depending on the complexity of the relevant inline
operator >>=):
- GCC sometimes comes up with the insight that b is definitely assigned
at (1), which is correct, and in which case GCC does not issue a warning
(why should it?);
- GCC sometimes cannot decide whether or not b is definitely assigned at
(1), although it *is*, and in which case GCC conservatively does not
issue a warning;
- GCC sometimes comes up with the insight that b is not definitely
assigned at (1), which is an incorrect analysis, and in which case GCC
erroneously issues a warning.
So, it seems that we have used the wrong operator here. Therefor I tend
to agree to Frank, that we may want to fix this.
The choice of operator is indeed unfortunate. However, I do not agree that
- T b;
+ T b = T();
is in general a fix that improves code quality.
-Stephan
Kay
Frank Schönheit - Sun Microsystems Germany wrote:
Nonetheless, I would not consider initializing b with something
meaningful a mutilation. And be it just because months later, somebody
will be tempted to re-use b some lines below.
Ciao
Frank
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
