On Sep 11 2009 ver-2.8.1 of fuse was released as a stable release. Current ~arch version in portage is ver-2.7.4 which was released on Jul 25 2008. Reproducible: Always
*** Bug 277510 has been marked as a duplicate of this bug. ***
Created attachment 206324 [details] ebuild for fuse 2.8.1 It's just the old ebuild with kernel bits removed, as they're no longer there.
If you wonder about the patch, it was recently removed from the tree (as unneeded).
I've just tried the ebuild with gcc-4.4.1 and glibc-2.10.1 and all compiled fine. Also just recompiled ntfs-3g and that works with fuse-2.8.1 too.
Afaik newer versions of GlusterFS use (but doesn't require) features specific for fuse-2.8+.
Rafał, Can you rebase your ebuild off of the new 2.7.4-r1 (look at the USE=modules code) and I will bump it for you?
Well, the problem that, as I said in comment 2, there's no point - one of the main changes in 2.8 was: "Out-of-tree fuse module removed". There is no need for those kernel bits, as they don't do anything now. Of course, I've got no idea if it still works for FreeBSD, I only corrected the part for Linux. And that update seems broken anyway: opposite of 'use modules && use kernel_linux' is !(use modules) || !(use kernel_linux), so parts of 'if...else...fi' blocks meant for FreeBSD would get executed if USE='-modules'.
Though there one thing that could be added instead: as it now relies on kernel FUSE_FS, perhaps either in pkg_setup or in post-install note mention the need for setting that kernel option.
Created attachment 208645 [details] ebuild with kernel checks I'm not sure if those checks are correct, as I didn't have need before to examine closely linux-info eclass, but they do look that way. Kernel version with "die" is probably unneeded, as that version is very old.
Ahh, I see what you mean with comment #2 now. Ok. Well, the problem with the ebuild (why I *needed* USE=modules) is when you try to compile fuse with a kernel that *has* CONFIG_FUSE_FS=y, but doesn't have /usr/src/linux/.config it will bail out - this type of scenario happens on VPS guests where a kernel is managed for you. So, we need to drop that check or figure something else out.
Oh. You did CONFIG_CHECK="~FUSE_FS" which only warns - I guess I need to dig into the eclass or just try it to figure out what happens. :)
Well, while I haven't tested that, I've made that check non-fatal, as, AFAIK, only user-mode tools are built, so they should depend only on linux-headers (perhaps a dep on those should be added ?). That check should only be a warning, that you won't be able to use those tools on anything.
InCVS now. Thanks for the work, I cleaned it up a little bit more.
