- Boost-users - lists.boost.org

Interprocess condition and process crashes
by Gabriel Burca 17 Sep '09

17 Sep '09

I'm using Boost 1.39 on WindowsXP and I noticed that if a process crashes while waiting on an interprocess condition variable, the signaling process blocks forever in subsequent calls to notify_all(). It looks to me like boost is not using the native facilities, but some emulation, so is this scenario not supported by the emulation? The code is very simple: Process A: while (1) { { scoped_lock<named_mutex> lock(mut); cond.wait(lock); // Process A might be killed or die while in this call // do some work } } Process B: void notifyOthers() { scoped_lock<named_mutex> lock(mut); cond.notify_all(); } I've actually omitted one small detail in the description above. After process A dies, process B can execute notifyOthers() again only once. The second time it never makes it out of the notify_all(). -- Gabriel Burca

2 1

BGL: Very dense adjacency_matrix
by Mathieu Malaterre 17 Sep '09

17 Sep '09

Hi there, Has anyone worked on very dense graph ? Basically the relation is implicitly given by the position in the grid: Eg, consider the following structured grid A B C D E F G H I it would lead to the following (undirected) graph: Graph g(9); add_edge(A, B, g); add_edge(A, D, g); add_edge(B, C, g); add_edge(B, E, g); add_edge(C, F, g); add_edge(D, E, g); add_edge(D, G, g); add_edge(E, F, g); add_edge(E, H, g); add_edge(F, I, g); add_edge(G, H, g); add_edge(H, I, g); When the structured grid is full (no broken link) I would think there are better data structure (ie. no data at all) suited to represent the graph. And as we break link (say, edge(A,B)), we would use an opposite definition of adjacency_list, to list what edge is actually *missing*. As far as I know adjacency_matrix is not implemented in PBGL, so I'll need to work around that issue /real time soon/. Thanks for your help, -- Mathieu

1 0

PBGL vs (Parallel) BGL-VTK
by Mathieu Malaterre 17 Sep '09

17 Sep '09

Just out of curiosity what are the main differences in between PBGL and (Parallel) BGL-VTK ? What I am looking for is why not use PBGL directly ? What was the actual issue ? Thanks, -- Mathieu

1 0

Multiple async operations on a socket
by Kyle Tarplee 17 Sep '09

17 Sep '09

I have a rather simple question that I can't find in the boost::asio documentation. I have a boost::asio::ip::tcp::socket that I want to write to. Can I have multiple outstanding async_write() requests for a single socket? Are the writes performed in the order in which they are submitted? Or is this undefined behavior? How about multiple async_read() requests on the same socket? I know that I need to write to the socket from at most one thread but that doesn't preclude me from not waiting until the first one is done to put on the second async write or read. The reason I ask is because there seems to be two ways to write data to a socket. 1) The "fire and forget" approach (no queue, simply call async_write() when you want to write and ignore the completion handler callback) 2) Push on a queue when you want to write something. Pop from the queue when the preceeding async_write() operation completed. Thus at most one async_write() operation is associated with the socket at any point in time. Thanks in advance, Kyle

4 6

quaternion component access
by gtsml owevwr 17 Sep '09

17 Sep '09

Hi, Is there a way to set a boost::math::quaternion on an component basis? Also, is there to randomly access an component (operator [])? Finally, the current way to access compenents looks a bit too long to me: "R_component_1()". Is there a shorter way? Thanks, Frank

1 0

Re: [Boost-users] building boost 1_33_1 now 1_37_0
by Dan Smithers 17 Sep '09

17 Sep '09

> Message: 1 > Date: Wed, 16 Sep 2009 17:04:46 +0400 > From: Vladimir Prus <vladimir(a)codesourcery.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] building boost 1_33_1 now 1_37_0 > Message-ID: <h8qnpe$j3a$1(a)ger.gmane.org> > Content-Type: text/plain; charset=us-ascii > > Dan Smithers wrote: > >>>> How do I change the includes in jam? >>> Sorry, I don't understand the above. Boost.Build does not pass any >>> "intel includes" nor any "gcc 2.95 includes". Can you provide the commands >>> lines and errors in question and explain how exactly the command lines >>> should be altered. >> I've now found that as long as I make sure that gcc 3.4.3 is ahead of >> 2.95 in the search path I can get past the first hurdle. > > So, is there anything that still needs to be solved? > >> The command line that comes out is >> >> "icpc" -c -xc++ -O3 -Ob2 -w1 -fPIC -DBOOST_ALL_NO_LIB=1 >> -DBOOST_MATH_TR1_DYN_LINK -DNDEBUG -I"." -c -o >> "bin.v2/libs/math/build/intel-linux-release/threading-multi/hypot.o" >> "libs/math/build/../src/tr1/hypot.cpp" >> >> /usr/lib/pthread.h(163): error: expected a ")" >> extern int pthread_create(pthread_2 *__restrict __thread, > > Do you actually have file named /usr/lib/pthread.h? That is very strange > location for a header file. > >> I have written a short test program that includes pthread.h and that >> compiles. > > Anyway, it does not seem like anything wrong is passed to icpc. So, > this error suggests that it's not installed correctly, or some other > configuration issue. Do docs for the intel compiler say anything about > gcc version that is required? Update - I have managed to get this file (hypot.cpp) to build using the Intel 8 compiler. I wonder what they changed? thanks for your help dan

1 0

[serialization] some of the serialization docs are infected with a random single quote replacement
by Todd Day 17 Sep '09

17 Sep '09

The following files have a single character replaced by a ' (that's a single quote) at a fairly regular interval. This is in the latest trunk via SVN. serialization.html - Changeset 56240 special.html - Changeset 44743 traits.html - Changeset 55676

1 0

[serialization] 1.40.00 BOOST_CLASS_EXPORT_GUID creates memory leak
by Todd Day 17 Sep '09

17 Sep '09

MSVC 8.0 BOOST_CLASS_EXPORT_GUID is used to create some static objects for using in serialization. These memory errors are not too big a deal since they aren't runaway - ie, these objects are created at startup before main() is even called, and the objects are not created over and over again. They are just not deleted at program exit, which causes my debugger to complain about memory leaks. I tracked the source of these leaks to two spots in the code. Unfortunately, the code is so complicated at these points that I can't suggest a fix or workaround. Hopefully, the maintainers know what this is about. The first leak is created in \boost_1_40_0\boost\smart_ptr\detail\shared_count.hpp line @ line 124 pi_ = new sp_counted_impl_pd<P, D>(p, d); I could be wrong, but it appears to me that the above problem only occurs for serialized objects that are derived from other objects at least twice. For example, if B is derived from A, and C is derived from B, then BOOST_CLASS_EXPORT_GUID(B, "B") BOOST_CLASS_EXPORT_GUID(C, "C") results in a memory leak for C, but not for B. I think, at least it looks that way. The second leak is created in \boost_1_40_0\libs\serialization\src\void_cast.cpp @ line 205 if(* (*it)->m_derived == * m_base) new void_caster_shortcut( m_derived, (*it)->m_base, m_difference + (*it)->m_difference, includes_virtual_base ); Again, it appears the these leaks only happen for classes which are derived at least twice. The above allocations are not deleted at shutdown. In the second case, it almost appears to me as if a registration function is accidentally getting called twice, and a pointer is getting set first to one allocation and then to a second allocation, which means the first allocation is lost track of. Thanks for any help, -todd- PS - Just pulled down the latest trunk from SVN. It appears that many changes in the areas I've mentioned above have occurred, so maybe this issue is fixed. If so, I'm hoping I can just pull over the include and source file changes made to the serialization library by themselves, but I don't know if those changes rely on other changes in other Boost libraries.

1 0

Re: [Boost-users] building boost 1_33_1 now 1_37_0
by Dan Smithers 17 Sep '09

17 Sep '09

> > Message: 1 > Date: Wed, 16 Sep 2009 17:04:46 +0400 > From: Vladimir Prus <vladimir(a)codesourcery.com> > To: boost-users(a)lists.boost.org > Subject: Re: [Boost-users] building boost 1_33_1 now 1_37_0 > Message-ID: <h8qnpe$j3a$1(a)ger.gmane.org> > Content-Type: text/plain; charset=us-ascii > > Dan Smithers wrote: > >>>> How do I change the includes in jam? >>> Sorry, I don't understand the above. Boost.Build does not pass any >>> "intel includes" nor any "gcc 2.95 includes". Can you provide the commands >>> lines and errors in question and explain how exactly the command lines >>> should be altered. >> I've now found that as long as I make sure that gcc 3.4.3 is ahead of >> 2.95 in the search path I can get past the first hurdle. > > So, is there anything that still needs to be solved? > >> The command line that comes out is >> >> "icpc" -c -xc++ -O3 -Ob2 -w1 -fPIC -DBOOST_ALL_NO_LIB=1 >> -DBOOST_MATH_TR1_DYN_LINK -DNDEBUG -I"." -c -o >> "bin.v2/libs/math/build/intel-linux-release/threading-multi/hypot.o" >> "libs/math/build/../src/tr1/hypot.cpp" >> >> /usr/lib/pthread.h(163): error: expected a ")" >> extern int pthread_create(pthread_2 *__restrict __thread, > > Do you actually have file named /usr/lib/pthread.h? That is very strange > location for a header file. no, it's in /usr/include I am working on a computer that is not connected to the internet and had type the email. > >> I have written a short test program that includes pthread.h and that >> compiles. > > Anyway, it does not seem like anything wrong is passed to icpc. So, > this error suggests that it's not installed correctly, or some other > configuration issue. Do docs for the intel compiler say anything about > gcc version that is required? Nothing wrong is passed, but it doesn't compile. I don't have this problem building the non-boost code. thanks dan

1 0

Re: [Boost-users] Finding documentation for individual functions (uintmax_t)? An answer.
by s seelenluft 17 Sep '09

17 Sep '09

> in which library do I find documentation for the 'function': uintmax_t? > Since uintmax_t is such ubiquitous any search finds everything except its > actual description. Some more googling found the answer here: http://www.dinkumware.com/manuals/default.aspx Does such an equivalent, ie, a search function for functions/types only and not which does a full-text search, exist for Boost?

1 0