Mathieu Malaterre

Pay attention that nifti uses 32bits (ieee) representation to store a quaternion matrix. Hence round-trip operation dicom → nifti → dicom may discard trailing digits in IOP/IPP.

Also for PET dataset it should compute a per instance rescale slope/intercept to cope with the dynamic range.

@boxerab there is nothing _ideal_ in your description. Memory consumption should be limited to what is actually needed (~code block height x scanline length).

Duplicate of https://github.com/uclouvain/openjpeg/issues/794

this may be a long debate, but at a minimum I would start with -ffloat-store and/or -std=c99 just to check this is not a silly mistake. x86_64 use SSE by...

no, simply a `-m32` and `-m64` should be enough (if linux/maxosx is your preferred OS)

@detonin could attach issue399/0.jp2 to this issue ?

@yanxxd can you reproduce CVE-2019-6988 using opj 2.4.0 ?

Patch is at: * http://git.ghostscript.com/?p=ghostpdl.git;a=commitdiff;h=f8d299c;hp=8e8831a02f172490d7b175e54c7d15814acacea4

Well the original statement was about a specific linking issue. I am guessing this is some kind of hardened libc implementation. I could not find a specific one line reproducer,...