Re: [Boost-users] boost::interprocess boost::interprocess xsi_shared_memory
Is there a reason for not finishing the implementation of XSI (System V) shared memory? It seems like most of the work is already done. If there is no reason not to, I might be able to finish the implementation so that it can be included in the next release. I know that POSIX may be sufficient for some, but System V allows for much better control using SELinux since it lives in kernel space. FYI: This is in response to http://thread.gmane.org/gmane.comp.lib.boost.user/64537. Thoughts? Thanks. - Daniel
El 20/12/2010 19:41, Daniel Neuberger escribió:
Is there a reason for not finishing the implementation of XSI (System V) shared memory? It seems like most of the work is already done. If there is no reason not to, I might be able to finish the implementation so that it can be included in the next release.
XSI was supposed to be available to have a posix-like portable shared memory semantics for systems with only system v shared memory. However lack of time and portability issues froze the development. Maybe some class (xsi_shared_memory) is ready to be public, just like we have windows_shared_memory. It shouldn't be hard to implement managed_xsi_shared_memory. If there is interest in them I could add them and try posix emulation on top of XSI later. Best, Ion
XSI was supposed to be available to have a posix-like portable shared memory semantics for systems with only system v shared memory. However lack of time and portability issues froze the development. Maybe some class (xsi_shared_memory) is ready to be public, just like we have windows_shared_memory. It shouldn't be hard to implement managed_xsi_shared_memory. If there is interest in them I could add them and try posix emulation on top of XSI later.
Thanks for your quick reply, Ion. Yes, we are interested in both xsi_shared_memory and managed_xsi_shared_memory. I am also willing to help with the development if that would be helpful. Could you give me an idea of when a patch could be available for both of these so that I can decide whether we need to develop a workaround? Thanks again. - Daniel
El 21/12/2010 19:23, Daniel Neuberger escribió:
XSI was supposed to be available to have a posix-like portable shared memory semantics for systems with only system v shared memory. However lack of time and portability issues froze the development. Maybe some class (xsi_shared_memory) is ready to be public, just like we have windows_shared_memory. It shouldn't be hard to implement managed_xsi_shared_memory. If there is interest in them I could add them and try posix emulation on top of XSI later.
Thanks for your quick reply, Ion. Yes, we are interested in both xsi_shared_memory and managed_xsi_shared_memory. I am also willing to help with the development if that would be helpful. Could you give me an idea of when a patch could be available for both of these so that I can decide whether we need to develop a workaround?
If you just need those two classes, I think they could be ready in two weeks, I'll reserve some time for this. Best, Ion
Thanks for your quick reply, Ion. Yes, we are interested in both xsi_shared_memory and managed_xsi_shared_memory. I am also willing to help with the development if that would be helpful. Could you give me an idea of when a patch could be available for both of these so that I can decide whether we need to develop a workaround?
If you just need those two classes, I think they could be ready in two weeks, I'll reserve some time for this.
Awesome. Thanks Ion. That's exactly what we need. I'm fairly sure the other two things we need will be supported by the above. The first is using xsi_shared_memory with mapped_region and it looks those changes are already done. The second is being able to choose which implementation of shared memory we're using (System V over POSIX), which would be as simple using the right classes. If I'm misunderstanding anything though, please let me know. Thanks again for your help and please let me know if there is anything I can do to help. - Daniel
El 22/12/2010 15:47, deuberger escribió:
Thanks for your quick reply, Ion. Yes, we are interested in both xsi_shared_memory and managed_xsi_shared_memory. I am also willing to help with the development if that would be helpful. Could you give me an idea of when a patch could be available for both of these so that I can decide whether we need to develop a workaround?
If you just need those two classes, I think they could be ready in two weeks, I'll reserve some time for this.
Awesome. Thanks Ion. That's exactly what we need.
I'm fairly sure the other two things we need will be supported by the above. The first is using xsi_shared_memory with mapped_region and it looks those changes are already done. The second is being able to choose which implementation of shared memory we're using (System V over POSIX), which would be as simple using the right classes. If I'm misunderstanding anything though, please let me know.
Yes, mapped region supports mapping System V shared memory
Thanks again for your help and please let me know if there is anything I can do to help.
I think I need to add some tests and some minor tweaks. If I see the work is gettin too big, I'll ask for help. Best, Ion
Hi Ion, I saw that you made a commit for XSI shared memory (https:// svn.boost.org/trac/boost/changeset/67451). Could you confirm whether it is ready to use? We're using Boost 1.43; do you think applying that changeset would be sufficient to be able to use XSI shared memory? Lastly, do you think these changes will make it into the upcoming 1.46 release? Thanks again, - Daniel
participants (4)
-
Daniel -
Daniel Neuberger
-
deuberger
-
Ion Gaztañaga