[PierreRaybaut]

Follows DataLab-Platform/Sigima#110 (same branch name, so CI uses that Sigima branch).

Summary

Sigima now does two things differently:

• It gives the analysis results of images with non-uniform coordinates in physical coordinates. Concerned: Centroid, Minimum enclosing circle center, 2D peak detection, Contour detection, Circle Hough transform and Blob detection. These results were given in pixel indices, or according to a stale uniform calibration, so they were drawn away from the features they describe.
• It refuses to extract regions of interest holding no pixel of the image.

The contracts of these operations are at version 2.

• Provenance: provenance_desktop_families.h5 is regenerated: centroid and peak_detection are recorded in version 2 and replayed exactly. provenance_web_families.h5 still records version 1 for them. These activities are refused, as checked by the data-driven test, until DataLab-Web regenerates the file.
• ROI > Extract... shows an explicit error for each region of interest holding no pixel (e.g. ROI grid cells narrower than a pixel) and still extracts the others, instead of creating empty images. Checked in an offscreen window: a 4×4 image with the default grid gives 4 extracted images and 5 error messages.
• Release notes: two bug-fix entries, with their French translations.

Validation

With the Sigima branch:

• provenance tests (common, hdf5): 123 passed;
• image features, processing preview, replay-only features, signal ROI, example scenario, Web API extra and NeXus import tests: 157 passed, 3 skipped;
• families tests after regeneration: 4 passed.