Date: Mon, 10 Nov 2008 01:30:56 +0300 From: "Alexander Churanov" <alexanderchuranov@gmail.com> To: "Peter Jeremy" <peterjeremy@optushome.com.au> Cc: Jeremy Messenger <mezz7@cox.net>, freebsd-ports@freebsd.org Subject: Re: Boost 1.37 released! Message-ID: <3cb459ed0811091430u46bb1eb9y5dccdab051e8a981@mail.gmail.com> In-Reply-To: <20081105180245.GA51239@server.vk2pj.dyndns.org> References: <20c4814a0811030712n4105c765p23983b55f729b25e@mail.gmail.com> <20081103192933.e4aeedef.ehaupt@FreeBSD.org> <3cb459ed0811031237l3adeef89t86a70030cca57fff@mail.gmail.com> <20081103223115.78fbbaef.ehaupt@FreeBSD.org> <3cb459ed0811031506g4b450865q34d33988a688c3@mail.gmail.com> <20081104002942.f136adec.ehaupt@FreeBSD.org> <op.uj2w7tm19aq2h7@localhost> <20c4814a0811050113i1bed72abu84b16c7343f93c4f@mail.gmail.com> <op.uj5k1xc59aq2h7@localhost> <20081105180245.GA51239@server.vk2pj.dyndns.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi folks! I'm glad to announce that I have working patch that makes possible to build boost-1.37 on FreeBSD. At present, I am investigating into port for boost-1.34 in order to create a working port for 1.37 with no regressions. Three things do not enable boost to compile out-of-box: First is the math library. It uses c99 and SUSv3 long double functions that are absent in 6.2-RELEASE and 7.0-RELEASE (for example. sinhl() ). I've checked out CURRENT source and found these functions, so probably 8.0-RELEASE would not suffer of this issue. The second is the test framework. Authors made error reports more detailed and rely on UNIX 03 features that provide information about the reason why a signal was delivered (see siginfo_t::si_code). FreeBSD 6.x does not comply with this in full and some signal-specific codes are missing. This is however, not a problem for 7.x branch - as I know all necessary constants are in place. The solution for the second issue is obvious: make error reporting in Boost.Test more coarse-grained on 6.x. I mean report that SIGILL was received, but do not tell what exactly has happened. My patch does exactly that. The Boost.Math issue is more complicated since we have either to disable the whole math library or to disable the part of it that deals with std::sinhl(), etc. A more expensive option is to drop support for particular long double variants of functions. Could anybody comment on that? Alexander Churanov
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3cb459ed0811091430u46bb1eb9y5dccdab051e8a981>