BUG: ROIs not deleting on Nightshade

[rkferguson]

Michael reported that he and Sandra were working on an image with an ROI propagated across 300 time points. They deleted the ROI on 100 of the time points and this appeared to be successful, but when they ran Michael's matlab script on the image, it analysed all of the original ROIs over the 300 time points, so obviously the ROIs had not been deleted on the server.

Michael then tried a test with an image of his ID 4610 and drew and then deleted an ROI that was on it. The delete progress in the activity window continues to spin, and although the ROI has disappeared from the viewer, it was still visible when opened in another window or in the web client. I could also see it logged in as myself looking at Michael's image.

Image ID 4610 - ROI "deleted" 22/5/14 at about 1540 hrs. on Nightshade.

[jburel]

#12248 describes the same issue.

[jburel]

Merging content of #12248. Similar problem
Bug found by Sandra using Nightshade.
Has image with 305 time points

• creates ROI and projects it across all time points
• once analysis done deletes ROI
• disappears from ROI Manager window
• creates new ROI, projects across all time points
• closes viewer
• saves changes when prompted
• new ROI is not saved

I suspect this is because the deletion of the old ROI is still being processed, so new one not actually saved before thread is killed.

Need some kind of block and warning e.g "Cannot close this window until save has been completed".

[jburel]

Fix error while turning a line into a polyline. Problem noticed while investigating this ticket.
see ​https://github.com/openmicroscopy/openmicroscopy/pull/2616

[mtbcarroll]

Conversation with Michael indicates a rarely occurring but long-standing problem in which ROI updates in Insight are not available to mtools from nightshade for many minutes.

[mtbcarroll]

I'll give this back to JM for now, as he's been looking at this code lately and so far I'm not managing to reproduce this problem at all.

[jburel]

Discussed today with Michael, The shapes are deleted. It was explaining a delay that he noticed when creating ROI in insight (Java) and being able to access them in mtools (matlab)
Next step: write a script in java creating ROI+save and a script in matlab to check if they are immediately accessible

Balaji: did you notice some strange behaviour when creating/accessing ROI?

[jburel]

Some ROIS were also created using mtools.
Balaji will have a look when J-M is away

[bramalingam]

Did some independent testing (using the matlab toolbox) to see if the ROI service is getting updated in time.
There seems to be a lag (non-linear lag), potentially based on the number of operations queue'd on the server,
in retrieving the ROI's linked to the image.

Here's a small snippet of code that was used to do the test,

% Create the roi.
roi = omero.model.RoiI;
ellipse = createEllipse(10, 10, 5);
% Indicate on which plane to attach the shape
setShapeCoordinates(ellipse, 0, 0, 0);

roi.addShape(ellipse);
roi.setImage(omero.model.ImageI(imageId, false));

%Retrieve the roi linked to an image

%%%%%%Non linear pause required at this stage , tried : pause(3)%%%%%%%%%%%

service = session.getRoiService();
roiResult = service.findByImage(imageId, []);
% Save
iUpdate = session.getUpdateService();
roi = iUpdate.saveAndReturnObject(roi);
rois = roiResult.rois;
n = rois.size;

The test was able to retrieve the ROI's post this non-linear lag.

[jburel]

Ran similar test in Java, I did not notice any issue.

[jburel]

Ran test in Matlab too but I cannot see the problem reported by balaji.
If the measurement tool is opened, the roi created in matlab will not be visible.

[jburel]

Tested call in Java and Matlab. I have not been able to reproduce the issue.
I discussed the issue with Michael, rois are always visible in insight but not in mtools. The problem is also difficult to reproduce.
Passing it to Michael to investigate mtools

[jamoore]

Balaji: can you explain your Matlab code above? I don't see any "saves" before your non-linear wait.

[jburel]

We have not been able to reproduce the issue. Marking it as invalid. We can re-open if we see it happening again