Peter Lindstrom

You should not have to cast all operands as the integral promotion rules should do any necessary casting automatically. Or the compiler, not zfp, is broken. Regarding different sizes, I...

Ah, this indeed is necessary as the integral promotion does not extend the right-hand-side operand of `&` with one-bits, as needed to keep all the MSBs in the 64-bit left-hand-side...

I pushed some changes to `develop` that implement `round_up` via "C templates." This hopefully takes care of any remaining bugs.

I'll defer to @shaomeng as I have little knowledge about the implementations of SPERR and its HDF5 filter, but my initial reaction is that the SPERR CLI uses Fortran ordering...

Indeed, that looks like a faulty assumption that integer overflow wraps. One obvious fix would be to use unsigned arithmetic here, though that requires some changes that propagate across a...

It seems we're about to open up Pandora's box. These UB issues are not confined just to signed integer overflow in the reversible zfp algorithm, but the same can evidently...

I believe this should be fixed in zfp, as valid input data should never invoke UB. This is somewhat different from #242, where the user violates zfp's assumptions on all-numerical...

@mjwillson: #241 is being addressed on the `bugfix/integer-overflow` branch. Please check if the fix works for you. We're currently struggling with tests passing in general because of outdated scripts and...

@mjwillson I would like to merge this fix into `develop`, but I would prefer to hear from you first to make sure it addresses all of your concerns.

I have merged this fix and am closing the issue. Please reopen if it does not address your needs.