I've found a cause of problems with HARMONY-5617 patch. Extensive testing has showed that shared memory is not a good choice for sharing data between static library instances - shared segments are not deleted after program termination, and there is a system limit for shared objects.
Now I'm working on another way - with using shared library as a placeholder for unique data. But this new approach will also require accurate testing, so I propose making M6 without HARMONY-5617 changes. Thanks, Ilya. 2008/4/25, Tim Ellison <[EMAIL PROTECTED]>: > Ilya Berezhniuk wrote: > > > I would like to finish HARMONY-5617. Now the problem is just workarounded. > > > > Although it's a bugfix, actually it includes changes in VM signal > > handling to make it platform-independent, and moving stack and guard > > page operations and OS-level threading to porting layer - so I think > > it should be considered as a feature. > > > > The patch is almost ready - I'm now investigating the last > > intermittent problem with SIGUSR2 handling. Also I want to check > > performance impact of the patch - this can take few days. > > > > Thanks for the heads-up Ilya. Clearly there is work in-flight that we need > to complete before calling for the feature freeze. > > Let's wait to see if anyone else has any major items in their immediate > plans, and set a feature freeze date based on what we learn (i.e. it will > likely be during next week). > > Regards, > Tim >
