A few questions about std.algorithm.swap() that came up when I was
reading the code to fix Bug 4789:

1.  Do we really need to use memcpy() for structs w/o a postblit?  I can't see any advantages and it seems like it might be less efficient, since for a struct without a postblit just using plain old assignment should be about as efficient as it gets.

2.  memcpy() isn't guaranteed to work for overlapping memory blocks. Shouldn't there be a check in to make sure a and b don't have the same address?

This has been lying in my inbox for a good while. I see now the bug is fixed. Any trailing questions regarding it?

Andrei

On 9/2/10 10:58 PM, David Simcha wrote:
> A few questions about std.algorithm.swap() that came up when I was
> reading the code to fix Bug 4789:
>
> 1. Do we really need to use memcpy() for structs w/o a postblit? I can't see any advantages and it seems like it might be less efficient, since for a struct without a postblit just using plain old assignment should be about as efficient as it gets.
>
> 2. memcpy() isn't guaranteed to work for overlapping memory blocks.
> Shouldn't there be a check in to make sure a and b don't have the same
> address?
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos

Well, bug 4789 is fixed for static arrays but reintroduced for structs
with elaborate assign.
Array copy is not allowed for overlapping memory.

           ubyte[T.sizeof] t = void;
           auto a = (cast(ubyte*) &lhs)[0 .. T.sizeof];
           auto b = (cast(ubyte*) &rhs)[0 .. T.sizeof];
           t[] = a[];
           a[] = b[]; <-- will raise exception
           b[] = t[];

Martin

On Sun, 02 Jan 2011 00:42:44 +0100, Andrei Alexandrescu <andrei at erdani.com> wrote:

> This has been lying in my inbox for a good while. I see now the bug is fixed. Any trailing questions regarding it?
>
> Andrei
>
> On 9/2/10 10:58 PM, David Simcha wrote:
>> A few questions about std.algorithm.swap() that came up when I was
>> reading the code to fix Bug 4789:
>>
>> 1. Do we really need to use memcpy() for structs w/o a postblit? I can't see any advantages and it seems like it might be less efficient, since for a struct without a postblit just using plain old assignment should be about as efficient as it gets.
>>
>> 2. memcpy() isn't guaranteed to work for overlapping memory blocks.
>> Shouldn't there be a check in to make sure a and b don't have the same
>> address?
>> _______________________________________________
>> phobos mailing list
>> phobos at puremagic.com
>> http://lists.puremagic.com/mailman/listinfo/phobos

I'm not sure I understand where the brokenness is. Could you please post to http://d.puremagic.com/issues/show_bug.cgi?id=4789 an example that does break even after David Simcha's fix (http://dsource.org/projects/phobos/changeset/1948)?

Andrei

On 1/2/11 9:19 AM, Martin Nowak wrote:
> Well, bug 4789 is fixed for static arrays but reintroduced for structs
> with elaborate assign.
> Array copy is not allowed for overlapping memory.
>
> ubyte[T.sizeof] t = void;
> auto a = (cast(ubyte*) &lhs)[0 .. T.sizeof];
> auto b = (cast(ubyte*) &rhs)[0 .. T.sizeof];
> t[] = a[];
> a[] = b[]; <-- will raise exception
> b[] = t[];
>
> Martin
>
> On Sun, 02 Jan 2011 00:42:44 +0100, Andrei Alexandrescu <andrei at erdani.com> wrote:
>
>> This has been lying in my inbox for a good while. I see now the bug is fixed. Any trailing questions regarding it?
>>
>> Andrei
>>
>> On 9/2/10 10:58 PM, David Simcha wrote:
>>> A few questions about std.algorithm.swap() that came up when I was
>>> reading the code to fix Bug 4789:
>>>
>>> 1. Do we really need to use memcpy() for structs w/o a postblit? I can't see any advantages and it seems like it might be less efficient, since for a struct without a postblit just using plain old assignment should be about as efficient as it gets.
>>>
>>> 2. memcpy() isn't guaranteed to work for overlapping memory blocks.
>>> Shouldn't there be a check in to make sure a and b don't have the same
>>> address?
>>> _______________________________________________
>>> phobos mailing list
>>> phobos at puremagic.com
>>> http://lists.puremagic.com/mailman/listinfo/phobos
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos