> On Oct 11, 2015, at 23:10, Christiaan Hofman <cmhof...@gmail.com> wrote: > >> >> On Oct 11, 2015, at 22:58, Adam R. Maxwell <amaxw...@me.com >> <mailto:amaxw...@me.com>> wrote: >> >> >>> On Oct 11, 2015, at 13:46 , Christiaan Hofman <cmhof...@gmail.com >>> <mailto:cmhof...@gmail.com>> wrote: >>> >>>> >>>> On Oct 11, 2015, at 20:52, Adam R. Maxwell <amaxw...@mac.com >>>> <mailto:amaxw...@mac.com>> wrote: >>>> >>>> BibDesk will probably need >>>> >>>> <key>NSAppTransportSecurity</key> >>>> <dict> >>>> <key>NSAllowsArbitraryLoads</key> >>>> <true/> >>>> </dict> >>>> >>>> to compile with the 10.11 SDK. >>> >>> It does not need this for compilation, >> >> Yeah, misspoke; I meant if you compile on the 10.11 SDK, it will need it at >> runtime, since my understanding is that compiling against the 10.11 SDK >> triggers the new App Transport security, which broke TeXShop's Sparkle >> updates (and I assume would break stuff like PubMed support unless it's >> https). Of course, if you're on 10.10, you can't even test that, which is >> unbelievable. >> >>> apart from stupid warnings and the open-ssl error I have no problems >>> compiling with Xcode 7.0.1. But perhaps it will be needed to have BibDesk >>> work fully on 10.11. I haven’t seen any reports to that effect though. It >>> may give problems with the updater, as we cannot retroactively add this >>> before updating. >> >> Speaking of stupid warnings, I sprinkled these all over my code yesterday. >> >> #pragma clang diagnostic push >> #pragma clang diagnostic ignored "-Wnonnull" >> [self tableViewSelectionDidChange:nil]; >> #pragma clang diagnostic pop >> > > I handled those by adding dummy arguments, either a variable with a nil > value, or a dummy object (like an empty notification in this case). > > The one place where I needed the pragma’s was with the -init for the > operation(queue) class clusters in the file view project. > >> Adding WARNING_CFLAGS = -Wno-tautological-pointer-compare to FileView >> helped, too. >> > > Where does that have an effect? > >>> Yea, I did the HAVE_OPENSSL_H=0 for now. At least it compiles, I did not >>> check whether zoom servers work yet. >>> >>> Looking in the code, it seems they just use open-ssl when the server >>> communicates through the ssl: protocol, at least that’s in the part where >>> they create a connection. So the effect may be limited. >> >> Yeah, I don't know what else to do. AFAIK z39.50 isn't encrypted, so it >> should work just fine. It's been years since I looked into that, though, and >> yaz has probably changed a lot; sadly, they never took my objc bindings >> upstream :(. >> >> — adam > > I guess for now see whether there will be any reports. > > Christiaan
I did have some deeper look into this, and I came to the conclusion we really don’t need openssl. It is only used when the server uses a ssl: or https: protocol, which it gets from the host address’s scheme. This is in the cs_create_host function in cosmetic.c. But we expressly don’t have a scheme (look on BDSKServerInfo assumes either a simple address without scheme, and possibly a port and database attached to it, but removes any scheme). All code that uses openssl will be ignored when these schemes are not used, or an SSL context is explicitly set from the outside using cs_set_ssl_ctx (which we never do). So we can safely remove the use of openssl. And of course I first wasted a lot of time figuring out how to build and include openssl. I guess I learned a bit about building and linking libraries in Xcode. Christiaan
------------------------------------------------------------------------------
_______________________________________________ Bibdesk-develop mailing list Bibdesk-develop@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bibdesk-develop
