Wait, in the interim, the brave user who wants ICU support system-wide
now, and who has ICU already installed in the system, can just switch
the "--without-icu" part to "--with-icu" in
/usr/ports/devel/boost/Makefile , and do "make; make install" right?
That will not break binary compatibility with precompiled packages, the
only thing would be that if some package is specifically incompatible
with Boost's ICU support, it'd go into undefined behavior, right?
On 2015-12-17 02:40, Landry Breuil wrote:
On Wed, Dec 16, 2015 at 01:40:47PM +0100, Antoine Jacoutot wrote:
On Wed, Dec 16, 2015 at 01:32:21PM +0100, Jérémie Courrèges-Anglas
wrote:
> Tinker <ti...@openmailbox.org> writes:
>
> > On 2015-12-16 05:04, j...@wxcvbn.org wrote:
> >> Tinker <ti...@openmailbox.org> writes:
> >>
> >>> What would the decision be based on?
> >>
> >> I think that those points should be enough.
> >> - good reasons to use ICU in boost, not just "I need the ICU parts of
> >> Boost.". What would be the benefit for the ports tree?
> >
> > I need normalize() to do Unicode normalization!
>
> I'm glad to hear that you want to do Unicode normalization using ICU,
> that is not a valid answer to what I said above.
>
> We have one report here:
>
> http://marc.info/?l=openbsd-ports&m=144171504417490&w=2
>
> jirib didn't confirm that ICU was the only thing needed to make his
> aegisub port work, and to my knowledge no existing port requires ICU in
> boost. The only benefit I can see so far is to have a boost package
> similar to other distros.
I wonder if adding ICU support to boost would unbreak manik.
On that topic i've filed https://github.com/mapnik/mapnik/issues/3202
upstream but didnt thought about that boost/icu thing.
Either way, touching boost... here be dragons.
Landry