On Sun, Jan 22 2023, Volker Schlecht <openbsd-po...@schlecht.dev> wrote: > On 1/21/23 19:46, Theo Buehler wrote: >> To be clear, I'm not excluding that this is part of the solution to the >> problem. However, with this diff, I see the exact same segfault that I >> saw without any diff at all, so I do not think we should land this just >> yet. > > I don't have a CPU with the PKU register, and I was using the last amd64 > snapshot that didn't have this enabled as default. With that combination > it neither crashed, nor showed any of those warnings. Which is expected > behavior, I guess ... > >> My suggestion for a diff that would only affect the troublesome amd64 >> architecture and avoid any possible PIC issues on i386 would be the >> below: > > Tested with the same hardware setup and the latest amd64 snapshot, so > I couldn't see a crash coming from xonly if there was one. But here it > works fine, so OK from me.
IIUC tb's diff isn't ready for commit yet since it still results in a segfault on a PKU-enabled amd64 system. I do however can confirm that I also need USE_NOEXECONLY to avoid a segfault on riscv64. This was used in the previous bulk build and will be needed for the next bulk build too. ok? Index: Makefile =================================================================== RCS file: /cvs/ports/lang/node/Makefile,v retrieving revision 1.114 diff -u -p -r1.114 Makefile --- Makefile 21 Jan 2023 15:48:25 -0000 1.114 +++ Makefile 23 Jan 2023 11:30:08 -0000 @@ -2,7 +2,7 @@ ONLY_FOR_ARCHS = amd64 aarch64 i386 powe DPB_PROPERTIES = parallel USE_WXNEEDED = Yes -.if ${MACHINE_ARCH} == amd64 +.if ${MACHINE_ARCH} == amd64 || ${MACHINE_ARCH} == riscv64 USE_NOEXECONLY = Yes .endif @@ -13,7 +13,7 @@ PLEDGE_VER = 1.1.3 DISTFILES = node-pledge-{}${PLEDGE_VER}.tar.gz:0 \ ${DISTNAME}-headers.tar.gz \ ${DISTNAME}.tar.xz -REVISION = 0 +REVISION = 1 DISTNAME = node-${NODE_VERSION} PKGNAME = ${DISTNAME:S/v//g} -- jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE