Ivan, I tracked down the problem to a difference between the generic and optimized versions of the P256 implementations. I'm not 100% sure of the correct fix, so I submitted it as golang.org/issues/20215 with a link to your example.
- Todd On Tuesday, May 2, 2017 at 12:33:56 PM UTC-5, Ivan Voras wrote: > > Hi, > > I'm using ECDSA with the P256 curve and I've found a problem which I've > tracked to a different result of signature verification on the Raspberry Pi > and x64 (both Linux and Windows). Later I've verified that i386 behaves as > RPI. I've tracked the problem down to this self-contained case: > > https://gist.github.com/ivoras/32b2abd16b5984fa43c006486bfb7e2c > > It's trivial to build. When run on the RPI or i386, verification succeeds. > On x64, it fails. RPI is running little-endian Ubuntu 16.04.2, 32-bit. > > Am I doing something wrong or is it a bug in the libs? I've updated all > systems to Go 1.8.1, but I've noticed this behaviour on 1.7 (and the update > didn't solve it). -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.