[DNGDecoder] Images consisting of a single padded tile are well-formed

"Topaz Photo AI" seems to be producing such rather weird files. Looking at the actual images, it's *clearly* a bug in said software. But there is an argument to be made that [DNG] tiles can have padding rows/cols, and while it may seem, err, non-conventional, to have a single tile with size much larger than the actual image size, that isn’t exactly illegal. https://discuss.pixls.us/t/darktable-dxo-topaz-raw-denoise-dngs/37127/14 Co-authored-by: Miloš Komarčević <[email protected]>
darktable-org · May 19, 2023 · 9c5e0f9 · 9c5e0f9
1 parent 8f42ba6
commit 9c5e0f9
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/librawspeed/decoders/DngDecoder.cpp b/src/librawspeed/decoders/DngDecoder.cpp
@@ -334,7 +334,7 @@ DngDecoder::getTilingDescription(const TiffIFD* raw) const {
                            ? raw->getEntry(TiffTag::ROWSPERSTRIP)->getU32()
                            : mRaw->dim.y;
 
-  if (yPerSlice == 0 || yPerSlice > static_cast<uint32_t>(mRaw->dim.y) ||
+  if (yPerSlice == 0 ||
       roundUpDivision(mRaw->dim.y, yPerSlice) != counts->count) {
     ThrowRDE("Invalid y per slice %u or strip count %u (height = %u)",
              yPerSlice, counts->count, mRaw->dim.y);