update fails on 5d data with regular tiling

[Vlad Merticariu]

Updating a 5d array tiled regularly works when the update target is small, however it fails with larger areas.

To reproduce:

1. create the types and the collection

CREATE TYPE fivedmarray AS float MDARRAY [a0,a1,a2,a3,a4]
CREATE TYPE fivedset AS SET (fivedmarray)

CREATE COLLECTION test5D fivedset

2. insert dummy value and specify tiling

INSERT INTO test5D VALUES marray x in [0:0,0:0,0:0,0:0,0:0] values 1f TILING REGULAR [0:100, 0:100, 0:10, 0:10, 0:10]

3. update small area works

UPDATE test5D SET test5D[0:120,0:120,0,0,0] ASSIGN marray x in [0:120,0:120] values 2f

4. update larger area fails

UPDATE test5D SET test5D[0:3500,0:1500,0,0,0] ASSIGN marray x in [0:3500,0:1500] values 2f

[FATAL] - 05/04/2016 16:11:33.163592, minterval.cc:276: r_Minterval::intersects_with([0:3500,0:1500]) do not share the same dimension
[FATAL] - 05/04/2016 16:11:33.164065, minterval.cc:276: r_Minterval::intersects_with([0:3500,0:1500]) do not share the same dimension

[Dimitar Misev]

The update in the second case will generate 27977625324 B = 28 GB of tiles which is probably causing an issue of some sort.

The "fatal" message is a warning actually, it's not the real reason.

We need more info on what happened, does rasdaman segfault?

[Dimitar Misev]

I will close it as invalid use of the REGULAR tiling. In rasdaman 10 the regular tiling could be fixed to work in a better way.