Date: Mon, 22 Feb 2016 15:42:22 -0500 From: Lowell Gilbert <freebsd-ports-local@be-well.ilk.org> To: Andriy Gapon <avg@FreeBSD.org> Cc: Perry Hutchison <perryh@pluto.rain.com>, freebsd-ports@FreeBSD.org Subject: Re: unexpected package dependency Message-ID: <44mvqs4gup.fsf@lowell-desk.lan> In-Reply-To: <56CB2BD4.1040908@FreeBSD.org> (Andriy Gapon's message of "Mon, 22 Feb 2016 17:40:04 %2B0200") References: <56c43d57.Pot24goK72QkTKqk%perryh@pluto.rain.com> <56C45B9C.7090808@FreeBSD.org> <56c6760d.nR7fjvuf3gEK3yNY%perryh@pluto.rain.com> <56CB2BD4.1040908@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Andriy Gapon <avg@FreeBSD.org> writes: > On 19/02/2016 03:55, Perry Hutchison wrote: >> Andriy Gapon <avg@freebsd.org> wrote: >>> On 17/02/2016 11:28, Perry Hutchison wrote: >>>> I had not expected to find gcc listed (in packagesite.yaml) as a >>>> dependency of the sysutils/cpuburn package. I can understand a >>>> _port_ needing gcc (at build time), but does the cpuburn _package_ >>>> actually require gcc at _runtime_? >>> >>> I don't believe so. AFAIR, it builds static binaries. >> >> So would the inclusion of gcc in the "deps" for sysutils/cpuburn (in >> packagesite.yaml) be caused by a problem with the way the dependencies >> are specified in the port, or with the way they are handled by the >> package-generation mechanism? (I'm trying to figure out which to file >> a PR against -- and I'm not all that familiar with pkgng details.) >> > > My recollection is that the ports infrastructure does not allow to specify > whether a non-base compiler (like GCC for FreeBSD 11) is required only as a > compiler (that is, only during the build time) or if its run-time is required as > well. The latter is always assumed. USES_GCC doesn't support that, but using BUILD_DEPENDS with RUN_DEPENDS does. The downside to that is you have to specify a particular version when you otherwise would not have needed to do so. Adding a knob to bsd.gcc.mk to allow a port to say it doesn't need the RUN_DEPENDS would do the right thing. > But I could be mistaken. I've probably overlooked a lot of things, but I did *test* my suggestion, so I'm probably not completely wrong.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?44mvqs4gup.fsf>