Date: Sun, 29 Sep 2019 01:55:53 +0200 From: Jan Beich <jbeich@FreeBSD.org> To: freebsd-ports@freebsd.org Subject: Re: make.conf for a poudriere jail and CPUTYPE Message-ID: <pnjk-7z7a-wny@FreeBSD.org> In-Reply-To: <20190928224615.GH49516@bastion.zyxst.net> (tech-lists@zyxst.net's message of "Sat, 28 Sep 2019 23:46:15 %2B0100") References: <20190928135343.GE49516@bastion.zyxst.net> <r240-cxo8-wny@FreeBSD.org> <20190928145055.GF49516@bastion.zyxst.net> <y2y8-bcuc-wny@FreeBSD.org> <20190928220050.GG49516@bastion.zyxst.net> <20190928224615.GH49516@bastion.zyxst.net>
next in thread | previous in thread | raw e-mail | index | archive | help
tech-lists <tech-lists@zyxst.net> writes:
> On Sat, Sep 28, 2019 at 11:00:50PM +0100, tech-lists wrote:
>
> [...]
>
> in the meantime, for a test, I made a fresh poudriere jail, and tried to build
> lang/python36 with CPUTYPE?=btver1 in its make.conf and it fais the same way
Could be a compiler[1], kernel or hardware bug then. For compiler try using
clang80 from devel/llvm80 or upgrade to the upcoming FreeBSD 12.1. For kernel
try booting -CURRENT kernel (e.g., from snapshot) while keeping old world/base.
For hardware try reproducing on another machine or virtual machine.
Alternatively, track down which function returns unexpected result,
disassemble and/or minimize then ask for feedback on toolchain@ list.
To get debugging symbols build WITH_DEBUG=1.
Or maybe someone has better ideas.
--
[1] If the code uses instructions not supported by CPU then kernel
should return SIGILL (and dump core) but maybe CPython catches the
signal or similar.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?pnjk-7z7a-wny>