Adam Stylinski

There was a zlib API change since the release. Are you sure that that crate isn't trying to build against zlib 1.2.11?

Right but often crates and other self contained build systems will pack their own specific version of a library with it. Arch definitely uses 1.2.12 and if you are sure...

Those pkgbuilds are not enough to go on. Looking at the source to that crate I can't even find a direct reference to zlib, so it has to be somewhere...

Where are you seeing alignment issues? Arm has no such alignment requirement but even so the neon implementation of adler32 does scalar sums until it's aligned to 16 bytes for...

What instruction is 0x320 bytes into the adler32_neon function? It'd be good to know which load is causing your issues. As far as I can tell though everything should be...

Also interesting is that the address it's faulting on modulo 16 is zero. That means that even widest possible load would be a naturally aligned one. This sigbus makes no...

@asuka-mio is this reproducible with a debug build? Might it be possible to at least provide a core dump for some post mortem? On paper it seems like this shouldn't...

Any way you can get me the final binary as well? I should be able to attach the coredump in gdb and inspect the state of your heap and stack...

If it's not crashing with direct linking it seems very likely to not be an issue with the adler32 function itself but with an API incompatibility. It seems probable that...

Just to double check your use case as close as possible (I'm on aarch64 and using Linux, not android, but perhaps this is a good enough test): ``` [adam@odroid ~]$...