This patch is needed for the /usr-move feature
https://fedoraproject.org/wiki/Features/UsrMove
This package requires now 'filesystem' >= 3, which is only installable
on a system which has /bin, /sbin, /lib, /lib64 as symlinks to /usr and
not regular directories. The 'filesystem' package acts as a guard, to
prevent *this* package to be installed on old unconverted systems.
New installations will have the 'filesystem' >=3 layout right away, old
installations need to be converted with anaconda or dracut first; only
after that, the 'filesystem' package, and also *this* package can be
installed.
Packages *should* not install files in /bin, /sbin, /lib, /lib64, but
only in the corresponding directories in /usr. Packages *must* not
install conflicting files with the same names in the corresponding
directories in / and /usr. Especially compatibility symlinks must not be
installed.
Feel free to modify any of the changes to the spec file, but keep the
above in mind.
When selabel_lookup found an invalid context with validation enabled, it
always stated it was 'file_contexts' whether media, x, db or file.
The fix is to store the spec file name in the selabel_lookup_rec on
selabel_open and use this as output for logs. Also a minor fix if key is
NULL to stop seg faults.
Fix setenforce manage page.
* selinuxswig_python.i: don't make syscall if it won't change anything
* Remove assert in security_get_boolean_names(3)
* Mapped compute functions now obey deny_unknown flag
* get_default_type now sets EINVAL if no entry.
* return EINVAL if invalid role selected
* Updated selabel_file(5) man page
* Updated selabel_db(5) man page
* Updated selabel_media(5) man page
* Updated selabel_x(5) man page
* Add man/man5 man pages
* Add man/man5 man pages
* Add man/man5 man pages
* use -W and -Werror in utils
* load_policy: handle selinux=0 and /sys/fs/selinux not exist
* regenerate .pc on VERSION change
* label: cosmetic cleanups
* simple interface for access checks
* Don't reinitialize avc_init if it has been called previously
* seusers: fix to handle large sets of groups
* audit2why: close fd on enomem
* rename and export symlink_realpath
* label_file: style changes to make Eric happy.
* utils: matchpathcon: remove duplicate declaration
* src: matchpathcon: use myprintf not fprintf
* src: matchpathcon: make sure resolved path starts
* put libselinux.so.1 in /lib not /usr/lib
* tree: default make target to all not
* utils: matchpathcon: remove duplicate declaration
* src: matchpathcon: use myprintf not fprintf
* src: matchpathcon: make sure resolved path starts
* put libselinux.so.1 in /lib not /usr/lib
* tree: default make target to all not
2.1.4 2011-0817
* mapping fix for invalid class/perms after selinux_set_mapping
* audit2why: work around python bug not defining
* resolv symlinks and dot directories before matching
* Release, minor version bump
* Give correct names to mount points in load_policy by Dan Walsh.
* Make sure selinux state is reported correctly if selinux is disabled or
fails to load by Dan Walsh.
* Fix crash if selinux_key_create was never called by Dan Walsh.
* Add new file_context.subs_dist for distro specific filecon substitutions
by Dan Walsh.
* Update man pages for selinux_color_* functions by Richard Haines.
* Give correct names to mount points in load_policy by Dan Walsh.
* Make sure selinux state is reported correctly if selinux is disabled or
fails to load by Dan Walsh.
* Fix crash if selinux_key_create was never called by Dan Walsh.
* Add new file_context.subs_dist for distro specific filecon substitutions
by Dan Walsh.
* Update man pages for selinux_color_* functions by Richard Haines.
Errors were being seen in libpthread/libdl that were related
to corrupt thread specific keys. Global destructors that are called on dl
unload. During destruction delete a thread specific key without checking
if it has been initialized. Since the constructor is not called each time
(i.e. key is not initialized with pthread_key_create each time), and the
default is 0, there is a possibility that key 0 for an active thread gets
deleted. This is exactly what is happening in case of OpenJDK.
