On 02/02/2012 07:41, Joshua Isom wrote: > I know that build cluster lists some ports that have problems with > clang, but it doesn't say if they're tested or not. I set up a clang > jail to test out things before switching to clang for general use. When > I try running mencoder to encode a file to x264, it seg faults. Changing > options doesn't change anything. A gdb backtrace points to x264 being > the problem. Everything compiles and installs, but the build's > useless. Is clang ready for ports, or is it only safe for kernel/world?
It is certainly true that more ports will compile with the base gcc-4.2 compiler than will compile with clang. It's also true that of the ports that do compile, there will be more run-time failures with clang than with gcc. At the last test it was something like 18,000 out of 23,000 ports that built successfully with clang -- remember though that figure (a) includes a lot of ports that don't use a compiler at all (like shell scripts or pure-perl code) and (b) some of the failures are because a dependency failed to compile, which blocks testing on anything further down the tree, or for reasons completely unrelated to the compiler, like being unable to fetch distfiles. See: http://wiki.freebsd.org/PortsAndClang I personally have a FreeBSD 9-STABLE VM where everything (system+ports) is compiled with clang, but it's just for testing, has nothing particularly valuable on it and no one would care if it laid down and died. I am planning on upgrading my primary machine to 9.0 sometime soon, and while still undecided whether to build the system with clang or not, I certainly won't be enabling it for ports just yet. Cheers, Matthew -- Dr Matthew J Seaman MA, D.Phil. 7 Priory Courtyard Flat 3 PGP: http://www.infracaninophile.co.uk/pgpkey Ramsgate JID: matt...@infracaninophile.co.uk Kent, CT11 9PW
signature.asc
Description: OpenPGP digital signature