------- Comment #4 from ciprian dot craciun at gmail dot com 2006-01-29 09:31
-------
(In reply to comment #3)
> int t1;
> int t2;
>
> int * a = &t1;
> int const * const & b = a;
> ----
> This does something different than you think.
> It creates a temparory variable for the type, "const int *" as you need to
> first convert a to that and then to "const int * const" as "const int * const"
> is not compatiable with "int *".
>
Please scuse me to bother you, but your explanation doesn't seem right.
If I try:
int * a = 0;
int const * const b = a;
it compiles and even using the -Wall option the compiler does'nt complain.
Also if I try:
int t1;
int * a = &t1;
int const * & b = a;
it gives a compile time error... But I think the code is corect.
You said that int * is not compatible with int const * const (I know they
are compatible), but here I use int * as a int const *, which to my knowledge
are compatible (and used without & it compiles without errors).
These situations are exemplified in the file (bug-01.cpp) I have atached...
All I want to do is declare an alias to a variable, and weaken the access
to this alias. This is a valid operation. This is done in a class:
class astring { // ascii string
public :
astring (void) : length (actual_length), chars (actual_chars) {}
size_t const * length; // an alias to actual_length
char const * const & chars; // an alias to actual_chars
private :
size_t actual_length;
char * actual_chars;
};
length works fine when I update actual_length. Only chars don't get updated
when I update actual_chars.
--
ciprian dot craciun at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ciprian dot craciun at gmail
| |dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=26011
