Boost C++ Libraries: Ticket #979: ::boost::detail::empty_base could be improved

status, milestone changed; owner set

Daniel Frey — Wed, 23 May 2007 20:38:35 GMT

owner set to Daniel Frey
status new → assigned
milestone Boost 1.34.1 → Boost 1.35.0

Accepted, but raised milestone to 1.35.0 as the risk of breaking other compilers is not worth it. Once regression tests show that everything is fine, I might consider moving it back to 1.34.1, but it should not delay 1.34.1.

component, severity changed

Daniel Frey — Wed, 23 May 2007 20:41:20 GMT

component iterator → operators
severity Problem → Optimization

status changed; resolution set

Daniel Frey — Wed, 23 May 2007 23:04:47 GMT

status assigned → closed
resolution → fixed

Fixed in CVS HEAD

status, milestone changed; resolution deleted

Thomas Witt — Tue, 05 Jun 2007 18:21:52 GMT

status closed → reopened
resolution fixed
milestone Boost 1.35.0 → Boost 1.34.1

Reopened for inclusion in 1.34.1

Dave Abrahams — Tue, 05 Jun 2007 19:28:59 GMT

I can't see how this is appropriate for 1.34.1

It's NAD, in my opinion, and should be put off until the next full release.

sparent@… — Tue, 05 Jun 2007 19:34:06 GMT

Although I have a work around so putting off until 1.35.0 is fine with me - the work around is ugly and I do consider a helper template that adds unnecessary words to my objects a defect.

Thomas Witt — Tue, 05 Jun 2007 21:16:38 GMT

Replying to dave:

I can't see how this is appropriate for 1.34.1

It's a "simple" change, passes testing on HEAD and it won't delay 1.34.1 even if we have to back it out again.

It's NAD, in my opinion, and should be put off until the next full release.

If this were a commercial project I would agree. My experience is that a more opportunistic approach is required for boost. There is no way prioritizing changes when you can not prioritize work items.

Dave Abrahams — Tue, 05 Jun 2007 22:45:05 GMT

Replying to witt:

Replying to dave:

I can't see how this is appropriate for 1.34.1

It's a "simple" change, passes testing on HEAD and it won't delay 1.34.1 even if we have to back it out again.

Unless the change is not exactly as proposed in the summary of this ticket, it is a breaking change for users. I can't imagine any good excuse for doing that in a point release, and I object strongly to doing so.

attachment set

Thomas Witt — Wed, 06 Jun 2007 00:01:13 GMT

attachment → operators.hpp.diff

Thomas Witt — Wed, 06 Jun 2007 00:04:56 GMT

Replying to dave:

Replying to witt:

Replying to dave:

I can't see how this is appropriate for 1.34.1

It's a "simple" change, passes testing on HEAD and it won't delay 1.34.1 even if we have to back it out again.

Unless the change is not exactly as proposed in the summary of this ticket, it is a breaking change for users. I can't imagine any good excuse for doing that in a point release, and I object strongly to doing so.

Dave, I have attached the proposed patch. This may very well be my fault. I never intended to add a breaking change. Looking at the patch I still can't figure it out. What am I missing?

cc changed

Thomas Witt — Wed, 06 Jun 2007 00:10:02 GMT

cc witt@… added

Dave Abrahams — Wed, 06 Jun 2007 00:27:23 GMT

Sorry, the mistake is mine. I didn't realize, from the example, which used no qualification and showed empty_base named in what is presumably user code, that empty_base was in boost::detail. As long as users aren't touching empty_base the patch is OK.

Thomas Witt — Wed, 06 Jun 2007 00:31:46 GMT

Replying to dave:

I always assumed everything in boost::detail is fair game. Should we be more conservative?

anonymous — Wed, 06 Jun 2007 00:47:33 GMT

Replying to witt:

Replying to dave:

I always assumed everything in boost::detail is fair game.

Well, it should be... but occasionally you see stuff like this:

http://www.boost.org/libs/serialization/doc/archive_reference.html http://www.boost.org/libs/serialization/doc/derivation.html

Should we be more conservative?

I don't think so. Did http://svn.boost.org/trac/boost/ticket/979#comment:11 lead you to think I would say otherwise.

Thomas Witt — Wed, 06 Jun 2007 00:57:52 GMT

Replying to anonymous:

Replying to witt:

Replying to dave:

I always assumed everything in boost::detail is fair game.

Well, it should be... but occasionally you see stuff like this:

http://www.boost.org/libs/serialization/doc/archive_reference.html http://www.boost.org/libs/serialization/doc/derivation.html

Sigh!

Should we be more conservative?

I don't think so. Did http://svn.boost.org/trac/boost/ticket/979#comment:11 lead you to think I would say otherwise.

No, not really. I am just paranoid about misunderstandings with release issues ;-).

cc, status changed; resolution set

Daniel Frey — Wed, 06 Jun 2007 21:07:18 GMT

cc d.frey@… added
status reopened → closed
resolution → fixed

Applied patch to RC_1_34_0. Sorry Dave, I should have added the patch to the ticket from the very beginning.