> 2016-12-21 11:45 GMT+03:00 Jonathan Gray <[email protected]>: > > On Wed, Dec 21, 2016 at 12:43:57AM -0700, Peter Hessler wrote: > >> CVSROOT: /cvs > >> Module name: ports > >> Changes by: [email protected] 2016/12/21 00:43:57 > >> > >> Modified files: > >> databases/postgresql: Makefile > >> > >> Log message: > >> disable spinlocks on arm, so we can have a postgresql package > >> > > > > Wouldn't it be better to depend on a newer gcc/clang to > > get the atomic/sync builtins that base gcc on arm lacks?
If it works, then yes I think so. On 2016/12/21 11:52, Vadim Zhukov wrote: > Until we have working libc++ in base, the generic answer would be > "please, no, if you can". Using egcc/clang forces the port binaries > being linked to libestdc++, which in turn causes havoc in all > dependant ports which are build with base compiler. OpenBSD has been living with that for years, both with base gcc3/ports gcc4, and then base gcc4.2/ports 4.9. It used to be very obvious before you hid the _S_debug_messages symbol warnings ;) If you use any of the usual gui web browsers on OpenBSD you are running code with this kind of conflict.
