On 2010/11/19 22:50, Christian Weisgerber wrote: > Do we want to use .xz distfiles for ports that otherwise do not > depend on archivers/xz?
I would certainly use .xz distfiles if I'm packaging/hosting my own snapshot tarball from an upstream repository and it's a non-trivial size. (e.g. when we were building llvm from a snapshot). For texmf, admittedly an extreme case, it saves 87MB compared to bzip2 - for most things I've checked, xz saves as much over bzip2 as bzip2 saves over gzip. If it's something small I probably wouldn't use either bzip2 or xz, I'd stick with gzip as it's faster to decompress and avoids additional dependencies. And for something huge I wouldn't think twice if there's an .xz distfile available. For mid-sized things given a choice of distfiles, I'd probably use bzip2 at the moment because writing a separate EXTRACT_CASES is messy, but that's about the only reason, I don't see any major advantage of bzip2 over xz. > If we want to use more .xz distfiles, should we drop the gettext > dependency from the xz port (--disable-nls) to minimize the dependency > tree? I think that makes sense.. > A complicating factor is that xz requires at least gcc3 to build. > (The code is full of declarations after statements. They didn't > slip in accidentally, the author purposely uses them all the time.) Can gcc3 be built on vax at the moment?
