----- Original Message ----- From: Peter Dimov To: Sent: Friday, 14 December, 2001 2:33 PM Subject: Re: [Boost-Users] Why not "if (px != rpx) " ?

From: "Louis Lavery"

...

Hello,

The following is use by the boost shared_ptr assignment operators (see boost\smart_ptr.hpp)...

void share(T* rpx, long* rpn) { if (pn != rpn) { // Q: why not px != rpx? A: fails when both == 0 ++*rpn; // done before dispose() in case rpn transitively // dependent on *this (bug reported by Ken Johnson) dispose(); px = rpx; pn = rpn; } } // share

...I curious to know what it means by "fails when both == 0".

It's possible that px == rpx == 0, yet pn != rpn.

-- Peter Dimov Multi Media Ltd.

Hello, Yes, I see that, but I can't see what's wrong with using the test "if ( px != rpx )". Or, to put it another way, if px == rpx == 0 who cares whether pn equals rpn or not? Louis.

On 12/14/01 1:40 PM, "Louis Lavery" wrote:

Yes, I see that, but I can't see what's wrong with using the test "if ( px != rpx )".

Or, to put it another way, if px == rpx == 0 who cares whether pn equals rpn or not?

Oops, my last message was a mistake. I don't know the answer. -- Darin

From: "Louis Lavery"

From: Peter Dimov

...

From: "Louis Lavery"

...

void share(T* rpx, long* rpn) { if (pn != rpn) { // Q: why not px != rpx? A: fails when both ==

0

...

++*rpn; // done before dispose() in case rpn transitively // dependent on *this (bug reported by Ken Johnson) dispose(); px = rpx; pn = rpn; } } // share

...I curious to know what it means by "fails when both == 0".

It's possible that px == rpx == 0, yet pn != rpn.

Yes, I see that, but I can't see what's wrong with using the test "if ( px != rpx )".

Or, to put it another way, if px == rpx == 0 who cares whether pn equals rpn or not?

Two shared_ptr's are either copies of each other (pn == rpn) or they are not (pn != rpn). It is possible for two shared_ptr's to be both zero, yet to have different counts: shared_ptr<int> a; shared_ptr<int> b; // b.use_count() == 1 b = a; // b.use_count() == 2 It's also possible to have two NULL shared_ptr's that have the same count: shared_ptr<int> a; shared_ptr<int> b(a); // b.use_count() == 2 b = a; // b.use_count() == 2 If you change the test to px != rpx, the first example will leave b with the same count. This is not correct since after the assignment a and b are guaranteed to share the same count (and hence use_count() must be at least 2.) I realize that this is a bit confusing; think about it this way. The test is an "optimization." The results of share() must be the same even without the test. -- Peter Dimov Multi Media Ltd.