[Ady0333]

Problem

In raster/r.lake/main.c the statement that plants the seed point sits outside the braceless if (sdxy_opt->answer) block that is meant to guard it, so it runs on every invocation:

if (sdxy_opt->answer)
    /* Check is water level higher than seed point */
    if (in_terran[start_row][start_col] >= water_level)
        G_fatal_error(...);
out_water[start_row][start_col] = 1;

start_row and start_col are initialised to 0 and are only assigned in the coordinates= branch. The parser makes seed= and coordinates= mutually exclusive, so on every seed=<raster> run the line above sets out_water[0][0] = 1, adding a seed at the north-west corner cell of the computational region that the user never asked for.

The stray seed is not inert. If the terrain at that corner is below water_level, the flood fill spreads from it exactly as it would from a real seed, producing a second water body unconnected to the seed map. If the corner is above water_level it is cleared on the first sweep and nothing happens, which is why the defect is easy to miss.

Impact

The tool exits 0 and emits no warning, so the result looks correct. The reported Lake area and Lake volume include the spurious body, and the documented workflow of feeding a lake map from one run back in as the seed for the next carries the contamination forward.

Reproduction

A 20x20 region with a basin in the south-east at elevation 5, a separate low area in the north-west at elevation 8, a ridge at elevation 50 between them, one seed cell inside the south-east basin, and water_level=10. Only the south-east basin is reachable from the seed, so the expected result is 36 wet cells in a single water body.

	before	after
wet cells	72	36
distinct water bodies	2	1
north-west cell (2.5, 17.5)	2	NULL
reported lake area	72.000000	36.000000
reported lake volume	252.000000	180.000000

Fix

Add the braces so the seed point is set only when coordinates are given. With seed=, out_water is already fully populated from the seed raster, so nothing else is needed.

Tests

r.lake had no tests. This adds raster/r.lake/tests/r_lake_test.py with three cases covering the seed map path: total wet cells, the unseeded low area staying dry, and the number of distinct water bodies. All three fail on main and pass with the fix.

The coordinates= path is unchanged. Seeding inside the basin still yields 36 cells, and seeding on the dry ridge still fails with Given water level at seed point is below earth surface.

[echoix]

I'm waiting for the CI results, the description and changes seem fine

[marisn]

Looks good to me, but I can not judge Python code in tests, as I am not that familiar with latest trends in GRASS Python code writing.