- Boost - lists.preview.boost.org

More boostcon videos
by Marshall Clow 30 Jun '11

30 Jun '11

Bartoz (Tuesday): http://blip.tv/boostcon/haskell-the-pseudocode-language-for-c-template-meta… Tony (Wednesday): http://blip.tv/boostcon/lockfree-programming-part-2-data-structures-5258064 Cliff (Wednesday); http://blip.tv/boostcon/design-journeys-with-skoot-a-peer-oriented-generic-… I'm hoping to get about half of the videos up before I go on holidays next week. -- Marshall -- Marshall Marshall Clow Idio Software <mailto:mclow.lists@gmail.com> A.D. 1517: Martin Luther nails his 95 Theses to the church door and is promptly moderated down to (-1, Flamebait). -- Yu Suzuki

5 5

[phoenix] Implementing phoenix::make_shared<T>(...)
by Pyry Jahkola 30 Jun '11

30 Jun '11

Hi, Thomas and other Boosters! I promised on #boost yesterday to try implementing make_shared<T>(...) for Boost.Phoenix. Well, I got the first version working, it was mostly a find-and-replace task to the existing phoenix::new_<T>(...) implementation. I put the modifications to the current Phoenix trunk version here: https://github.com/pyrtsa/boost-svn/commit/b37cb155ff157199a88e59e44f1d8ede… First, the questions, then some remarks and ideas for improvement: 1) If I've understood correctly, you and Hartmut used Wave to generate the preprocessed headers. What command did you use for them? (Currently, I just added a false condition to the preprocessor #if's to fallback to the Boost.Preprocessor iteration.) 2) Where would you put the new make_shared stuff? The first implementation resides below boost/phoenix/object, but I think we should use another directory for all the smart pointer stuff - especially if we're trying to support all of (boost|std::tr1|std)! How about the following pattern? boost/phoenix/smart_ptr/... for all Boost.SmartPtr support stuff boost/phoenix/tr1/... for TR1 support (alternatively, boost/phoenix/std/tr1) boost/phoenix/std/... for C++0x (with #ifdef guards checking against std version) 3) To my pleasant surprise, I didn't need any hacks to create a functor of type function<shared_ptr<Parent>()> using a Phoenix construct like phx::make_shared<Child>(args), where Child inherits Parent. You'll find a working example in https://gist.github.com/1051564. Nevertheless, I think there's also a need for implementing the Phoenix versions of static_pointer_cast etc. 4) Another feature worth implementing might be phoenix::make_unique<T>(...). 5) We also discussed about an API that allows you to make "deep" copies of pointed objects in Phoenix. I think it might already work using something like phoenix::make_shared<T>(ref(x)) if x is of type T, or phoenix::make_shared<T>(*ref(p)) if p points to T. 6) When we get this far, I'll promise help in the documentation effort too. ;-) -- Pyry Jahkola (pyrtsa @ IRC, Twitter, etc.) pyry.jahkola(a)iki.fi +358 40 7529128

3 2

[1.47.0] Beta 1 Release candidate files available
by Beman Dawes 29 Jun '11

29 Jun '11

Release candidate files for 1.47.0 beta 1 are available at http://boost.cowic.de/rc/ As always, the release managers would appreciate it if you download the candidate of your choice and give building it a try. Please report both success and failure, and anything else that is noteworthy. This helps ensure the candidates build OK before we push them out to SourceForge. Thanks, --The release managers PS: I've downloaded and built the Windows 7z file. No problems.

19 34

[1.47.0] permission to merge lexical_cast tests failure markup
by Antony Polukhin 29 Jun '11

29 Jun '11

You can see the changes at https://svn.boost.org/trac/boost/changeset/72798 It adds markup for new versions of some compilers, that continue to fail very strict test. Best regards, Antony Polukhin

2 1

[1.47.0] permission to merge phoenix bugfix
by Thomas Heller 29 Jun '11

29 Jun '11

Hi, I just managed to get my hands on this bug report: https://svn.boost.org/trac/boost/ticket/5626 The bug pointed out a sever failure in the phoenix return type calculation meta programs. The changeset fixing this problem is: http://svn.boost.org/trac/boost/changeset/72794 Please don't be intimidated by the huge number of changed lines. Most of them are due to a regenaration of the (partially) preprocessed header files. The risk of this fix should be relatively low. It just fixes a problem introduced by emulating perfect forwarding which wasn't addressed before. Hope it is not too late for the fix. Thanks, Thomas

2 2

Question about useful libraries
by Denis Shevchenko 29 Jun '11

29 Jun '11

Hello all! I want to ask an important question to the leading developers of Boost. The fact that many programmers write to mailing list of Boost developers about their libraries. And they raised the questions about interest in their libraries and about inclusion their libraries in Boost. But very often proposed library (or library with similar functionality) already exists in Boost, or nobody is interested in a such library. Could you write about the subject areas and technical directions for libraries which would be *REALLY* useful in Boost today? Thanks. - Denis

8 11

Request for Feedback on the Singularity Design Pattern
by Ben Robinson 29 Jun '11

29 Jun '11

The Singularity Design Pattern seeks to address the three weaknesses of the Singleton Design Pattern. Namely: 1) Lifetime: Singleton introduces lifetime management complexities (for a detailed discussion, see Chapter 6 of Modern C++ Design by Andrei * Alexandrescu*). 2) Initialization: Singleton restricts the class to the default constructor. This makes Singleton unusable for classes which require non-default constructors. 3) Encapsulation: Singleton provides global access to the class. Some may view this as a feature, but advocates of unit testing, dependency injection, and context independent objects would disagree. Singularity guarantees a single instance of a class, but does not suffer from 1-3 above. Usage is as follows: // Usage as a Factory: typedef singularity<Horizon, single_threaded> HorizonSingularityType; Horizon & horizonF = HorizonSingularityType::create(value, &event, event); HorizonSingularityType::destroy(); // Usage as a Base Class: class Horizon : public singularity<Horizon, multi_threaded, global_access> Horizon & horizonB = Horizon::create(value, &event, event); Horizon & horizonC = Horizon::get(); Horizon::destroy(); The usage example above shows the three advantages of Singularity over Singleton: 1) Singularity does not permit lazy initialization by design. The lifetime of the single object is bounded by create() and destroy(), and this lifetime is managed just like objects bounded by new() and delete(). 2) Singularities of objects can be created using any available constructor for that object. 3) Singularity prevents global access to the instance by default. However, if this is required, or will help to transition from Singleton, a policy can be supplied which enables a get() accessor to the instance. It is also worth pointing out that thread safety is policy based, and both a single_threaded, and multi_threaded policy are provided. The source files are available here: http://www.mediafire.com/?d8kn5fx2n5d22 I look forward to your feedback. Thank you, Ben Robinson, Ph.D.

3 6

[GSoC] Request for Feedback on Boost.Bloom Filter Project
by Alejandro Cabrera 29 Jun '11

29 Jun '11

Hello all, I'm writing to request feedback, suggestions, and advice from the Boost community. I've been working on a Bloom filter data structure for about a month and a half now. At this point, the following is complete and available in the Boost Sandbox: - working, basic prototype (insert & query operations, union & intersect) - a test suite covering the public interface - five examples - documentation covering many aspects of the project (tested on Chrome 12 and 13, Firefox 4 and 6) Before I proceed to the next phase, I need feedback. Specifically, I need comments on: - code quality: is there anything that stands out about my design, naming convention, or implementation that could use improvement? - documentation quality: was it informative, clear, and easy to understand? Is there anything missing? At this point, there are many directions the Bloom filter project can go, and I need to know which direction would best serve the interests of the community. For this, I would be very interested in hearing about: - desired usage: do you have a project that could use a Bloom filter? What kind of Bloom filter would you need? - compression policy: would you be interested Bloom filters that could be serialized to a compressed form and deserialized back into a decompressed form? - variants: there are a few planned variants listed in the design page of the documentation (future directions). Do any of these seem particularly useful to you? To access the source, visit the sandbox here: http://svn.boost.org/svn/boost/sandbox/bloom_filter/trunk/ To access the documentation, visit the sandbox here: http://svn.boost.org/svn/boost/sandbox/bloom_filter/trunk/libs/bloom_filter… All portions of the project assume the canonical Boost release model. Namely: $(BOOST_ROOT)/boost/bloom_filter $(BOOST_ROOT)/libs/bloom_filter Thanks! -Alejandro Cabrera

7 16

Re: [boost] [GSoC] Request for Feedback on Boost.Bloom Filter Project
by Alejandro Cabrera 29 Jun '11

29 Jun '11

Jakub, > Having bloom_filter in boost:: is great idea! It is super useful, we use in > production code (our own implementation (actually we use Bloomier filters > with murmur hash perfect hashing from your wiki reference). > It lets us use very memory and lookup efficient data structure for very > large datasets (think 100 GB file with strings on SSD disk indexed by e.g. > 200 MB Bloomier filter in memory) I've read about Bloomier filters, though those seem a ways off still. I'm building from the most basic to the more complex variants, in general. Your statistics sound amazing, to say the least! Working with very large data isn't something I've had the opportunity to do just yet, though I'm interested and aware of such scenarios, as my research area is in operating systems and storage issues. > Did you considered adding serialization of a bloom_filter to your > implementation? I haven't considered it at length, but there must be a way to encode the type of the Bloom filter uniquely prior to writing out its bits, so that when deserialization time comes, a user wouldn't accidentally instantiate a Bloom filter of the wrong type. The idea I have in mind at the moment is to first serialize the template parameters of the Bloom filter (Type, Size, HashFunctions). The Type and the HashFunctions appear to be the trickiest, since their structure can take on as many possibilities as the type system allows. Writing out the bits should be the easiest part. If you have any ideas in this regard, I'd appreciate it if you shared them. > In general reconstructing hash based containers with series of inserts is > pretty inefficient. I can imagine this to be the case, especially if the containers are storing a very large number of elements. > Use case that I'm talking about: e.g. for you web proxy scenario proxy > service keeps running and downloading, caching URLs and adding them to bloom > filter. Than the process needs to be restarted for some reason. All > documents downloaded and stored on disk will have to be reitereted and their > URLs reinserted to newly created bloom_filter, which makes the startup of > the process slow. > btw I have the same problem with standard containers (except std::vector). > There is no efficient serialization / deserialization for them rendering > them useless for any larger side project (like unordered_set of 1m strings). > > There was a Boost Serialization library in the works. It seems that the last time it was maintained was 7 years ago. I'm not sure what generic, type-safe [de]serialization would look like - it sounds like a project all on its own! I agree that the C++ standard would benefit greatly from a means to [de]serialize data structures, much like Python has the pickle module. Thank you for your feedback, Jakub. -Alej

2 5

[1.47.0] permission to merge filesystem unmerged changesets
by Beman Dawes 28 Jun '11

28 Jun '11

These change sets from the end of May were never merged to release. They include bug fixes, and have been stable in trunk. https://svn.boost.org/trac/boost/changeset/72309 https://svn.boost.org/trac/boost/changeset/72306 https://svn.boost.org/trac/boost/changeset/72278 My apologies for not catching this earlier. OK to merge? --Beman

2 3

[preprocessor] pp-strings
by Lorenzo Caminiti 28 Jun '11

28 Jun '11

Hello all, As I understand it, Boost.Preprocessor was largely based on another preprocessor library from the same authors named CHAOS. Is this correct? Do you know where I can find CHAOS source and its documentation? Specifically, I am curious if CHAOS supports "pp space-separated strings" -- i.e., macros like the followings: PP_STRING_CAT( b o o s t ) // expand to `boost` PP_STRING_AT(3, b o o s t ) // expand to `s` etc... If yes, why were these pp-strings not ported from CHAOS to Boost.Preprocessor? Thanks a lot. --Lorenzo

8 12

[algorithm] to_lower_copy / to_upper_copy with no output param assumes sequence to be same as input
by Sebastian Karlsson 28 Jun '11

28 Jun '11

It would seem: std::string foo = boost::algorithm::to_lower_copy< std::string >( "bar" ); isn't valid. Instead one has to do for example: std::string foo = boost::algorithm::to_lower_copy( std::string( "bar" ) ), which while perhaps just as terse and natural creates an extra unneeded temporary. I propose the following addition, as well as one for to_upper_copy: template<typename OutputSequenceT, typename RangeT> inline OutputSequenceT to_lower_copy( const RangeT& Input, const std::locale& Loc=std::locale()) { OutputSequenceT Output( ::boost::begin( Input ), ::boost::end( Input ) ); ::boost::algorithm::detail::transform_range( ::boost::as_literal( Output ), ::boost::algorithm::detail::to_lowerF< typename range_value<OutputSequenceT>::type >(Loc)); return Output; } Or perhaps I've missed something altogether :) Kind regards, Sebastian Karlsson

3 9

filesystem3::rename error codes
by Andreas Haferburg 28 Jun '11

28 Jun '11

Greetings, I was wondering how to handle errors of filesystem3::rename(). I'd like to use the system::error_code& overload, but I can't seem to find out how to use it. The documentation at [1] points to POSIX rename(), so I'm guessing the error_code somehow corresponds to the POSIX rename() return value. I'm not a native speaker, but I find the documentation pretty vague here. The tutorial section [2] doesn't help either. For example if the target file doesn't exist, the error_code.value() is 3. But I'm not supposed to use the constant 3 in the code, am I? Are there any constants defined that correspond to these error code values? What's the proper way to do the following: TEST( BoostFilesystem3Test, testRenameFails ) { boost::system::error_code ec; boost::filesystem3::rename("c:/does/not/exist", "asdf", ec); ASSERT_EQ(3, ec.value()); } I'm going to use QFile for now, but only because their documentation is better. I would like to use boost for as much as possible, so any explanation or pointers to better documentation would be much appreciated. Thank you. Best regards Andreas [1] http://www.boost.org/doc/libs/1_46_1/libs/filesystem/v3/doc/reference.html#… [2] http://www.boost.org/doc/libs/1_46_1/libs/filesystem/v3/doc/tutorial.html#E…

1 0

Re: [boost] Is there interest in an alternative to the Singleton anti-pattern?
by Ben Robinson 28 Jun '11

28 Jun '11

> > On 24 June 2011 02:06, Ben Robinson <icaretaker(a)gmail.com> wrote: > > > If you choose the default policy of single_threaded, there is no > > thread safety, but there is also no overhead in using a mutex and using > a > > volatile pointer. > > > > Volatile in C++ doesn't really help with threading... Check out < > http://software.intel.com/en-us/blogs/2007/11/30/volatile > -almost-useless-for-multi-threaded-programming/> > for one of many articles on the subject. > -- > Nevin ":-)" Liber <mailto:nevin@eviloverlord.com> (847) 691-1404 Nevin, I greatly appreciate the input you have provided. I retract my original response, as I now better understand why volatile does not add any value to Singularity. Singularity is already acquiring a mutex object, before accessing the shared instance pointer, and that mutex acquisition is already providing memory fencing. I have simplified the implementation of Singularity and the latest version is available at: https://github.com/icaretaker/Singularity. Any additional input would be appreciated. Thank you, Ben Robinson, Ph.D.

1 0

Logging library
by Tigran Hayrapetyan 28 Jun '11

28 Jun '11

Hello dear Boost community I'm Tigran Hayrapetyan from Armenia, 22 years old. I have started learning C++ when I was 14, after 1.5 years practicing on BASIC. I want introduce to you an idea related to logging, that I came up with recently. Existing logging libraries which I used, are writing log messages linearly, just one after another. In contrast, as we know, almost any application's working structure is not linear, but hierarchical instead, i.e. functions which call other functions, and so on. When viewing large log file produced by an application, it is not very convenient to find log messages of functions, which took middle place in call hierarchy. Because between that messages there are a lot of other messages produced by lower-level functions in that hierarchy. So, generally, it takes some time to understand called functions sequence, and to find the point from which something goes wrong. After spending time looking for log messages I need in my applications, I think that it is possible to organize logging not linearly, but hierarchically, so the messages will form something like a tree. Log messages which are related to some subtask, can be grouped as child messages of the caller task. Also, in my opinion, priority of the caller task's message can be equal to maximal priority of child messages, so it will show whole status of that task (higher priority is for more serious messages). So I decided to implement such logging library. I put it in namespace "tlog" by now, for "Tree Logging". In my implementation hierarchical messages are presented in XML format, as it is some standard for tree-like data. There are two attached files. The library itself is in "tlog.hpp" file, and "main.cpp" contains a sample application which is using the library and writes log to "tlog.xml" file. They are written in MS Visual Studio 2008. Log messages are written with help of 4 macros: TLOG_INFO( Logger, Text ) : logs message with specified text to specified logger with priority "information" TLOG_WARNING( Logger, Text ) : logs message with specified text to specified logger with priority "warning" TLOG_ERROR( Logger, Text ) : logs message with specified text to specified logger with priority "error" TLOG_SCOPED_NODE( Logger, NodeName ) : opens xml element in the log file. The element will be closed at the end of scope, so there is no need to close it manually. I think that the best way to implement entering/leaving subnodes of the log file, is to use scopes of the code. Please review run result and the code, and see if it can be useful for boost. Best Regards, Tigran Hayrapetyan

1 0

Re: [boost] numeric_cast
by Tom Tan 27 Jun '11

27 Jun '11

It seems to me the outer #ifdef works for MSVC as well. What's the extra benefit of embedding an inner #pragma once in this case? 在 2011-06-24五的 04:22 -0400，boost-request(a)lists.boost.xn--org:-7z8fk634a > Message: 1 > Date: Thu, 23 Jun 2011 21:43:02 +0000 > From: "Stephan T. Lavavej" <stl(a)exchange.microsoft.com> > To: "boost(a)lists.boost.org" <boost(a)lists.boost.org> > Subject: Re: [boost] numeric_cast > Message-ID: > > Actually, the best thing to do is this: > > #ifndef MYPROJECT_MYDIRECTORY_FOOBAR_HPP > #define MYPROJECT_MYDIRECTORY_FOOBAR_HPP > > #ifdef _MSC_VER > #pragma once > #endif // _MSC_VER > > meow; > > #endif // MYPROJECT_MYDIRECTORY_FOOBAR_HPP > > This way, all compilers get the portable idempotency guard. Because it's on the outside, GCC (and possibly other compilers) will magically recognize it and avoid opening the header file twice. And the #pragma once for MSVC also instructs it to avoid opening the header file twice, which can improve build perf and can't hurt. > > Stephan T. Lavavej > Visual C++ Libraries Developer >

7 8

Re: [boost] [Boost-build] [1.47.0] Beta 1 Release candidate files available
by Matthew Chambers 27 Jun '11

27 Jun '11

On 6/21/2011 5:28 PM, KTC wrote: > On 21/06/2011 20:47, Jürgen Hunold wrote: > >>> bjam --without-python --stagedir=x86 --build-type=complete -j2 stage >>> >>> Both result in: "...failed updating 2 targets..." >>> See<http://pastebin.com/s7Cvw2Tm> for failure. >> >> Boost.Exceptions is a library on Win32. It builds >> "libs/exception/src/clone_current_exception_msvc.cpp" >> >> which states at the top: >> >> #if defined(_MSC_VER)&& defined(_M_IX86)&& !defined(_M_X64) >> >> so this is for x86 only. On x64, no code gets compiled. >> >> The msvc linker then "helpfully" silently omits the creation of an empty dll >> and the corresponding copy step(s) fail(s). >> >> The correct fix would be to build this file only on for the<address-model>32 >> case but I don't know if this can be specified with the current Boost.Build. > > There's the problem that this is also occurring when building for x86 while on a x64 system. Does "<address-model>32:<build>no" as a requirement work? -Matt

1 0

[msm]how to distinguish region ids?
by Takatoshi Kondo 27 Jun '11

27 Jun '11

Hello, On Boost.Msm 1.46.1, When I use explicit entries and entry pseudo states, I have to provide a template argument "region id". http://www.boost.org/doc/libs/1_46_1/libs/msm/doc/HTML/ch03s02.html#d0e930 But I couldn't find documents about numbering policy of region id. I also check the example below. http://www.boost.org/doc/libs/1_46_1/libs/msm/doc/HTML/examples/DirectEntry… The example passes the region id 0 and 1. I'd like to know how to decide which region is 0(or1). Am I missing something? Thanks, Takatoshi

2 5

[1.47.0] Permission to merge filesystem HP-UX issue fix?
by Beman Dawes 27 Jun '11

27 Jun '11

OK to merge: https://svn.boost.org/trac/boost/changeset/72521 Thanks, --Beman

2 3

[1.47.0] Merge requests for iostreams + quickbook
by Daniel James 27 Jun '11

27 Jun '11

Hi, Is it okay to merge a couple of changes to release? This change activates an existing workaround for Visual C++ 8 in iostreams. https://svn.boost.org/trac/boost/changeset/72726/ This change reverts the support for escaped characters in 'simple markup' in quickbook because it breaks the regex documentation. It isn't used anywhere else, and the documentation still builds, so it's safe. Will be reintroduced for quickbook 1.6. https://svn.boost.org/trac/boost/changeset/72695/

2 1

[review][assign] Formal review of Assign v2 ends today
by John Bytheway 27 Jun '11

27 Jun '11

The formal review of the Assign v2 library upgrade by Erwann Rogard is scheduled to end today. Although there have been no actual reviews with formal votes, there has been useful discussion of the library, and we have chosen not to request an extension of the review. Despite the end of the review, please feel free to discuss the library further, make any final comments, or even submit a late review. I will certainly not declare the review request before next Monday, and will incorporate any comments made before then. I would like to thank all those who participated in the discussions, and Erwann for developing the library, and being so prompt in his responses during the review. A quick reminder of where to find more about the library: The documentation is online at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/libs/assign/v2/doc/html/in…> With the change log at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/libs/assign/v2/doc/html/bo…> The Assign v2 source can be found in the Boost sandbox at: <http://svn.boost.org/svn/boost/sandbox/assign_v2/> or downloaded from: <http://www.boostpro.com/vault/index.php?action=downloadfile&filename=assign…> John Bytheway Review Manager

4 3

Re: [boost] [GSoC] Request for Feedback on Boost.Bloom Filter Project
by Pierre Morcello 26 Jun '11

26 Jun '11

Hello Alejandro, In your first example, you wrote : for (int i = INSERT_MAX; i < CONTAINS_MAX; ++i) { if (bloom.probably_contains(i)) cout << "collision" << endl; } I suppose that you meant ++collision instead. Best regards, Pierre Morcello

2 1

Re: [boost] [GSoC] Request for Feedback on Boost.Bloom Filter Project
by Phil Endecott 26 Jun '11

26 Jun '11

Arash Partow wrote: > Phil Endecott wrote: > > Sorry Arash but I don't understand your reply; what exactly > > is correct? > > My point was that you are correct, the "and" does not provide anything of > value. Only "or" and "xor" are practical for non-counting bloom filters. Well no, that's not what I said. For clarity, I believe that: 1. Bitwise OR of two bloom filters achieves the set union without any side effects. 2. Bitwise AND of two bloom filters achieves the set intersection, but it will introduce additional false positives, which may or may not be acceptable since they are already inherent in bloom filters. 3. Bitwise XOR of two bloom filters achieves something like the set symetric difference, but will introduce false negatives, which probably are not acceptable because bloom filters do not otherwise suffer from them. > I'm not sure what you mean by overlapping hashes I simply mean the case where two distinct values have the same hash for some of the hash functions. Regards, Phil.

1 0

[1.48.0] Schedule discussion
by Beman Dawes 26 Jun '11

26 Jun '11

In the "[boost] [1.47.0] Release branch now closed" thread, Daniel James and I discussed possible schedules for 1.48: Beman>>> While the release managers haven't talked about the 1.48.0 schedule Beman>>> yet, I personally would like to get back to our traditional quarterly Beman>>> schedule. That would mean 1.48.0 would be scheduled for the first Beman>>> Monday in August. Daniel>> We haven't got enough time for that. If we release 1.47 after a 2 week Daniel>> beta (i.e. pretty quickly), it will be released on 4th July, leaving Daniel>> just under a month before the first Monday in August. Beman> Agreed, but we could aim for as short a 1.48 schedule as is reasonable Beman> so that the 1.49 schedule is then back on the quarterly cycle. A 1.48 beta 5 weeks after 1.47 ships seems doable. That would be about August 8th, and then the release 3 weeks later, or about August 29. The 1.49 release then would be back on schedule for the first Monday in November. To make the above happen, we would need to open branches/release for merges immediately after 1.47 ships. Or maybe create a branch now for 1.48. Thoughts? --Beman

9 11

Re: [boost] Is there interest in an alternative to the Singleton anti-pattern?
by Ben Robinson 26 Jun '11

26 Jun '11

> > > If you choose the default policy of single_threaded, there is no > > thread safety, but there is also no overhead in using a mutex and using a > > volatile pointer. > > > > Volatile in C++ doesn't really help with threading... Check out < > > http://software.intel.com/en-us/blogs/2007/11/30/volatile-almost-useless-fo… > > > for one of many articles on the subject. > -- > Nevin ":-)" Liber <mailto:nevin@eviloverlord.com> (847) 691-1404 > > Thank you for the feedback. That article and the accompanying discussion is very interesting and informative. However, given the design of Singularity, I feel I still must make "instance_ptr" volatile for thread safety. If I do not, the compiler can store the pointer in a register, and other threads of execution will not see changes to its value. The linked article seems mostly concerned with fencing access to the variable, which I am achieving by using a mutex object. I only need volatile to prevent effectively reading a "cached" value for the pointer. I found the following article by Andrei Alexandrescu especially interesting: http://drdobbs.com/cpp/184403766 Also note, that the ThreadingModel for Singularity is policy based, so users of the class are free to develop alternative thread-safe policies, using whatever techniques are appropriate for their environment. Thank you, Ben Robinson, Ph.D.

2 1