Fix TileOffsets for DNGs containing one tile

[SimonSegerblomRex]

The old code looks messed up.

Fixes #6575

[SimonSegerblomRex]

At least this fixes the issues with decoding the file that I uploaded, but I'm not 100 % sure about what's happening... Let me know if you have any image files for regression testing.

[SimonSegerblomRex]

Removing this will affect the logic here:

RawTherapee/rtengine/dcraw.cc

Lines 7065 to 7066 in 2ce5b82

	if (!tile_width ) tile_width = INT_MAX;
	if (!tile_length) tile_length = INT_MAX;

...but I'm not sure why you would like to set tile_width/tile_length to 0 or INT_MAX...

[SimonSegerblomRex]

Seems like this breaks lossless JPEG decoding for some reason...

[SimonSegerblomRex]

Removing

RawTherapee/rtengine/dcraw.cc

Lines 1136 to 1137 in 2ce5b82

	if (tile_length < INT_MAX)
	fseek (ifp, get4(), SEEK_SET);

seems to solve the problem for lossless JPEG encoded single tiled DNGs, but will at the same break things for some multi-tiled images...

It might also be worth looking into the logic deciding between the two functions lossless_jpeg_load_raw and lossless_dng_load_raw.

[SimonSegerblomRex]

At least this image works again with the latest patch set.

EDIT: This image is also fine.

[Thanatomanic]

I must say, you're a brave man trying to disect and change dcraw.cc code! I will need some time to do proper checking here, especially since this code is embedded deep in the bowels of the beast, so to speak...

[SimonSegerblomRex]

I must say, you're a brave man trying to disect and change dcraw.cc code! I will need some time to do proper checking here, especially since this code is embedded deep in the bowels of the beast, so to speak...

Yes, we need to be really careful... I'm not even sure if this problem is related to the singled tiled nature of the image I uploaded or something else. At least I think this change will only affect images were tile_length was set to INT_MAX (after first being set to 0...). There is one more occurrence of tile_length < INT_MAX that we might need to update.

[Beep6581]

Tagging this for 6.0 as it touches radioactive elements. Let's merge as soon as 5.10 is out to allow as much time for testing as possible.

[Beep6581]

@SimonSegerblomRex can you provide a list of affected cameras?

[SimonSegerblomRex]

@SimonSegerblomRex can you provide a list of affected cameras?

I actually don't know of any cameras affected by the bug that I tried to fix here, but a lot of cameras could potentially be affected by this change, so I would be really careful merging this without thorough regression testing. Maybe it's better to convert this pull request to draft.

[Lawrence37]

5.10 has just been released. Now is a good time to merge if this is ready for broader testing.