Re: Any one compiled OpenAFS1.4.2fc3 on freeebsd6.1
Marcus Watts <[email protected]> Tue, 03 Oct 2006 15:28:33 -0400
| Newsgroups | gmane.comp.file-systems.openafs.freebsd |
|---|---|
| Message-ID | <[email protected]> |
Datastream <[email protected]> writes: ... > openafs-1.4.2fc3/src -O2 -pipe -c asetkey.c > asetkey.c: In function `main': > asetkey.c:85: error: structure has no member named `length' > asetkey.c:87: error: structure has no member named `length' > asetkey.c:91: error: structure has no member named `contents' > *** Error code 1 ... > I changed configure to fit freebsd. and Jim Rees <[email protected]> replied: > The FreeBSD client is not reliable. I have fixed the Readme to say this. > > I doubt you will get asetkey to work on FreeBSD. It's not something you > need every day, so I recommend you just run it somewhere else. > > You might find aklog useful if you are using MIT. At one time it built on > FreeBSD. I don't know if it still does. > > There shouldn't be anything else you need from --with-krb5. If you run asetkey from a different machine, make sure it's the same byte sex. The keyfile has machine endian integers inside of it, so won't work without byte swapping if you move it. Somewhere around I have a perl script that will swap the bytes, but I don't seem to have it handy. The problems you ran into with asetkey are most likely due to providing the wrong flavor of kerberos; presumably heimdal instead of mit. If you feel very experimental, you might try this patch: /afs/umich.edu/group/itd/build/mdw/tmp/afs-setkey.patch You'll need to find some way to set USING_HEIMDAL (you could just add "#define USING_HEIMDAL 1" at the start of asetkey.c) [ this comes from diffs for rxk5 against 1.5.8 which of course doesn't directly need asetkey, but does need to be more portable to different flavors of kerberos 5. Probably it will "just work" for you - but who knows? ] The parts you will most likely run into problems with on freebsd are (1) the fileserver, and (2) the cache manager. The fileserver depends a bit on native filesystem semantics. With namei, those dependendencies are much less -- in theory, things should "just work". Since the cache manager knows a lot about the kernel, this is much more problemmatical. A freebsd afs db server (without file service) is very likely to just work. -Marcus