Bugzilla Bug 147446

<snip>
pkg_config() {
        # remove stale wrappers from previous broken versions
        [[ -d "${ROOT}/usr/lib/ccache/bin" ]] && rm -rf
"${ROOT}/usr/lib/ccache/bin"
}
</snip>

# emerge -uavl portage

...
existing file /usr/lib/ccache/bin/c++ is not owned by this package
existing file /usr/lib/ccache/bin/g++ is not owned by this package
existing file /usr/lib/ccache/bin/gcc is not owned by this package
existing file /usr/lib/ccache/bin/i686-pc-linux-gnu-c++ is not owned by this
package
existing file /usr/lib/ccache/bin/i686-pc-linux-gnu-g++ is not owned by this
package
existing file /usr/lib/ccache/bin/i686-pc-linux-gnu-gcc is not owned by this
package
* spent 0.00635409355164 seconds checking for file collisions
* This package is blocked because it wants to overwrite
* files belonging to other packages (see messages above).
* If you have no clue what this is all about report it 
* as a bug for this package on http://bugs.gentoo.org

package dev-util/ccache-2.4-r3 NOT merged

# ls -ld /usr/lib/portage/bin
# drwxr-xr-x 2 root root 328 2006-09-13 14:50 /usr/lib/ccache/bin

Now, for something a bit different:

<snip>
pkg_setup() {
        # remove stale wrappers from previous broken versions
        [[ -d "${ROOT}/usr/lib/ccache/bin" ]] && rm -rf
"${ROOT}/usr/lib/ccache/bin"
}
<snip>

>>> dev-util/ccache-2.4-r3 merged.

WTH???

Portage 2.1.1 (default-linux/x86/2006.1/desktop, gcc-4.1.1, glibc-2.4-r3,
2.6.17-gentoo-r7 i686)
=================================================================
System uname: 2.6.17-gentoo-r7 i686 AMD Athlon(tm) XP 1600+
Gentoo Base System version 1.12.4
Last Sync: Wed, 13 Sep 2006 12:00:07 +0000
ccache version 2.4 [enabled]
app-admin/eselect-compiler: [Not Present]
dev-lang/python:     2.4.3-r3
dev-python/pycrypto: 2.0.1-r5
dev-util/ccache:     2.4-r3
dev-util/confcache:  [Not Present]
sys-apps/sandbox:    1.2.18.1
sys-devel/autoconf:  2.13, 2.60
sys-devel/automake:  1.4_p6, 1.5, 1.6.3, 1.7.9-r1, 1.8.5-r3, 1.9.6-r2
sys-devel/binutils:  2.16.93, 2.17, 2.17.50.0.3
sys-devel/gcc-config: 1.3.13-r3
sys-devel/libtool:   1.5.22
virtual/os-headers:  2.6.17
ACCEPT_KEYWORDS="x86 ~x86"
AUTOCLEAN="yes"
CBUILD="i686-pc-linux-gnu"
CFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer -fforce-addr
-ftree-vectorize"
CHOST="i686-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/kde/3.5/env /usr/kde/3.5/share/config
/usr/kde/3.5/shutdown /usr/share/X11/xkb /usr/share/config"
CONFIG_PROTECT_MASK="/etc/env.d /etc/env.d/java/ /etc/gconf
/etc/java-config/vms/ /etc/revdep-rebuild /etc/splash /etc/terminfo"
CXXFLAGS="-march=athlon-xp -O2 -pipe -fomit-frame-pointer -fforce-addr
-ftree-vectorize"
DISTDIR="/usr/portage/distfiles"
EMERGE_DEFAULT_OPTS="--alphabetical"
FEATURES="autoconfig ccache collision-protect distlocks metadata-transfer
parallel-fetch sandbox sfperms splitdebug strict userfetch userpriv
usersandbox"
GENTOO_MIRRORS="ftp://ftp.sh.cvut.cz/MIRRORS/gentoo/gentoo
ftp://ftp.fi.muni.cz/pub/linux/gentoo/"
LANG="en_US.UTF-8"
LDFLAGS="-Wl,-O1 -Wl,--sort-common"
LINGUAS="cs en"
MAKEOPTS="-j2"
PKGDIR="/usr/portage/packages"
PORTAGE_RSYNC_EXTRA_OPTS="--progress"
PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress
--force --whole-file --delete --delete-after --stats --timeout=180
--exclude='/distfiles' --exclude='/local' --exclude='/packages'"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/local/portage"
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
USE="x86 3dnow 3dnowext 7zip X X509 a52 aac acl acpi alsa amr asf audiofile
bash-completion berkdb bluetooth bzip2 cairo caps cddb cdparanoia cdr chroot
cli crypt cscope css cups curl curlwrappers dbx dga dlloader dri dts dv dvd
dvdr dvdread elibc_glibc encode ethereal exif expat fam fbcon ffmpeg fftw
firefox flac flash flatfile foomaticdb gd gdbm gif glibc-omitfp glut gmp gpm
gstreamer iconv icq idn imagemagick imap imlib inifile input_devices_joystick
input_devices_keyboard input_devices_mouse input_devices_vmmouse ipv6 irda jack
javascript jbig joystick jpeg jpeg2k kdeenablefinal kdehiddenvisibility
kernel_linux lcms libcaca libg++ libsamplerate linguas_cs linguas_en lirc
lirc_devices_cph06x lm_sensors logrotate mad maildir matroska mikmod mime mmap
mmx mng mp3 mpeg musepack musicbrainz ncurses nls nodrm nptl nptlonly nsplugin
nvidia offensive ogg openal opengl pam pcre pdf perl png ppds python qt3
quicktime readline real reflection samba sdl session sftplogging skey sndfile
speex spell spl sse ssl svg symlink tcpd theora threads tiff truetype udev
unicode urandom usb userland_GNU v4l v4l2 vcd video_cards_fbdev video_cards_nv
video_cards_nvidia video_cards_v4l video_cards_vesa video_cards_vmware vorbis
win32codecs wmf x264 xine xinerama xinetd xml xml2 xmlrpc xorg xosd xpm xv xvid
xvmc zlib"
Unset:  CTARGET, INSTALL_MASK, LC_ALL

Sigh, the above non-working example should obviously be:

<snip>
pkg_preinst() {
        # remove stale wrappers from previous broken versions
        [[ -d "${ROOT}/usr/lib/ccache/bin" ]] && rm -rf
"${ROOT}/usr/lib/ccache/bin"
}
</snip>

Well, anyway, to reproduce this:

1. emerge =ccache-2.4-r2
2. check that you have orphaned symlinks in /usr/lib/ccache/bin/
3. emerge =ccache-2.4-r4
4. watch the collision
5. move the stuff to pkg_setup() -> directory nuked, no collision

Only thing I can think of is userpriv, but looking at the code preinst isn't
userpriv'd ;)

The problem is that pkg_preinst runs *after* the collision-protect check.  That
makes sense, because the preinst phase should not be executed if the merge has
been aborted due to collisions.  An alternative, for the ccache ebuild, would
be to create the symlinks in pkg_postinst.

(In reply to comment #4)
> That makes sense, because the preinst phase should not be executed if the >merge has been aborted due to collisions.

Uhm, I don't understand, I'm afraid. There would be no merge collisions if
portage did check for them when relevant, not before any merge is actually
done.

> An alternative, for the ccache ebuild, would be to create the symlinks in >pkg_postinst.

No, that's exactly what has caused this bug in the first place. :=) We need to
nuke the orphaned symlinks, and it fails in pkg_preinst. The only workaround is
to move it to pkg_setup.

pkg_setup isn't proper because the system will be left in an inconsistent state
if the build fails or if something like --buildpkgonly is enabled.

There's really no proper solution that doesn't require user intervention
(doesn't affect users that don't have collision-protect enabled).

(In reply to comment #6)
> pkg_setup isn't proper because the system will be left in an inconsistent state
> if the build fails or if something like --buildpkgonly is enabled.

We need a way to nuke files on live FS somehow. See Bug 145167 for reason why
this is needed (And that's definitely not the only case in the current tree.)

> There's really no proper solution that doesn't require user intervention
> (doesn't affect users that don't have collision-protect enabled).

collision-protect doesn't matter here, we simply need a way to kill orphaned
files that just break users and that plain shouldn't be there. Even with
collision-protect disabled, users will still be broken if there's no way to do
so.

So, what's your proper solution to this? Die in ebuilds? That plain sucks.

(In reply to comment #7)
> collision-protect doesn't matter here, we simply need a way to kill orphaned
> files that just break users and that plain shouldn't be there. Even with
> collision-protect disabled, users will still be broken if there's no way to do
> so.
> 
> So, what's your proper solution to this? Die in ebuilds? That plain sucks.
> 

pkg_preinst is the most proper solution.  However, it doesn't work in the case
of collisions.

Shrug... I still can't see why checking for collision takes place at the point
when no collisions whatsoever can occur. Just moving this check would fix this
issue.

If we move the check inside portage, then portage itself is actively putting
the system into a potentially inconsistent state, which is something that we
should try to avoid.

(In reply to comment #9)
> Shrug... I still can't see why checking for collision takes place at the point
> when no collisions whatsoever can occur. Just moving this check would fix this
> issue.

NFC what you mean with "when no collisions can occur" (they obviously can).
preinst runs after the collision protect check as it's a potentially
destructive operation and therefore should run as late as possible.