From: Johannes Thyssen Tishman Subject: Re: [new] devel/scnlib To: Theo Buehler Cc: Rafael Sadowski , ports@openbsd.org Date: Mon, 20 Apr 2026 13:20:17 +0000 2026-04-20T11:04:33+0200 Theo Buehler : > On Mon, Apr 20, 2026 at 08:15:28AM +0000, Johannes Thyssen Tishman wrote: > > 2026-04-17T08:12:01+0200 Rafael Sadowski : > > > On Sat Apr 04, 2026 at 02:40:43PM +0000, Johannes Thyssen Tishman wrote: > > > > Please find attached a port for devel/scnlib. > > > > > > > > COMMENT: > > > > scanf for modern C++ > > > > > > > > DESCR: > > > > scnlib is a modern C++ library for replacing scanf and std::istream. > > > > This library attempts to move us ever so much closer to replacing > > > > iostreams and C stdio altogether. It's faster than iostream, and > > > > type-safe, unlike scanf. Think {fmt} or C++20 std::format, but in the > > > > other direction. > > > > > > > > This port is needed to update graphics/vtk to version 9.6.0. Most tests > > > > are passing (96% tests passed, 1 tests failed out of 25). I've > > > > reported[1] the set of tests that failed upstream. > > > > > > > > [1] https://github.com/eliaskosunen/scnlib/issues/155 > > > > > > > > Feedback? OKs? Thank you. > > > > > > OK (I'm in favour of post 7.9!) but ... > > > > Hi rsadowski. Thanks for reviewing these ports and sorry for the late > > reply. Attached a tarball of the port. > > > > ... > > > > > > +COMPILER = base-clang ports-gcc > > > > > > why COMPILER? > > > > IIRC, to override the default of 'base-clang base-gcc', as I believe > > base-gcc is never used for cmake ports, right? > > I don't know if that's true in general but this particular port fails > to configure with the compiler line removed (see further down): > > -- Compiler frontend not gcc or clang -- build failure tests don't check the compiler output > -- Configuring incomplete, errors occurred! > > In general, essentially everything c++ wants to be compiled with > ports-gcc on base-gcc architectures, unless it's on the path to > lang/gcc/15. > > So please add the compiler line back before import. The port then builds > on sparc64 and only fails one test, which I assume is the same failure > as the one you had. Will do, thanks. Yes it's the same test.