- Boost - lists.boost.org

[release mgt] Eric Niebler joining release management team
by Beman Dawes 14 Oct '08

14 Oct '08

Eric Niebler is joining the release management team. Eric will be helping out with general release management and quality control. Like Rene Rivera and Daniel James, he will backup Beman so that release management isn't dependent on any single person. Thanks to Eric for agreeing to help out! --Beman

4 3

Ticket #2115 Avoid bad Apple macros
by Beman Dawes 14 Oct '08

14 Oct '08

> Apple has a header called <AssertMacros.h> that #defines both “check” > and “check_error.” These macros naturally collide with the same names > in Boost. We should consider a global change. What global change? That Boost not use these names? What about the other horrid macros in that header? Here is a list from Marshall: debug_string, check, ncheck, check_string, ncheck_string, check_noerr, check_noerr_string, verify, nverify, verify_string, nverify_string, verify_noerr, verify_noerr_string, verify_action, require, nrequire, require_action, nrequire_action, require_quiet, nrequire_quiet, require_action_quiet, nrequire_action_quiet, require_string, nrequire_string, requireaction_string, nrequireaction_string, require_noerr, require_noerr_action, require_noerr_quiet, require_noerr_action_quiet, require_noerr_string, require_noerr_action_string. Marshall comments: IMHO, the really nasty ones are: check, verify, require and check_error. > That would also > probably require a global policy change for checkins. > > Assigning to Beman, Cc'ing John M, component "inspection script" > assigned to John although I might be wrong about that, but obviously > we need to discuss this. Maybe we'll decide to make it "wontfix" I really don't want to start a policy of avoiding every name that is defined in any vendor header. I forget the details, but at least one vendor has shipped a header defining "read" and "write" macros! The fix for this one is to complain to Apple. They need to provide a new new header that supplies the functionality but using macro naming discipline, deprecate the old header, and add a warning to the old header suggesting that users migrate to the new header. If they wanted to be really helpful, they could provide a script to automatically change the old names to the new names. And if they have any other headers that fail to apply macro naming discipline, they should do the same for those headers too. By "naming discipline", I mean something like the Boost policy for macro names always being all uppercase and beginning with BOOST_. --Beman

8 14

New library in Vault: Msm (Meta State Machine)
by christophe henry 14 Oct '08

14 Oct '08

Hi, Following advice from Dave Abrahams and Aleksey Gurtovoy who were nice enough to have a look at the doc, I just uploaded to the Vault the first version of Msm and would be delighted to have as many people as possible have a look at it and give their opinion before I dare request a formal review. Msm offers an alternative to Boost.Statechart by implementing a relatively big subset of the UML State machine specification at high run speed and staying as close as possible to the beautiful Domain-Specific-Language described in the "C++ template metaprogramming" book. Tutorials and features can be found in doc/index.html. In doc you will also find the tutorials' code. Christophe _________________________________________________________________ Discover the new Windows Vista http://search.msn.com/results.aspx?q=windows+vista&mkt=en-US&form=QBRE

2 1

[parameter] error in BOOST_PARAMETER_NAME reference docs
by Ryan Gallagher 14 Oct '08

14 Oct '08

There appears to be an error in the BOOST_PARAMETER_NAME reference docs. See: http://tinyurl.com/4my48e This lists that the macro parameter "name" can be of the form: "(tag-name, namespace-name) object-name" Where "tag-name" is a structure defined in namespace "namespace-name" and "object-name" is an instance of: "::boost::parameter::keyword<tag-namespace::tag-name>::instance" (Also note that "tag-namespace" should be "namespace-name" here.) However, the implementation appears to actually be "(object-name, namespace-name) tag-name" However, this information is repeated under the "Fine-Grained Name Control" section, in which it appears to be correct. -Ryan

1 0

[1.37.0][Spirit] Spirit branches/release regression test failures
by Beman Dawes 14 Oct '08

14 Oct '08

See http://beta.boost.org/development/tests/release/developer/spirit.html Note that many of these tests were run after the revision 49310 branches/release commits. Please resolve these failures as soon as possible. Thanks, --Beman

2 1

[date_time] Patch for ticket #1740: date_time includes <windows.h>
by Emil Mieilica 14 Oct '08

14 Oct '08

Hello, date_time on Windows needs windows.h in order to have the definitions for FILETIME and GetSystemTimeAsFileTime and friends and does not use the same technique as other boost headers : if BOOST_USE_WINDOWS_H is defined include <windows.h> otherwise forward define functions as needed. In date_time is more difficult because it is a header_only library. Please find attached a patch to solve this issue and the test report of the library on the compiler I have at my disposal. I have taken precautions in order to ensure it will work on x64 Windows (please see comments related to the conversion between FILETIME and __int64) and I see no reason it should have changed for other platforms, but only testing will tell. Thank you, -- Emil Mieilica __________ Information from ESET NOD32 Antivirus, version of virus signature database 3506 (20081009) __________ The message was checked by ESET NOD32 Antivirus. http://www.eset.com

6 7

Identifying members
by coppro 14 Oct '08

14 Oct '08

In C++ it is possible to determine whether there exists a member of a class of a specified type (including member functions) with a given name. As far as I know, there exists no such functionality in Boost. Why is this? It can easily be implemented as a traits class for a specific identifier/signature, which can then be made into a macro.

7 8

Merge permission sought
by Nicola Musatti 13 Oct '08

13 Oct '08

Hallo, I would like to merge the following changes from trunk to the release branch: http://svn.boost.org/trac/boost/attachment/ticket/2341/call_traits.patch http://svn.boost.org/trac/boost/attachment/ticket/2342/sp_convertible.patch http://svn.boost.org/trac/boost/attachment/ticket/2352/boost_filesystem.pat… http://svn.boost.org/trac/boost/attachment/ticket/2344/boost_iterator.patch They all relate to supporting the latest CodeGear (ex Borland) compiler and all went through at least one successful test cycle on trunk. May I go ahead? Is the merge procedure outlined here: http://svn.boost.org/trac/boost/wiki/ImprovingPractices#Mergingchangesfromd… still considered the correct one? Cheers, Nicola -- Nicola.Musatti <at> gmail <dot> com Home: http://nicola.musatti.googlepages.com/home Blog: http://wthwdik.wordpress.com/

3 5

[filesystem]Setting permissions/chmod
by Johan Torp 13 Oct '08

13 Oct '08

After looking through boost.filesystem's documentation and code, I can't find anything about permissions. I assume boost.filesystem has not implemented this because it aims to be portable and POSIX style permissions isn't applicable on all operating systems. Am I correct? Best Regards, Johan Torp www.johantorp.com -- View this message in context: http://www.nabble.com/-filesystem-Setting-permissions-chmod-tp19950894p1995… Sent from the Boost - Dev mailing list archive at Nabble.com.

4 4

Geometry and spatial indexes, my opinion
by Mathias Gaunard 13 Oct '08

13 Oct '08

Hi, Wanting to code a simple 3D virtual world, I looked into the recent spatial indexes SOC work, which depends on the proposed geometry library. I have to admit I was fairly disappointed. So I'm going to put some critics, some quite harsh maybe, along with some ramblings about what I would have liked to have. Note that I am no geometry or space indexing expert, so what I say may be perfectly silly. For starters, I think those libraries should be dimension agnostic and generic. Geometry not only has a strong 2D bias, but the documentation also says geometry::box only works with point_xy (and point_ll). Since it is used everywhere in spatial index, that means it's unusable on 3D. The geometry documentation doesn't clearly state what a geometry::box is. Is it (assuming 2D) any rectangle or only axis-aligned rectangles? In case it is just any rectangle, it mind be interesting to introduce a new type for axis-aligned ones, which can always be defined from two points whatever the dimension (well, at least it seems enough for 3D, I'm not good at projecting in other dimensions to see if that'd work). I suppose the spatial index interface expects axis-aligned boxes, so it's fairly important to clarify what we're talking about. Apart from that, there is also lots of point_xy-dependent code in spatial indexes. Also the quadtree implementation isn't dimension agnostic in particular (it should have 2^n children nodes, and not 4). I don't really get why it uses shared_ptr everywhere either (is it supposed to support COW or something?) I don't really get what boost::spatial_index::spatial_index is for. It does nothing but forward everything to the last template parameter. Instead of writing boost::spatial_index::spatial_index< point_type, value_type, boost::spatial_index::rtree<point_type, value_type> > index(args...); I might as well write boost::spatial_index::rtree<point_type, value_type> index(args...); Also, I think the spatial indexes do not provide enough operations. The only things they allow is putting point => value pairs, retrieving the value associated to a point in O(log n) time (I assume, complexity is not documented), and finding all values within a certain box (in O(log n) too?). Many things one typically wants to do with spatial indexes are not possible: - Iterate all (points, value) pairs in the index - Find the element closest (or k-closest) to some point in space or other element. - Find all elements within a certain range of some point in space or other element - Find the elements that intersect with a given object (ray tracing, for example, searches for an element that intersects with a ray and that is the closest to its source) - Consider only a part of the indexed space (e.g. an area delimited by a range/sphere or box) - Being able to index arbitrary objects and not just points - Traverse the spatial index as a tree I believe there is no need to clutter the interface with a lot of functions, most algorithms can be implemented generically as free functions. Giving a read-only tree interface to all spatial indexes ought to be enough to implement all algorithms. A spatial index, after all (correct me if I'm wrong), is nothing else but a tree that partitions space, where every node defines a partition of space that includes all of its children (which might overlap or not). Subtrees should be full-fledged spatial indexes (so that I can perform a search in a subtree), albeit read-only. Apart from that they should allow insertion and removal of object => value pairs. Quadtrees and r-trees should be fairly easy to work with: they both partition space into axis-aligned boxes (which are even cubes in quadtrees). As for allowing all kinds of objects to be stored instead of just points, that should be possible by using generic distance and intersection functions. The real difference is that the same object could eventually be in multiple places in the tree, because it spans over multiple space partitions. Whenever you put an object in the tree or split a partition, however, it would be necessary to calculate in which partitions to put the object, whereas with points you only have to find the one partition. Of course, some cases can be specialized and optimized, like axis-aligned boxes in a R-tree, which can directly be used as partitions and thus never span across multiple ones. I find it weird having a different interface for R-trees and Quadtrees. Indeed, one can only put points in a quadtree while in a r-tree you can put boxes as well. A generic approach allowing to put any object with boxes specialized seems like a better approach to me. About geometry, I think there should be more basic objects. While it is true polygons and multipolygons (polyhedrons) are enough to represent everything in 3D space, having rays, segments, triangles, quads, circles, spheres, squares, rectangles, cubes, cuboids (boxes), planes, half-planes and more, with ability to compute all intersections and distances, would be the most practical thing for 3D development. I do not know how to organize this in a really dimension-agnostic manner, however. In a nutshell, I think the spatial indexes design and how geometry can help it ought to be discussed.

16 61