Alignment issues on 32-bit archs: TestAlignedBuff32 & TestAlignedBuffInt32 failures about nftables HOT 9 CLOSED

Confirmed building and testing fine with 318b4af on armel and armhf, in addition to amd64 and i386.

from nftables.

cc @thediveo who contributed alignedbuff

from nftables.

I suspect my test to be (obviously) incorrectly done, I need to test the test next time. Somehow the 32bit archs return to exact revenge on me. Unfortunately, I need to setup a test VM or similar first.

from nftables.

If you want to push a branch, commits, patches, etc. somewhere, I'm happy to run tests and report back if that lowers the barrier for you.

from nftables.

thank you very much! but I need to first explore this, as Go behaves differently from the C compilers in some aspects of 64bit versus 32bit architectures. nftables need to use the C alignment rules...

from nftables.

I've now set up a Debian 11 i386 VM on kvm/qemu ... the only bummer is that there's no 32bit versions of VSCode anymore. I can reproduce using go version 1.19.4 on linux/386.

from nftables.

If I'm not mistaken, then the expected length differs due to the uint64 padding being 4 on 32bit, compared to 8 on 64bit platforms.

from nftables.

With your commit, it builds and tests fine at least on amd64 and i386 indeed, thanks! Would like me to run some tests on arm* systems (I have access to a few porter boxes via my Debian hat, including armel and armhf).

(receipe→recipe × 2 by the way)

from nftables.

yes, please, if that is possible.

from nftables.