Björn Harrtell
Björn Harrtell
I will surely try as this is blocking https://github.com/postgis/postgis/pull/583.. but unfortunately I lack knowledge of flatbuffers internals so it will likely not be easy for me. Also I find C...
Hmm ok. I'm very far from knowing what I'm getting into.. :) but.. an immediate reaction is to this part: https://github.com/dvidelabs/flatcc/blob/a889e487609ec0f2b6aa9b6a920bc8129a8b547f/src/runtime/builder.c#L1380-L1382 ... in that the get_min_align and set_min_align does not...
@mikkelfj thanks for trying to explain - but it will take a long while before I understand all of it. :) Meanwhile, I have gotten closer to find out more...
@mikkelfj your note about tail padding the buffers is interesting and I've thought about it before. It is likely and unfortunately a design flaw in my format (FlatGeobuf) which is...
@mikkelfj agreed that would be nice. I believe tail padding is also neglected in other language supports including the reference C++ (but I might be wrong).
Hmm interesting! :) Oh well.. I still have work to do.. As I wrote I will see if I can confirm your findings about the potential verifier bug by verifying...
Thanks for your analysis @mikkelfj. I will see if I can confirm this by verifying the buffers with the C++ implementation.
@mikkelfj ok :( But I went ahead and did it anyway and got strange results... The "empty" doublevec table verifies, i.e this passes: ```c++ const unsigned char flatbuf2[] = {...
I also tried producing buffers with the same content using C++ to compare. I know it doesn't have to be the same binary representation for the same content and I...
@mikkelfj while I'm far from sure about my findings above and not sure exactly what they mean it doesn't seem to confirm that the problem we see here with flatcc...