[CFT] boost 1.55.0

[Justin Hibbits]

On Sat, 14 Dec 2013 20:51:03 +1100
Kubilay Kocak <koobs.freebsd at gmail.com> wrote:

> On 14/12/2013 8:47 PM, Kubilay Kocak wrote:
> > On 13/12/2013 7:47 PM, Baptiste Daroussin wrote:
> >> Hi,
> >>
> >> I have been deseperatly waiting for someone to handle updating
> >> boost to the latest version, this didn't happened so I did it.
> >>
> >> http://people.freebsd.org/~bapt/boost-1.55.0.diff
> >>
> >> Changes compared to previous version (from the ports point of
> >> view) the libraries are now having the name upstream would expect
> >> aka: libboost_bla.so.1.55.0
> >> For compatibility I added a symbol link to libboost_bla.so.5 so
> >> everything does not have to be rebuilt
> >> The ports is now staged aware
> >> It uses a modern compiler (sorry I haven't been able to make it
> >> use gcc from base at all, so it uses clang33 or gcc47+ from ports
> >> on FreeBSD 9.1 and FreeBSD 8.*, clang33 from base in FreeBSD 9.2
> >> (if available) and cc on 10+) It removes the wrongly added support
> >> for long long double.
> >>
> >> Exp-run show everything is ok from package building side of view.
> >>
> >> What about users?
> >>
> >> Can you test?
> >>
> >> to test
> >> cd /usr/ports/devel
> >> patch -p0 < /path/to/boost-1.55.0.diff
> >>
> >> regards,
> >> Bapt
> >>
> > 
> > Hi bapt,
> > 
> > There is a PR to fix building boost-libs on non-x86 platforms that
> > is not in the 1.55.x release branch, that I have tested and is
> > ready to commit:
> > 
> > http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/182549
> > 
> > I have additionally re-opened the upstream issue and pinged
> > upstream on IRC about the merge back to the release branch.
> > 
> > Would you prefer a) I commit now and you merge a one line change to
> > a single source file into your CFT patch, or b) something else ?
> > 
> > jhibbits (submitter) CC'd
> > 
> > Thanks!
> > 
> > --
> > koobs
> > 
> 
> Patch attached this time :)

Thanks, koobs, for fixing ports/182549.

Bapt, thanks for taking care of Boost 1.55, this fixes ports/181126
when it gets in.

However, I have some problems with testing your patch:

1) it doesn't seem to acknowledge USE_GCC at the command line (are
ports supposed to?), and with the base assembler requires passing
-maltivec to get altivec instructions (maybe using -integrated-as for
clang would fix this?  Regardless, gcc needs to be used for powerpc
target)
2) I see the following messages when configuring, are they known?:
    - iconv (libc)             : no
    - iconv (separate)         : no
    - icu                      : no
    - icu (lib64)              : no
- Boost.Locale needs either iconv or ICU library to be built.

  I do have ICU set in the options.  Is Boost.Locale ignored for
  boost-libs?

- Justin