On 12/11/20 1:10 PM, Jean-Marc Lasgouttes wrote: > Le 11/12/2020 à 15:29, Yuriy Skalko a écrit : >>> The only issue left is concerning tracker_/trackable_ members of >>> Trackable class used in call to track_foreign. As I understand it is >>> some hack to have lifetime of slots be the same as of containing >>> class (PreviewLoader::Impl/Converter::Impl). Is it right? >> >> I've resolved this issue. Looking at the comment for Trackable class >> in signals.h, I've implemented more thread-safe solution that tracks >> parents directly with a shared_ptr as recommended by signals2. Now >> deletion of the parent by a concurrent thread is prevented. >> >> Trackable class is still used in Server.h/.cpp so I haven't deleted >> it. Probably it should be moved there from signals.h. > > To be frank I am not competent to review this. > > The code in the nod library seems well written and commented. It has > not be touched for 2 years, but it is maybe because it does not need > to. This code should be maintainable at worst. > > The fact that Yuriy has researched all the possible alternatives is > good too. > > If we get rid of Boost.signal2, then the size of boost will not be a > concern anymore. > > Any thoughts?
I'm not really competent, either. But if there were going to be problems, I think we would see them fairly quickly. Maybe it would be a good time, after this were committed, to do another alpha-ish release. Riki -- lyx-devel mailing list lyx-devel@lists.lyx.org http://lists.lyx.org/mailman/listinfo/lyx-devel