Ticket #802 (new defect)
Opened 2011-10-04T13:48:54-05:00
Last modified 2013-06-10T14:47:38-05:00
Imglib can hang up when memory is low
| Reported by: | bdezonia | Owned by: | bdezonia |
|---|---|---|---|
| Priority: | major | Milestone: | imagej2-b8-analysis |
| Component: | ImgLib2 | Version: | |
| Severity: | serious | Keywords: | |
| Cc: | Blocked By: | ||
| Blocking: | #1403 |
Description
Run IJ2 and open Boats. Now repeatedly Add Axis and choose axis size as 256. The number is not significant - other choices fail too. After as few as two iterations Add Axis fails. This is because memory runs low. IJ2 gets stuck in an imglib call. Its in the factory.create() code of RestructureUtils::createNewImgPlus() and it never returns.
Change History
comment:1 Changed 2011-10-05T11:52:43-05:00 by bdezonia
comment:2 Changed 2011-10-11T14:47:21-05:00 by bdezonia
Tried to duplicate with a stub program that only uses imglib code and cannot reproduce the problem. Factory create exceptions out with an out of memory exception like one would want.
comment:3 Changed 2011-10-13T10:49:16-05:00 by bdezonia
This issue happens if you create a new 20k x 20k image too.
comment:4 Changed 2011-10-14T13:39:30-05:00 by bdezonia
Ignore last comment. That issue is probably bug #825.
comment:5 Changed 2011-11-30T11:24:06-06:00 by bdezonia
Debugging this is puzzling. Imglib code looks fine. No exceptions being caught. The debugger seems to switch threads and is in the middle of a FutureTask thread where it is stuck. This might be a Java bug of some kind.
comment:6 Changed 2012-02-27T11:00:58-06:00 by curtis
- Milestone changed from imagej-2.0-beta1 to imagej-2.0-beta2
comment:8 Changed 2012-05-22T13:46:13-05:00 by bdezonia
Note that this is easy to trigger when run from Eclipse (not getting a lot of memory from launcher).
comment:10 Changed 2012-07-12T10:57:17-05:00 by bdezonia
- Milestone changed from imagej-2.0.0-beta3 to imagej-2.0.0-beta4
comment:13 Changed 2012-08-16T12:27:33-05:00 by bdezonia
I found a similar issue that I fixed in the SamplerService (#1374). If you call ImgPlus::initializeColorTables() with a large enough number Imglib gets stuck too. Again trying to allocate a lot of memory. A Future Task is spawned but it never returns. We need to find why exceptions aren't thrown and caught in these cases.
comment:14 Changed 2012-09-10T09:11:19-05:00 by bdezonia
- Milestone changed from imagej-2.0.0-beta4 to imagej-2.0.0-beta5
comment:15 Changed 2012-11-07T15:04:26-06:00 by bdezonia
Yet another example of silent failures: change PrimitiveDoubleArray to allocate memory not as (top * 2) but (top + top/2). Now run StatisticsDemo on embryos (or organ of corti) and try to calc median. Nothing happens. PrimitiveDoubleArray is unable to get a hold of the ram it needs. For some reason Imglib does not throw an exception about not being able to get such ram.
comment:16 Changed 2013-06-07T16:01:19-05:00 by bdezonia
- Milestone changed from imagej2-b7-ndim-data to imagej2-b8-analysis
comment:17 Changed 2013-06-10T14:47:38-05:00 by bdezonia
Can no longer duplicate the Add Axis hang. It just reports OutOfMemoryException.
Also may manifest itself as #746