This might be attributed to the users responsibility: If a user generates a very large network (i tried a regular lattice with 50 nodes per side) the program freezes and starts eating up memory until it crashes.
One might think about offering a graceful recovery from this, or maybe thread the netzwork creation?
| Status | Assigned | Task | ||
|---|---|---|---|---|
| Resolved | None | T11549 Create the 2012 MITKDiffusion release | ||
| Wontfix | kislinsk | T11724 Connectomics: Freeze when generating large random networks |
Moving the generation algorithm to another thread would definitely be an improvement, though probably not something I would do for this release.
As for the crash that is currently an MITK wide problem, as any data set that eats up all the memory usually crashes MITK, this issue has been mentioned in T6264 .
Considering a regular lattice of 50 nodes side length contains roughly 125000 nodes with 666552 edges I would put this down to user error. A graceful exit would certainly be better, I just do not know whether I will do anything about it for this release.
[4295fe]: Merge branch 'bug-11724-warn-about-possible-crash'
2012-04-20 17:09:48 Caspar Goch [019949]
set limit to 5000 and added missing break
2012-04-20 16:33:55 Caspar Goch [19ccf4]
Added warning when trying to generate very large networks
This is now partly fixed, further improvements will be done at a later date, definitely post-release.