Date: Mon, 17 Feb 2020 09:03:49 +1100 From: Dewayne Geraghty <dewayne.geraghty@heuristicsystems.com.au> To: freebsd-hackers@freebsd.org Subject: Re: Why is not llvm-config executable included? Message-ID: <0f258cc4-7aae-b757-add0-e44c329c8255@heuristicsystems.com.au> In-Reply-To: <20200215185821.GV4808@kib.kiev.ua> References: <D1A947BC-BCD9-4BAE-9D1B-EB1B433C1452@FreeBSD.org> <20200215185821.GV4808@kib.kiev.ua>
next in thread | previous in thread | raw e-mail | index | archive | help
On 16/02/2020 9:27 am, Konstantin Belousov wrote: > One of the reason why llvm in base should not be used as llvm infrastructure > is because llvm API and ABI is not stable across llvm releases, and exposing > that would make compiler updates in stable impossible due to the stable > branches guarantee of ABI stability. I think you're saying - don't build ports in a base (without the llvm port), rather use a separate build jail with llvm port. We used to build all ports in base and ship them to clients. Around FreeBSD6 we started to build within a jailed environment, as we supported i386 & amd64; something we still do. Though I'm concerned that perhaps we should migrate (our jails) back to using gcc where-ever possible, in the hope of avoiding future ABI incompatibility? In my experience replacing a base 'something' with the port of 'something' must be done carefully. (our experience with binutils and libressl, ultimately required removing all base "stuff" and using workarounds like softlinks - a bit messy, but scriptable) Would using gcc, in the build jail, provide better insurance against ABI breakage?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?0f258cc4-7aae-b757-add0-e44c329c8255>