Le dim. 10 mai 2020 à 20:25, Joerg Sonnenberger <jo...@bec.de> a écrit : > I find it funny that you pick an example where correctly working > atomic implementation is essential and where the only reason it is > pulled in is a bug in GCC.
Yet the developers made conscious choice to depend on libatomic by explicitly requiring 128bit atomic operations, and the only platform where this doesn't work is apparently NetBSD. I admit I did not follow all the arguments, but I see no way to claim support for C11/C++14 without the lib. It's responsibility of the developers of the respective software which pulls dependency on libatomic to ensure their software works on hw platforms which don't have the hw support. It's pointless to unload on Kamil for trying to reduce amount of NetBSD-specific hacks. We can't fix Linux choices, and if we ask 3rd party developers for NetBSD-specific hacks they'll simply stop supporting it. I for one think that we need to accept the necessary reality, and indeed put libatomic to base. Jaromir
