Bugzilla Bug 98827

Darwin/OS X has no permissions for symlinks, therefore when portage attempts to
lchown a symlink 
with a non-existant target (or simply the target hasn't been moved from ${D}
yet) it raises Exception 
and stops the merge.

  lchown(dest,sstat[stat.ST_UID],sstat[stat.ST_GID])


Is the statement that fails.

No, I'm not attaching a patch yet, want to investigate a little more first, but
wanted to start the bug in 
case a portage guru has a good, clean, quick fix =)


Reproducible: Always
Steps to Reproduce:
1.
2.
3.




Portage 2.0.51.22-r1 (default-darwin/macos/progressive, gcc-3.3,
libsystem-7.1-r0, 8.2.0 Power 
Macintosh)
===============================================================
==
System uname: 8.2.0 Power Macintosh powerpc
Unknown Host Operating System
dev-lang/python:     2.4.1
sys-apps/sandbox:    [Not Present]
sys-devel/autoconf:  2.13, 2.59-r6
sys-devel/automake:  1.4_p6, 1.5, 1.7.9-r1, 1.8.5-r3, 1.9.5
sys-devel/binutils:  [Not Present]
sys-devel/libtool:   [Not Present]
virtual/os-headers:  7.1
ACCEPT_KEYWORDS="ppc-macos ~ppc-macos"
AUTOCLEAN="yes"
CBUILD="powerpc-apple-darwin8"
CFLAGS="-mcpu=G4 -O3 -pipe"
CHOST="powerpc-apple-darwin8"
CONFIG_PROTECT="/etc /usr/kde/2/share/config /usr/kde/3/share/config
/usr/share/config /var/
qmail/control"
CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d"
CXXFLAGS="-O2 -pipe"
DISTDIR="/usr/portage/distfiles"
FEATURES="autoconfig buildpkg distlocks fakeroot sfperms strict userpriv"
GENTOO_MIRRORS="http://gentoo.osuosl.org/ http://adelie.polymtl.ca/
http://distro.ibiblio.org/pub/
linux/distributions/gentoo/ http://cudlug.cudenver.edu/gentoo/"
LC_ALL=""
MAKEOPTS="-j2"
PKGDIR="/usr/portage/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY=""
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
USE="ppc bitmap-fonts dvd emboss gif jpeg ldap libwww nls pam png ppc-macos
python readline 
truetype truetype-fonts type1-fonts unicode userland_Darwin kernel_Darwin
elibc_Darwin"
Unset:  ASFLAGS, CTARGET, LANG, LDFLAGS, LINGUAS

Portage currently aliases lchown to chown on Darwin. It would make more sense 
for lchown be aliased to a noop. Does that work for you? 

I'd be careful about aliasing to a noop. lchown isn't only used on symlinks.

Yeah, I think Marius is correct. Ideally it would only noop if the symlink
target didn't exist, that way the 
symlinlk will just inherit perms from its parent.

Created an attachment (id=63710) [details]
Replaces lchown with os.chown when target is not a symlink

There's two case where lchown is used and the target is never a symlink;
changed both of those to use os.chown. Then there's one case where it can be
any type of file; converted that one to an if block based on stat information
that has already been read.

Created an attachment (id=63711) [details]
Turns lchown into a no-op on Darwin

The problem with lchown being aliased to chown is that file merging becomes
order dependent (and portage currently relies on it being not so). That is, the
target will get the wrong permissions if the symlink has different ownership
for some reason.

This looks good to me. thanks jstubbs.

ok, I spoke to soon. This still doesn't solve the problem of merging a symlink
with a non-existant 
target. On line ~2800 of portage.py in movefile() there is another lchown I
didn't see changed in your 
patch. If you change:

  lchown(dest,sstat[stat.ST_UID],sstat[stat.ST_GID])

to:

  os.chown(dest,sstat[stat.ST_UID],sstat[stat.ST_GID])

this will still raise:

  except Exception, e:
                        print "!!! failed to properly create symlink:"

Fixed in 2.0.53