From: Theo Buehler Subject: Re: [BUG] net/iamb: segfaults in latest 8.0 snapshot To: ports@openbsd.org Date: Mon, 5 Oct 2026 19:41:20 +0200 On Mon, Oct 05, 2026 at 05:52:37PM +0100, Stuart Henderson wrote: > On 2026/10/05 18:11, Theo Buehler wrote: > > > #0 0x00000596746d498f in aws_lc_0_45_0_curve25519_x25519base_alt () > > > > That's going to be the usual aws-lc xonly thing > > indeed, that was a new dep in the last update > > $ cvsps -s 11 -g | grep aws > +MODCARGO_CRATES += aws-lc-rs 1.18.1 # ISC AND (Apache-2.0 OR ISC) > +MODCARGO_CRATES += aws-lc-sys 0.45.0 # ISC AND (Apache-2.0 OR ISC) AND Apache-2.0 AND MIT AND BSD-3-Clause AND (Apache-2.0 OR ISC OR MIT) AND (Apache-2.0 OR ISC OR MIT-0) > +SHA256 (cargo/aws-lc-rs-1.18.1.tar.gz) = soHTB1iNY03pIIdIkHMmWeLnZy9yteEOgbrcGoqDYh4= > +SHA256 (cargo/aws-lc-sys-0.45.0.tar.gz) = m/9sO1T615ouYLgQLK9WWBlxFJfB9fCS9JUI4vXDGyc= > +SIZE (cargo/aws-lc-rs-1.18.1.tar.gz) = 249534 > +SIZE (cargo/aws-lc-sys-0.45.0.tar.gz) = 9725710 > > > Index: Makefile > > =================================================================== > > RCS file: /cvs/ports/net/iamb/Makefile,v > > diff -u -p -r1.10 Makefile > > --- Makefile 26 Sep 2026 16:54:28 -0000 1.10 > > +++ Makefile 5 Oct 2026 16:11:02 -0000 > > @@ -2,11 +2,18 @@ > > NOT_FOR_ARCHS = sparc64 > > BROKEN-i386 = compile fails, possibly out of memory ("fatal runtime error: Rust cannot catch foreign exception") > > > > +# aws-lc-sys has constants in .text > > +# https://github.com/awslabs/s2n-bignum/pull/242 > > +.if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "aarch64" > > +USE_NOEXECONLY = Yes > > +.endif > > I'm ok with that, but I wonder if we should automate it, or if that's > too much magic. Bumps included for all the ports pulling in aws-lc-sys > that don't yet set USE_NOEXECONLY. Not against applying a big hammer until this is resolved but I'm not sure all the things listing aws-lc in crates.inc actually use the aws-lc backend. For example, ncspot uses native-tls iiuc and rust-openssl-tests doesn't ship compiled code at all (it won't matter if it's tagged USE_NOEXECONLY, but the bump could be omitted).