Review: Fix allocation and stride bugs in ImageInput::read_tiles. #373
Conversation
When reading tiled images, where the image is not a whole number of tiles, this bug could misallocate and run off the end of the buffer while dealing with the partial tiles on the right or bottom edge of the image.
|
Updated with another fix to some bugs related to strides inside read_scanlines() for images with differing data formats per channel. Sorry, these are in some ways separate issues, but I was chasing down what I thought initially was one bug but turned out to lead to both of these problems. Forgive me. |
There was a problem hiding this comment.
What does this 0 mean here? Should it be related to z-depth instead?
There was a problem hiding this comment.
This is only converting within one z "plane", the zstride is not used. So I passed 0 instead of the more complicated expression that would be the correct zstride if the call needed it.
|
I just realized, however, that I may need to make the same fix when reading tiled images. Stay tuned for a possible update. |
|
Nah, the fix I already did is desperately needed. I'll address the tiled case separately after I get this fix out. |
…369/noop_profile_yaml Fix issue where loading profiles with yaml-cpp 0.5.x would crash.
2 participants
When reading tiled images, where the image is not a whole number of tiles,
this bug could misallocate and run off the end of the buffer while dealing
with the partial tiles on the right or bottom edge of the image.