Date: Fri, 8 Apr 2011 17:52:46 +0000 From: "b. f." <bf1783@googlemail.com> To: Alexey Dokuchaev <danfe@freebsd.org> Cc: Rene Ladan <rene@freebsd.org>, cvs-ports@freebsd.org, cvs-all@freebsd.org, ports-committers@freebsd.org Subject: Re: cvs commit: ports/www/chromium Makefile Message-ID: <BANLkTim2%2BNq9QaPpKQYjPmtnfOf5bBt7GA@mail.gmail.com> In-Reply-To: <20110408153710.GA28200@FreeBSD.org> References: <201104072239.p37Md9oh031253@repoman.freebsd.org> <BANLkTikmUZsHNjw2Jpe8Y7v5CD6WkZGtoQ@mail.gmail.com> <20110408022459.GA40411@FreeBSD.org> <20110408023021.GA47293@FreeBSD.org> <BANLkTik4HrVvVNU_nBtH8Oyt=81uKjn6og@mail.gmail.com> <20110408035430.GA56549@FreeBSD.org> <BANLkTinCm=6zuzopMKi1v=5ehohOGM5nYg@mail.gmail.com> <20110408153710.GA28200@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 4/8/11, Alexey Dokuchaev <danfe@freebsd.org> wrote:
> On Fri, Apr 08, 2011 at 12:24:57PM +0000, b. f. wrote:
>> I guess that we should just trust users to configure them properly.
>
> Yes, I also believe that if user wants to build a package for machine other
> than theirs, they are expected to know how these things work and what to do
> to get the work done. From our side, we should provide easy and straight-
> forward means of such customizations.
>
> Summarizing our discussion, what do you think of the following?
>
> .if ! ${MACHINE_CPU:Msse2} || defined(PACKAGE_BUILDING) && ${ARCH} !=
> "amd64"
> GYP_DEFINES+= disable_sse2=1
> .endif
>
> Default package would have reasonable defaults now, custom port build can
> be controlled by passing desired MACHINE_CPU. Building SSE2-enabled
> package for i386 is a bit tricky, but cannot be possibly polluted by
> stale cached OPTIONS value.
Sorry, I tried to convey in my previous message that I was mistaken in
thinking that the earlier changes would cause problems for the default
i386 packages. Packages for the machine hosting the build, and those
built for generic machines via the portbuild scripts, have the right
settings now, so I would leave the port as it is. As you wrote, those
who are building packages for machines other than the host should
ensure that CPUTYPE has the desired value (they should also ensure
that all OPTIONS and other relevant variables in included Makefiles
have the desired values, so the argument about build pollution for
only this knob is a bit overstated), so I think that adding the change
above just creates an unnecessary obstacle to building SSE2-enable
packages on i386 when such a build is desired.
b.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BANLkTim2%2BNq9QaPpKQYjPmtnfOf5bBt7GA>