Review: Bug fixes with per-channel data formats, lots of offsetting problems - #115
Conversation
|
Cool, i'll check it out tomorrow. I assume this was intended to fix the jrray itview crash, correct? It looks like these two commits are in files related to image writing. But the itview crash is on image load. Perhaps there are two issues? |
|
Right, right. The fix I posted so far fixes your test conversion program that goes through floats. Still stuck on half input, forgot to double check that, hang on... |
|
Updated the patch to fix the input problems as well as output. |
|
Thanks, Larry, I'll check this out ASAP. |
|
Negative, the bug persists. |
|
Leszek, could you email me the RLA that causes the problem? And make sure your latest RLA code is pushed to your GH repo, and I'll take a look at the problem. In the mean time, I think the patch as it stands fixes OpenEXR per-channel formats as well as some necessary infrastructure in ImageOutput. |
|
All the test rlas are here: |
|
Yup, all my up-to-date RLA code is in my RLA branch on GH, you just need to switch the two "#if 0"s to "#if 1"s. You can see the issue in any 16-bit RLA file (disregard the other artifacts that are visible in imgmake_rgba_nc16.rla, I'm working on resolving those; funny that the ginsu image displays fine). |
|
Let's take the RLA discussion off-line (into email), I don't think it's related to this ticket. Jeremy, are you able to confirm that this patch works for your OpenEXR issues? |
|
Great, I'm rolling it into all three branches (and testing one last time). |
3 participants
I'd like to backport this to 0.10 as well.
Jeremy and Leszek, I think this will solve both of your (independently found) problems.