Summary: | Cannot build unionfs-1.0.13 against kernel 2.6.13 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Alexander Skwar <askwar> |
Component: | Current packages | Assignee: | Stefan Schweizer (RETIRED) <genstef> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | charlie, nbensa, radek |
Priority: | High | ||
Version: | 2005.0 | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
Patch against unionfs-1.0.13.ebuild
Complete unionfs-1.0.14.ebuild Patch aganst Makefile, removing the call to ctags |
Description
Alexander Skwar
2005-09-03 05:06:51 UTC
I think, it is indeed so, that 1.0.13 is not compatible with 2.6.13. On the unionfs-cvs list, there's <http://www.fsl.cs.sunysb.edu/pipermail/unionfs-cvs/2005-September/000343.html>. In there, you can see, that: Log Message: ----------- * inode.c,stale_inode.c: Support for 2.6.13's new follow_link prototype. However, many of the regression tests fail for unrelated reasons (or at least I think they are unrelated). This really makes me think, that 1.0.13 isn't yet compatible with 2.6.13. FYI: I asked the unionfs mailing list reg. 2.6.13 support. See http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2005-September/001049.html Do you hve a patch for fixing the issue or is there something in the unionfs cvs? No, not yet. Not even the CVS has a fix - yet. Let me paste, what I got from the unioncfs list: | Right now fcntl locking doesn't work. If you cloes a file that has been | locked, you will get a kernel Oops. As flocking files might be important, I'd suggest to not use the CVS code - at least not now. Looks like a fix will be coming soon. http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2005-September/001093.html "Unionfs 1.0.13 was having issus with 2.6.13 it has been fixed in the latest snapshot. Dave" Looks like 1.0.14 is out and fixes the 2.6.13 issue. I have been using it on my laptop for a few days now under 2.6.13-gentoo-r1 w/o and problems. I removed the delbranch-fix patch from the ebuild and left in the amd64 bit patch to get it to compile on my box. # emerge info Portage 2.0.52-r1 (default-linux/amd64/2005.0, gcc-3.4.4, glibc-2.3.5-r1, 2.6.13-gentoo-r1 x86_64) ================================================================= System uname: 2.6.13-gentoo-r1 x86_64 AMD Athlon(tm) 64 Processor 3400+ Gentoo Base System version 1.12.0_pre8 ccache version 2.4 [disabled] dev-lang/python: 2.3.5, 2.4.1-r1 sys-apps/sandbox: 1.2.13 sys-devel/autoconf: 2.13, 2.59-r7 sys-devel/automake: 1.4_p6, 1.5, 1.6.3, 1.7.9-r1, 1.8.5-r3, 1.9.6 sys-devel/binutils: 2.16.1 sys-devel/libtool: 1.5.20 virtual/os-headers: 2.6.11-r2 ACCEPT_KEYWORDS="amd64 ~amd64" AUTOCLEAN="yes" CBUILD="x86_64-pc-linux-gnu" CFLAGS="-march=athlon64 -O3 -ffast-math -funroll-all-loops -funit-at-a-time -fpeel-loops -ftracer -funswitch-loops" CHOST="x86_64-pc-linux-gnu" CONFIG_PROTECT="/etc /usr/kde/2/share/config /usr/kde/3.4/env /usr/kde/3.4/share/config /usr/kde/3.4/shutdown /usr/kde/3/share/config /usr/lib/X11/xkb /usr/lib64/mozilla/defaults/pref /usr/share/config /var/qmail/control" CONFIG_PROTECT_MASK="/etc/gconf /etc/splash /etc/terminfo /etc/env.d" CXXFLAGS="-march=athlon64 -O3 -ffast-math -funroll-all-loops -funit-at-a-time -fpeel-loops -ftracer -funswitch-loops" DISTDIR="/usr/portage/distfiles" FEATURES="autoconfig distlocks sandbox sfperms strict" GENTOO_MIRRORS="http://gentoo.chem.wisc.edu/gentoo/ http://gentoo.mirrors.pair.com/" MAKEOPTS="-j2" PKGDIR="/usr/portage/packages" PORTAGE_TMPDIR="/var/tmp" PORTDIR="/usr/portage" PORTDIR_OVERLAY="/usr/local/overlays/gentopia /usr/local/overlays/bmg-main /usr/local/overlays/mine" SYNC="rsync://rsync.gentoo.org/gentoo-portage" USE="amd64 X acpi alsa apache2 audiofile avi bash-completion berkdb bitmap-fonts bluetooth bonobo cddb cdparanoia cdr cgi cli crypt ctype cups curl dba dbus dpms dts dvd dvdr eds emboss encode epiphany esd ethereal evo evo2 exif flac flash foomaticdb fortran freetype gd gdbm gif gimpprint glitz gnokii gnome gphoto2 gpm gstreamer gtk gtk2 gtkhtml guile hal hbci howl icq icu imagemagick imap imlib ipv6 irmc jabber java jpeg junit kde ldap libwww lirc lzw lzw-tiff mad matroska mikmod motif mozilla mp3 mpeg mpeg4 msn mysql nautilus ncurses network nfsv4 nls nptl nsplugin offensive ofx ogg oggvorbis opengl oscar oss pam pcmcia pcntl pcre pdf pdflib pdo-external perl php png pnp ppds python qt quicktime quotes readline real ruby samba sdl session silc simplexml sndfile soap speex spell sqlite ssl subversion svg tcpd theora tiff transcode truetype truetype-fonts type1-fonts usb userlocales vcd visualization vorbis wifi xanim xine xml xml2 xmms xpm xsl xv yahoo zeroconf zlib video_cards_radeon userland_GNU kernel_linux elibc_glibc" Unset: ASFLAGS, CTARGET, LANG, LC_ALL, LDFLAGS, LINGUAS (In reply to comment #6) > Looks like 1.0.14 is out and fixes the 2.6.13 issue. That's right. See the annoucement on their mailinglist at http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2005-September/001096.html Quoting it in its entirety: Unionfs 1.0.14 has just been released. The next version of unionfs that will be released will be 1.1.0. This release will signify a split in the 2.4 and 2.6 source tree. We will continue to deal with security issues under Linux 2.4, but feature enhancements and future development will be focused on 2.6 only. Others can continue to fix bugs and problems on the 1.0 branch for 2.4, and we will be glad to apply any patches they put forth, but want to focus our resources on stabilizing the 1.1 branch (for 2.6). Dave Quigley * Unionfs 1.0.14 Features: - Dropped unmaintained setattr,diropaque,delete=first and mount flags - Updated unionfs.4 to reflect default mount modes - You can use the root of the Union to remove branches with unionctl - Use official Debian packaging files. - Linux 2.6.13 support. Bug fixes: - Fixed several dentry refcount bug introduced by new deletion framework. - Fixed uninitialized fd_set in the query ioctl. - Branch reference counting now works across insertion and removal of branches. This prevents the branch counts from getting "confused". - Handle "/" as a branch in unionctl. - Removed static buffer for debug prints in favor of vprintk. - NFS silly renames avoided during several cases. - Attempting a write lock causes a copyup, so that the underying flock will work. - Cleaned up mount option parsing. - Improved link counting for directories. Could somebody please update unionfs in portage to the current version of 1.0.14? You'll need a new ebuild and a new patch. Both will be attached soon. The pkg now requires dev-util/ctags as a DEPEND. I guess in 1.0.15 this will go away again. See http://www.fsl.cs.sunysb.edu/pipermail/unionfs/2005-September/001120.html The patch is required, as the Makefile calls "ctags", which is not present in Gentoo. The patch replaces the call to "ctags" with one to "exuberant-ctags". However, according to the above mentioned mailinglist message, ctags might not be required at all. The message says: | Chip, Tags shouldn't be built automatically, unless someone runs "make tags" and also: | ctags $@ 2>/dev/null || etags $@ 2>/dev/null || :; What's the fealing? Make unionfs (for now) DEPEND on ctags or add that latter patch, which will do away with the requirement of ctags completely? I just read http://www.fsl.cs.sunysb.edu/pipermail/unionfs-cvs/2005-September/000366.html In this CVS log, the call to ctags is removed. I'll upload an ebuild and a patch which implement this. Created attachment 69354 [details, diff]
Patch against unionfs-1.0.13.ebuild
This is a patch against unionfs-1.0.13.ebuild resulting in a
unionfs-1.0.14.ebuild
Created attachment 69356 [details]
Complete unionfs-1.0.14.ebuild
The complete unionfs-1.0.14.ebuild
Created attachment 69357 [details] Patch aganst Makefile, removing the call to ctags This implements the change as shown in http://www.fsl.cs.sunysb.edu/pipermail/unionfs-cvs/2005-September/000366.html Please bump the unionfs version in portage as it works perfectly... Confirming bug and that solution works. I'm with gentoo-sources-2.6.13-r3 /C Thanks, I fixed it. Sorry for the long delay :( Anyone who really uses it wants to take over unionfs? I'm running linux-2.6.13-gentoo-r3 and unionfs-1.0.14 still won't emerge: /var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14/inode.c:1169: warning: initialization from incompatible pointer type /var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14/inode.c:1171: warning: initialization from incompatible pointer type /var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14/inode.c:1184: warning: initialization from incompatible pointer type /var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14/inode.c:1186: warning: initialization from incompatible pointer type make[2]: *** [/var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14/inode.o]Error 1 make[1]: *** [_module_/var/tmp/portage/unionfs-1.0.14/work/unionfs-1.0.14]Error 2 Oh, forgot to say it is amd64 system. Please open a separate bug for new issues(amd64) and can you please continue discussion there? I can suggest you to try renaming the ebuild to 1.1.0 and commenting out the patches. Then emerge --digest unionfs |