Merge gaps between subswaths for specific calibration routine data

[kautlenbachs]

Scene in question - S1A_IW_SLC__1SDV_20210722T005537_20210722T005604_038883_049695_2E58.SAFE

• Run the same on SNAP
• Run the same without AOI, all 3 subswaths included
• Run the same for coherence routine

[kautlenbachs]

Not present in SNAP. Present when no AOI

[kautlenbachs]

This happens during the merge. One of the sections zoomed out (between IW2 and IW3)

Debursted IW2 result

Debursted IW3 result

[kautlenbachs]

Same occurs on a scene with the same coverage - S1A_IW_SLC__1SDV_20210710T005537_20210710T005604_038708_049158_B74A

[kautlenbachs]

Same happens for coherence

[kautlenbachs]

No present when calibrating IW2 and IW3 AOI only.

[kautlenbachs]

Also present for perito moreno glacier scene

[kautlenbachs]

// TODO: add parallel_unsequenced_policy if minimal supported gcc version is changed to 9.1+ (Ubuntu 21.04 already uses 10+) topsar_merge.cc:134

[kautlenbachs]

Error happens when data is read from source (tile). Here can be seen a source read corrected by number of zeroes detected on a line after initial/original read (second read is done where offset is taken by number of zeroes detected - thats the initial offset/bug that is error). This is done in the bottom tile where data is fetched from a center subswath only. When read from both, a skewed pattern, where data read is not a rectangular but a parallelogram (seen in the middle upwards).

[kautlenbachs]

It is not write error, as can be seen below. All the tiles/data that is error initially is now filled with static values and they are correctly written in the merged result

[kautlenbachs]

All the tiles with error written into separate dataset:

[kautlenbachs]

Does also happen when about 3-4 bursts are selected when specifying AOI -

[kautlenbachs]

Source of the error is the case where source rectangle requests more data than there is in dataset.

But it is hard to tell the outer tile. It can detect that buffer sizes do not match, but it does not know which is off, either the width or height.

[kautlenbachs]

There was a broken golden image for end to end tests using maharashtra scene (S1A_IW_SLC__1SDV_20210722T005537_20210722T005604_038883_049695_2E58). rel diff

section of the originating scene

Other tests' golden ones did not need modification, hence current changes did not introduce any pixel offsets.