[wx-dev] (64 bit) hashmap/long long problem in HEAD and 2.7.2

[Vadim Zeitlin]

On Wed, 1 Nov 2006 14:21:02 +0000 Michael Wetherell <mike.wetherell at ntlworld.com> wrote:

MW> >  So I do not think this is a problem.
MW> 
MW> Yes it is the problem, long is 64-bit on that OS.

 But so what? As I wrote, wxLongLong_t is still a separate type so we
should allow defining hashes with wxLongLong_t type keys, shouldn't we?
Even if they're the same as long on this platform, they could be different
on another one.

MW> The function longlongHash is an implementation helper for the 
MW> wxLongLong_t overloads of operator(), which should only be defined when 
MW> long is not long long.

 Why? I.e. I though these operator()s had to be defined for all types which
can be used as hash keys.

MW> > but it is 8 under Windows. So I guess it's 
MW> > better to use wxUint32 explicitly here even though it's still wrong
MW> > for sizeof(size_t) == 8.
MW> 
MW> No, it's only used when long is 4 bytes anyhow.

 Supposing we define it when sizeof(long) is 8, of course.

MW> >  Anyhow, please let me know if you still have any problems with the
MW> > revision 1.61 of wx/hashmap.h.
MW> 
MW> Suggest you revert, why change it when it's been well tested by the 
MW> HashesTestCase though all of 2.6.x?

 It doesn't test wxLongLong_t keys in the hashes though if wxLongLongIsLong
is defined. Could it be because it doesn't work? And, if so, once again,
shouldn't it work?

 Thanks,
VZ