46 replies to this topic

[Servant of the Lord]

anonymous structures within an anonymous union are not legal C++. Its a visual studio extension.

Crap, I've been looking the standard up and down and I didn't find anything that would make this illegal. Though I could've looked harder. Maybe there's a difference between C11 and C++11 in this regard? But I do know that it's not just a visual studio extension; it works well on gcc and clang as well.

GCC does it just because Visual Studio does it, and people would complain if GCC was too incompatible with Microsoft's compiler.

If you have compiler warnings set to their proper level, GCC would tell you that it's non-standard, iirc (I ran into this a few months back). Though, the documentation for GCC does imply that C11 permits it.

[joew]

I should note that the only places I've actually seen people use anonymous unions/structures was usually almost always a hack that could have been avoided.

I used to use it awhile back in a matrix class but killed it off when fixing standards issues in the codebase.

[source lang="cpp"]struct Matrix44{ union { struct { float _m[16]; }; struct { float m[4][4]; }; };};[/source]

Edited by joew, 23 October 2012 - 03:15 PM.

[swiftcoder]

I should note that the only places I've actually seen people use anonymous unions/structures was usually almost always a hack that could have been avoided.

I used to use it awhile back in a matrix class but killed it off when fixing standards issues in the codebase.

If you really need to do something like this, then an array decays to a pointer. You can always store just the 2-dimensional array and cast back to the single dimension as needed.

Though I would submit that you shouldn't be handling the data in two separate fashions here, unless you are providing some sort of abstraction over both.

[ChaosEngine]

variable arguments
...
^ is awesome

No, it's bloody not.
"Basically whenever you invoke the dread ellipses construct you leave the happy world of type safety." — SiCrane

[swiftcoder]

variable arguments
...
^ is awesome

No, it's bloody not.
"Basically whenever you invoke the dread ellipses construct you leave the happy world of type safety." — SiCrane

Thus why we now need variable template arguments, to fix the deficiencies previously solved via variadic arguments

[joew]

If you really need to do something like this, then an array decays to a pointer. You can always store just the 2-dimensional array and cast back to the single dimension as needed.

I have no idea why it was like that as the code was extremely old and just never touched. I believe it was due to passing to an API but that still doesn't make any sense considering the cast is a NOP so it wouldn't even save anything.

Edited by joew, 23 October 2012 - 03:58 PM.

[Radikalizm]

Thus why we now need variable template arguments, to fix the deficiencies previously solved via variadic arguments

I was so looking forward to this feature, only to find out that visual studio 2012 would not be supporting it. I really hope microsoft implements at least the really important features of C++11 soon, they're seriously running behind gcc right now.

Edited by Radikalizm, 23 October 2012 - 04:02 PM.

[Washu]

If you really need to do something like this, then an array decays to a pointer. You can always store just the 2-dimensional array and cast back to the single dimension as needed.

I have no idea why it was like that as the code was extremely old and just never touched. I believe it was due to passing to an API but that still doesn't make any sense considering the cast is a NOP so it wouldn't even save anything.

The most common cases I've seen that in is crappy game tutorials code.

Also, I like your nickname. It's much better...now that I've changed it.

Edited by Washu, 23 October 2012 - 04:04 PM.

[joew]

Also, I like your nickname. It's much better...now that I've changed it.

rofl bah fix it!

[jwezorek]

Also, I like your nickname. It's much better...now that I've changed it.

rofl bah fix it!

Personally I liked "joew" seeing as it is the *ultimate* first name plus last initial.

[Washu]

Also, I like your nickname. It's much better...now that I've changed it.

rofl bah fix it!

Personally I liked "joew" seeing as it is the *ultimate* first name plus last initial.

Well, I was going to go for [Joe the:Objective c:Lover] but it wouldn't accept that.

[joew]

Don't make me whine to Drew!

[Washu]

Don't make me whine to Drew!

Your wish has been granted.

Edited by Washu, 23 October 2012 - 04:35 PM.

[joew]

Your wish has been granted.

To joew!

EDIT: not funny!

Edited by joew aka The Woynillowicz, 23 October 2012 - 04:39 PM.

[Washu]

Your wish has been granted.

To joew!

I'm CONFUSED?!?

[joew]

I'm CONFUSED?!?

So am I.... this change actually introduces a compile error for some wtf reason:

[source lang="cpp"]// OriginalRectangle& operator=(const Rectangle& rhs) : left(rhs.left), bottom(rhs.bottom), right(rhs.right), top(rhs.top) {}// Replaced withRectangle& operator=(const Rectangle&) = default;[/source]

EDIT: and yeah it's just a pod struct

Edited by joew, 23 October 2012 - 04:43 PM.

[Washu]

I'm CONFUSED?!?

So am I.... this change actually introduces a compile error for some wtf reason:

[source lang="cpp"]// OriginalRectangle& operator=(const Rectangle& rhs) : left(rhs.left), bottom(rhs.bottom), right(rhs.right), top(rhs.top) {}// Replaced withRectangle& operator=(const Rectangle&) = default;[/source]

EDIT: and yeah it's just a pod struct

If you're going to do one assignment operator, you should do both (move and copy).

As an additional note: VS2012 does not support default nor deleted

Edited by Washu, 23 October 2012 - 04:47 PM.

[joew]

If you're going to do one assignment operator, you should do both (move and copy).

I do

[joew]

As an additional note: VS2012 does not support default nor deleted

Gross! I'll stick with clang which has the majority of the standard in

EDIT: FYI the issue was just that another class using Rectangle didn't have operator= defined ... that's nice it makes it easy to track down during a cleaning day like today

Edited by joew, 23 October 2012 - 04:52 PM.

[mhagain]

I once used negative array indices in combination with a semi-Duff's Device - and winced in mixed horror/pride afterwards. But it was in the days when C-abuse like that used to please me (I've sinced learned).

Statements like this make me smile:

if ((time = (newtime = GetTime ()) - oldtime) > 0)

Wouldn't do it in real code, but it gets a "hey! you can do this!" comment in stuff I'm fooling around with.

Edited by mhagain, 23 October 2012 - 07:04 PM.