Fix destroy
#362
Fix destroy
#362
Conversation
* breaking change: reject shared static arrays (silently worked incorrect previously) * breaking change: do not reject nested static arrays (rejected previously) * also add unittests
|
@andralex, is it what you were asking for? |
| else | ||
| buf[] = init[]; | ||
| } | ||
| static assert(!is(T == const), "`destroy` doesn't work for `const` types"); |
There was a problem hiding this comment.
This is overly conservative. We should be able to e.g. destroy a const int or a pointer or an enum value etc. because that's a no-op. So we should only disallow overwriting const/immutable data, but other than that no problem. (There is this weird corner case of destroying const objects without indirections and without mutable indirections... we should discuss that later.)
There was a problem hiding this comment.
I will not finally fix current destroy implementation (like I did with emplace) as I hate it. You can accept or reject these fixes, I will not add more fixes here.
There was a problem hiding this comment.
@andralex,
this your comment shows that you think destroy can be used in templated code. And I'm sure it is incorrect as destroy behaves differently for different types and you always have to wrap it in static if or you will get compilable program that will fail at runtime once because you didn't predict some case of destroy behavior.
I was wrong. I don't understand what is going on in your brain at all. Please, do not be hurt, but when I'm reading your posts about destroy I think that you do not see this problem at all and you even don't want to try to see this problem.
There was a problem hiding this comment.
I will not add more fixes here.
More than that, you const int proposal isn't a fix, it only increases destroy danger. The more types destroy accept the higher probability one will write incorrect templated code with destroy and will spend tons of his time for debugging.
There was a problem hiding this comment.
@andralex,
sorry again as this time it really sounds offensive, do you understand that you can't show me (and, probably, even imagine) any example of templated function using destroy for arbitrary types and still trying to make it work for more and more types?
|
It seems we're not making progress in this matter. I will close this and get to improving |
This pull fixes incorrect
destroybehavior and makesdestroyreclaim non-memory resources also for dynamic arrays as Andrei requested.