Re: icu update to 77.1

[Greg Troxel <gdt%lexort.com@localhost>]

Thomas Klausner <wiz%NetBSD.org@localhost> writes:

> icu already required c++17.
> qt5-qtwebkit has a workaround in its Makefile:
>
> # XXX Workaround for icu-76 whose header files require C++17 unless you
> # define U_HIDE_DRAFT_API.  pkg/59028: qt5-qtwebkit doesn't build
> CXXFLAGS+=              -DU_HIDE_DRAFT_API
>
> that doesn't seem to be enough any longer with icu 77.
>
> (I also tried turning off its new header-only APIs with
> CXXFLAGS+=              -DU_SHOW_CPLUSPLUS_API=0
> but that didn't fix qt5-qtwebkit.)

Thanks, now I understand better.

>> What if in icu bl3, instead of USE_CXX_FEATURES+=c++17, it is c++11?
>> Or does icu document and truly require c++17 for header use (martin's
>> 2nd question)?
>
> Since icu documents it requires C++17 I'm not sure what this test is
> useful for. icu/buildlink3.mk does not require any c++ standard right
> now, just gcc 4.9+.

That seems like a bug, if icu documents C++17, and therefore taht adding
USE_CXX_FEATURES+=c++17 is fixing that bug.

I meant that if it turns out that despite icu saying 'c++17' is required
that if it is true that c++17 is required for the library but c++11 for
headers, that would be interesting.

But you point out above that building the headers in c++11 mode is known
to actually fail.

> https://unicode-org.github.io/icu/download/77.html

Thanks; doesn't seem to address headers vs libs, but you've explained
that it's headers too.

Given all this I think we're left with:

  A) someone who cares can fix qtwebkit

  B) we could version icu

  C) qtwebkit can stop working and perhaps be removed as unmaintained

and I feel like we are heading to C, which does not seem so terrible.  b