Pkgcore

-----------------------------

- fix py3k regression when trying to hash a PackageRestriction.

- drop CDEPEND tracking (unused, hold over from '04 days), and
newdepend (same era). Neither are used in >=EAPI0 ; if your
ebuild breaks, rebase the ebuild to a valid EAPI.


-----------------------------

-----------------------------

- force all einfo/elog/ewarn style bits to stderr.

- add path attribute to ebuild derived pkg instances; not a guaranteed
part of the api yet, but accessible via pquery --attr path


-----------------------------

-----------------------------

- silence spurious grep QA warnings during metadata sourcing.


-----------------------------

-----------------------------

- fix a major release bug; ebuild-env-utils.sh wasn't packaged in the
released 0.5.11, this version adds the missing file.

- more declare related fixups; this one a regression from 0.5.10- in
sourcing /etc/profile.env, its contents weren't being preserved
fully due to declare.

- add missing eapi3 phase support- basically just reuses eapi2's since
the only changes are environmental.


---------------------------

---------------------------

- took me a full night of debugging, but traced down yet another portage
incompatibility introduced. gentoo bug 303369; if you've been seeing
issues where portage merged ebuild envs aren't reused in pkgcore, this
is now fixed. Env handling in general was heavily rewritten to be as
robust as possible and protect against any further breakages from portage.

- env processing is a bit faster now- uses egrep where possible, falling
back to bash regex where not.

- shell scripts now are tabs based rather than spaces.

- FEATURES=splitdebug works once again.

- It's strongly suggested that you run >snakeoil-0.3.6.1 due to fixes
in extension building- specifically forcing -fno-strict-aliasing back
into cflags since python is invalidly dropping them out.

In addition, if you're running pkgcore on a py3k machine, installation
now is parallelized for 2to3 conversion- should be a fair bit faster.

- rename support for env var CONFIG_ROOT to PORTAGE_CONFIGROOT; seems
that changed in portage at some point. This should fully restore
crossdev support.


---------------------------

---------------------------

- ticket 235; CBUILD/CTARGET values were being stomped w/ CHOST.

- EAPI=3 support; pkgcore already preserved mtimes at the second level,
remaining bits were added for full EAPI3 support.

Pkgcore doesn't currently fully PREFIX offset merges, but that will be
added in the next release or two most likely.

- EBUILD_PHASE was set to setup-binpkg for pkg_setup phase w/ binpkgs-
ebuilds expected setup however, thus EBUILD_PHASE is now set to setup
always for pkg_setup phase.

- fixup env filtering- backslash escaping wasn't needed in the patterns
resulting in failed matches. Mostly protective cleanup.

- tweak cache backend to not stamp cache entries where mtime is no longer
external w/ an mtime of '-1'. Didn't hurt anything but was a pointless op.

- fix the cpy incremental_expansion implementation; not sure how it slipped
in being slower then native python, but the cpy version is now 60% faster
than the native equivalent.
Additionally, this extension is now disabled under py2.4 since it makes
heavy use of PySet apis.

- ticket 234; handle refs properly in dhcpformat/mke2fsformat.

- pkgcore atom objects blocks_temp_ignorable data is now stored in
blocks_strongly; the old attr is aliased, although will be removed.

- pkgcore now supports revisions of arbitrary length (previously was <31 bits).


--------------------------