April 2005 - Boost - lists.boost.org

Yet another smart pointer type: a new library proposal
by Marcin Kalici�ski 04 Apr '05

04 Apr '05

Hi Everybody, Smart pointers in boost work on a principle of preventing object destruction while references to it exist anywhere in the system. It causes that whoever owns a boost::shared_ptr to some object, has power to decide for how long will that object live. This is "ownership". In some situations it would be more convenient if object itself decided when to die (owned itself?), and notified all reference holders about this fact, so that they not try to use it any more. A real-life example from a computer game code. There are C++ objects representing tanks on the battlefield. When tank gets hit, it explodes and ceases to exist. The nice, straighforward way to do it is to write "delete this;" in C++. All references to deleted tank which are held by other tanks (for example those currently aiming at it, or following it), will be immediately updated to NULL, provided that they are live_ptrs: template<class T> class live_ptr { // Member functions closely resembling those of shared_ptr // When object pointed to is deleted, this pointer resets its value to NULL }; class Tank; void live_ptr_test() { Tank *tank = new Tank; live_ptr<Tank> p(tank); // not ownership, but a "live" reference delete tank; assert(p == 0); } I have been using (gradually improved) variations of such a smart pointer in some of my projects for last couple of years, and found it useful. Anybody interested? cheers, Marcin Kalicinski

5 6

[test] CodeWarrior+IsDebuggerPresent()
by Stefan Slapeta 04 Apr '05

04 Apr '05

Many tests on CW fail due to a missing IsDebuggerPresent() in execution_monitor.ipp. I assume this code should be disabled for CW, no? http://tinyurl.com/59vnj ### mwcc Compiler: # In: ..\boost\test\impl\execution_monitor.ipp # From: ..\libs\logic\test\tribool_io_test.cpp # ----------------------------------------------- # 199: if( IsDebuggerPresent() ) # Error: ^ # undefined identifier 'IsDebuggerPresent' Stefan

4 10

(Named) Parameter Library
by David Abrahams 04 Apr '05

04 Apr '05

Daniel and I are working on getting the Named Parameter Library up to the standards requested as part of its review. Over the past few months I realized I also needed "unnamed parameters" (which have no definite order and are discriminated by type), and it made sense to extend the library in that direction. There's also a need for unnamed (and maybe even named) template parameters, and all of these facilities can use a lot of code in common with what we've already got. I think it makes sense to just call this library Boost.Parameter and put it in namespace boost::parameters. I thought it made sense to give people a chance to object, just in case. But everyone loves this idea, right? ;-) -- Dave Abrahams Boost Consulting www.boost-consulting.com

4 5

Re: Proposal: intrusive containers
by Jaap Suter 04 Apr '05

04 Apr '05

> I have finished the first version of my STL-like intrusive containers > More information on: > http://people.freenet.de/turtle++/intrusive.html Hi Olaf, I am using intrusive lists a lot in my work, and I would love to see high-quality Boost versions. I have not looked at your code, but I'm impressed with your write-up, and it already addresses many issues. Here are my initial comments and thoughts: 1. Instead of providing the two options of deriving and embedding, let the user provide ADL-style 'next' and 'prev' lookup functions that for a given type provide hooks to the intrusive containers. Look at the boost::intrusive_ptr documentation for more information on this style. You can still provide the existing two methods as possible ways of providing these hooks, but if my legacy code (or for some other reason) has types with their own next and prev pointers, I still want to be able to use your code. 2. As you mention, inserting an object twice leads to problems. In general, intrusive lists need to be handled with more care. I would like to see a diagnostics mode that checks invariants and preconditions. Things to check for would be checking if an instance doesn't already exist on inserts, and checking whether there are no loops in a list. Because this will make list-manipulation really slow, I'd like to be able to enable this on a per-instance basis, probably through boolean-ish or policy-based template parameters. 3. I would love to see a single-linked intrusive container that only supports forward iteration, as well as a list decorator that turns an exiting single or doubly linked list into a ring. Thanks for the work, it looks very promising! Regards, Jaap Suter

3 2

Re: [boost] Re: [hash] Boost.Hash test results not showing at thereports
by JOAQUIN LOPEZ MU?Z 04 Apr '05

04 Apr '05

----- Mensaje original ----- De: Daniel James <daniel(a)calamity.org.uk> Fecha: Lunes, Abril 4, 2005 6:52 pm Asunto: [boost] Re: [hash] Boost.Hash test results not showing at the reports > Joaquín Mª López Muñoz wrote: > > Maybe a "sublibs" file is needed at libs/functional? > > Unfortunately, I can't add that because the functional tests are > not in > a subdirectory. So far now, once the tests are running, they're > going to > appear in 'functional'. It's probably best to live with that until > after > the release. > See my previous reply. The tests are being produced, only they're not shown. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

Re: [boost] Re: [hash] Boost.Hash test results not showing at thereports
by JOAQUIN LOPEZ MU?Z 04 Apr '05

04 Apr '05

Hi Daniel, ----- Mensaje original ----- De: Daniel James <daniel(a)calamity.org.uk> Fecha: Lunes, Abril 4, 2005 6:50 pm Asunto: [boost] Re: [hash] Boost.Hash test results not showing at the reports > Markus Schöpflin wrote: > > I just discovered the following in the regression run output: > [snip] > > don't know how to make hash_std > > don't know how to make hash_extensions > [snip] > > Maybe this has something to do with the problem? > > That could be it. There's an error in the Jamfile, which for some > reason > doesn't cause any problems when I run the tests locally. > This definitely looks like it should be fixed, but it is not the reason why the tests do not appear at the regression results: some engines are actually running your tests despite that problem, as you can see by manually inspecting: ftp://fx.meta-comm.com/boost-regression/CVS-HEAD/ For instance, Dirxion.zip has - <test-log library="functional" test-name="hash_custom_test" test- type="" test-program="" target- directory="bin/boost/libs/functional/hash/test/hash_custom_test.test/gcc -3_3-darwin/debug" toolset="gcc-3_3-darwin" show-run-output="true"> <link result="succeed" timestamp="2005-04-03 21:22:03 UTC">export DYLD_LIBRARY_PATH c++ -g - o "/Users/stephencarson/Development/boost/results/bin/boost/libs/functio nal/hash/test/hash_custom_test.test/gcc-3_3- darwin/debug/hash_custom_test" "/Users/stephencarson/Development/boost/r esults/bin/boost/libs/functional/hash/test/hash_custom_test.test/gcc- 3_3-darwin/debug/hash_custom_test.o" \ \ - L"/Users/stephencarson/Development/boost/results/bin/boost/libs/test/bui ld/libboost_unit_test_framework.a/gcc-3_3- darwin/debug" "/Users/stephencarson/Development/boost/results/bin/boost/ libs/test/build/libboost_unit_test_framework.a/gcc-3_3- darwin/debug/libboost_unit_test_framework.a" "/Users/stephencarson/Devel opment/boost/results/bin/boost/libs/test/build/libboost_unit_test_framew ork.a/gcc-3_3-darwin/debug/libboost_unit_test_framework.a" \ Chmod1 /Users/stephencarson/Development/boost/results/bin/boost/libs/fun ctional/hash/test/hash_custom_test.test/gcc-3_3- darwin/debug/hash_custom_test chmod 711 "/Users/stephencarson/Development/boost/results/bin/boost/libs/funct ional/hash/test/hash_custom_test.test/gcc-3_3- darwin/debug/hash_custom_test"</link> <run result="succeed" timestamp="2005-04-03 21:22:38 UTC">Running 1 test case... *** No errors detected EXIT STATUS: 0</run> </test-log> So, the problem must lie elsewhere. Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

1 0

[optional] Where are the fixes for self-assignment problem?
by Joe Gottman 04 Apr '05

04 Apr '05

Several weeks ago I reported that boost::optional<T> was not safe under self-assignment. The author of the class said he was fixing the library, but no new versions of the code have been checked into the CVS repository. With the code freeze for boost 1.33 only a couple of weeks away, I was wondering whether optional<T> would be fixed on time. Joe Gottman

2 1

[hash] Boost.Hash test results not showing at the reports
by Joaquín Mª López Muñoz 04 Apr '05

04 Apr '05

Seems like the report form is not showing the test results for Boost.Hash, yet some of the engines are already producing them. * The tests are defined at libs/functional/hash/test/JamFile * The tests are the subproject libs/functional/hash/test, and the name of the suite is "functional/hash" * The tests are included in /status/Jamfile with subinclude libs/functional/hash/test ; So everything looks like it is in place, but still the tests don't show. Ideas? Maybe a "sublibs" file is needed at libs/functional? Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

3 3

Re: GNU GPL
by Alexander Terekhov 04 Apr '05

04 Apr '05

David Abrahams wrote: > > Uzytkownik <uzytkownik2(a)poczta.onet.pl> writes: > > > Simple answer: yes or no? > > I understand yes, but I'm not sure and, for me, details doesn't matter > > for me. > > If you really want a simple answer, the only responsible thing for any > of us to say is "ask a lawyer." None of us are lawyers and we can > only describe our amateur understanding of the legal issues. You don't need to be a lawyer*** to realize that, apart from Law History, Mr. Moglen is Professor of GNU Copyright Law, not Copyright under Berne Convention and its US "codification", so to speak. All distributions of publicly available [L]GPL'd works (including distribution of derivative works under copyright law) fall under 17 USC 109 (nonexistent in the GNU Republic), not [L]GPL. As for "combinations", software is protected as literary works by definition, and "combination" of multiple works is protected as "compilation", not "derivative work." < Forward Online > -------- Original Message -------- Message-ID: <423FFBFC.5993610A(a)web.de> Newsgroups: gnu.misc.discuss Subject: Re: using GPL api to be used in a properietary software References: ...<slrnd3hm9n.8ci.isaac(a)latveria.castledoom.org> < for the sake of linking-back in future > Isaac wrote: > > On Mon, 14 Mar 2005 12:12:29 +0100, Martin Dickopp > <expires-2005-04-30(a)zero-based.org> wrote: > > > > That depends on how the program has been created and other details. If a > > program uses the ISO-standardized C library API, and uses no components > > of a particular C library while it is being created, then a derivative > > work of the program and a particular C library is created the moment the > > program is run (and therefore linked with the library). But I can also > > imagine different circumstances under which a derivative work is already > > created when the programm is written. > > Taking two works and pasting them together does not create a derivative > work, but instead creates some kind of compilation, most likely > a collective work. If there is no recasting, transforming or adaptation > of the original work, then I don't think the result fits the statutory > definition. > > However a collective work is separate combinations collected into a > whole which seems to accurately describe linking to a library for at > least some linking technologies. > > For static linked works, the diference between a derivative work and > a collective work is usually meaningless because distribution requires the > distributing the library. But the difference is critical if the two > joined works are to be distributed separately. Distribution under "first sale" aside for a moment, here's a nice explanation highlighting some other differences regarding compilations and derivative works. ---- HOUSE REPORT NO. 94-1476 Section 103 complements section 102: A compilation or derivative work is copyrightable if it represents an ''original work of authorship'' and falls within one or more of the categories listed in section 102. Read together, the two sections make plain that the criteria of copyrightable subject matter stated in section 102 apply with full force to works that are entirely original and to those containing preexisting material. Section 103(b) is also intended to define, more sharply and clearly than does section 7 of the present law (section 7 of former title 17), the important interrelationship and correlation between protection of preexisting and of ''new'' material in a particular work. The most important point here is one that is commonly misunderstood today: copyright in a ''new version'' covers only the material added by the later author, and has no effect one way or the other on the copyright or public domain status of the preexisting material. Between them the terms ''compilations'' and ''derivative works'' which are defined in section 101 comprehend every copyrightable work that employs preexisting material or data of any kind. There is necessarily some overlapping between the two, but they basically represent different concepts. A ''compilation'' results from a process of selecting, bringing together, organizing, and arranging previously existing material of all kinds, regardless of whether the individual items in the material have been or ever could have been subject to copyright. A ''derivative work,'' on the other hand, requires a process of recasting, transforming, or adapting ''one or more preexisting works''; the ''preexisting work'' must come within the general subject matter of copyright set forth in section 102, regardless of whether it is or was ever copyrighted. The second part of the sentence that makes up section 103(a) deals with the status of a compilation or derivative work unlawfully employing preexisting copyrighted material. In providing that protection does not extend to ''any part of the work in which such material has been used unlawfully,'' the bill prevents an infringer from benefiting, through copyright protection, from committing an unlawful act, but preserves protection for those parts of the work that do not employ the preexisting work. Thus, an unauthorized translation of a novel could not be copyrighted at all, but the owner of copyright in an anthology of poetry could sue someone who infringed the whole anthology, even though the infringer proves that publication of one of the poems was unauthorized. ---- regards, alexander. ***) http://www.google.de/groups?selm=42374045.B468A838%40web.de

1 0

[1.33 release] proposal: a little art for the ack section
by Joaquín Mª López Muñoz 04 Apr '05

04 Apr '05

Hi all, Continuing with our short-lived tradition of adorning the release ack section with a not so serious picture, I'm taking the liberty of proposing the attached bit. It can be seen in action in http://195.235.93.150/boost.html The guy at the computer is a medieval representation of our valiant release manager Mr. Gregor. If this is felt unappropriate or plain ugly, please say so, I won't have any kind of hard feeling against opposition to the idea: it is just that last Saturday it was ranining in Madrid and I had some vacant time to try my drawing skills. Best, Joaquín M López Muñoz Telefónica, Investigación y Desarrollo

5 4