Pkgcore

Latest version: v0.12.28

Safety actively analyzes 682361 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 16 of 27

0.6.1

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

- fix for "or_node.blocks" AttributeError, and related resolution
miscalculations.

- fix exit code return for ebuild helpers throwing warnings for <EAPI4

- fix typo in FEATURES=buildsyspkg, and FEATURES=buildpkg

- check to ensure pkgdir exists; if possible, create it, else turn off binpkg
features.


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

0.6

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

- Due to crazy work hours and moves, this release is fairly large, and frankly
repeatedly delayed. Future ones will be far more fine grained moving forward.

- Fix python2.7 incompatiblity in pkgcore.ebuild.misc

- It's suggested that folks use bash 4.1, primarily for regen
speed reasons- it is not required however.

- bash spawning now enforce --norc and --noprofile in full.

- RESTRICT is now properly use evaluated.

- pkgcore.restrictions.values.ContainmentMatch is deprecated in favor of
ContainmentMatch2. Update your code- by pkgcore 0.7, ContainmentMatch
will become ContainmentMatch2, and a shim will be left in place.

- introduction of EAPI objects (pkgcore.ebuild.eapi) for controlling/defining
new eapi's, capabilities, etc.

- pmaint regen is now cumulatively ~23x faster then the previous release.
This is via restoration of original metadata regeneration speeds, and
via enabling an eclass preloading optimization. No impact on metadata-
just far faster regeneration.

- Roughly a 15x speedup in general metadata regeneration; basically rework
a fix that was added to to 0.5.11 (dealing with portage induced
breakage in env loadup from their declare usage).

- filter-env regex backend now uses python's re always; previously
if the extension was active it would use posix regex.

This resolves occasional odd failures when running native filter-env.

- fix a truncation error in suffix version parsing resulting in
_p2010081516093 being less than _p2009072801401 .

- pkgcore.ebuild.restricts now contains some generally useful
building block restrictions for any api consumers

- full rewrite of EAPI helpers adding better error info, saner code,
double checked against PMS and portage/paludis to ensure no oddities.

- fix to buildpkg/pristine-binpkg saving. If you're looking for
something to contribute to pkgcore wise, tests for this would be
appreciated.

- write support for DEFINED_PHASES.

- bashrc hooks now run from ${S} or ${WORKDIR}, depending on
PMS rules for that phase.

- match the other PM's for econf; update ${WORKDIR} instances of
config.{sub,guess} from /usr/share/gnuconfig.

- added protection against bad environment dumps from other PMs for T
during env restoration.

- removed RESTRICT=autoconfig support.

- fix compatibility regression introduced in file-5.05 involving MAGIC_NONE.

- handle keyboard interrupts better during compilation; no longer display
die tracebacks if the user intentionally stopped the compilation.

- duplicate a portage workaround for emacs ebuild; specifically don't
regenerate infodir if the ebuild placed a .keepinfodir in the directory.
gentoo bug 257260.

- add workaround to disable unzip during unpack going interactive during
a failure; gentoo bug 336285.

- fixed traceback during displaying a summary for 'pinspect eapi_usage'

- add EAPI limitation to all portageq invocations, and support USE dep
usage with has_version and friends.

- handle portage's new interpretation of the sync retries variable for portage
configuration.

- pinspect distfiles_usage was added; this is primarily useful for getting
a repository level view of what the distfiles requirements are, what takes
what percentile of unique space, etc.

- FEATURES=allow-missing-manifests ; does exactly as it sounds, not advised to
use unless you know what you're doing.

- ospkg's fork of pkgcore has been folded in; FEATURES=save-deb is the primary
addition.

- extended atom syntax now allows '*' to be used w/in a string- for example
dev-\*kde, \*dev-\*k\*de\*, etc. This syntax is usable in user configs, and
from the commandline.

- new FEATURES=fixlafiles is on by default; basically folds
dev-util/lafilefixer functionality directly into the merger.
Note this version drops comments- it's about a 66% reduction in .la system
filespace requirements via doing so.

- triggers base class now carries a ConfigHint to provide a typename. If
a specific trigger cannot be specified by configuration directly, set
pkgcore_config_type = None to disable the hint removing it from being
directly configurable.

For users: this means basically all triggers are now directly usable in
configuration.

- object inspection for configuration can now handle object.__init__ for
config 'class' targets; no need to define an intermediate function.

- ConfigHints can now specify authorative=True to disable all introspection.
Mainly usedful for cpy objects, although useful if you want to limit what
the introspection exposes.

- api's for installing pkgs has changed; now to install a pkg to a domain,
you invoke domain.(install|uininstall|remove)_pkg. To just modify a repo,
access its operations for the appropriate operation.

- pkgcore.interfaces was moved to pkgcore.operations

- pkgcore.package.base derived objects no longer default to _get_attr dict
lookup- if you want it, set __getattr__ = dynamic_attr_dict.

- USE is now locked an intersection of the pkgs IUSE, with forced flags
(things like arch, userland, prefix, etc) added on. Mild speed up from
dealing with a reduced set, more importantly portage switched to controlled
USE here, so we can force it finally.

- USE collapsing now should match portage behaviour. Essentially now,
pkg IUSE + profile overrides + make.conf overrides + user config package.use
overrides. Previous behaviour didn't get edge cases correct.

- USE_EXPAND default iuse is now fully overridden if the target USE_EXPAND
groupping is defined in configuration. Mostly relevant for qemu-kvm.

- data_source.get_(text|bytes)_fileobj invocations now require writable=True
if you wish to mutate the data source. Via making the intention explicit,
consumers will get just what they need- a 3x speed up for
pquery --attr environment is from that internal optimization alone.

- pkgcore.fs.fsFile.data_source is deprecated; will be removed in the next
major version, use .data instead.

- pkgcore.interfaces.data_source moved to snakeoil.data_source.

- pkgcore.chksum moved to snakeoil.chksum. A compatibility shim was left in
for pkgcore-checks, which will be removed in 0.7 of pkgcore.

- pkgcore ticket 172; rely on snakeoil.osutils.access to paper over differing
os.access behaviours for certain broken userlands (SunOS primarily).


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

0.5.11.8

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

- ticket 221; add --color=(n|y) support

- pmaint perl_rebuild was added; right now it just identifies what needs
rebuilding on perl upgrades, but down the line it'll do the rebuilds as
needed.

- pkgcore now ignores ebuild postrm exit status- it logs failures, but there
isn't really anything that can be done at that stage (everything is already
unmerged after all).

- fixed pkgcore.fs.livefs.iter_scan to support a path pointing to a
nondirectory.

- force all sourcing to stderr; this protects against idiocy like the
python eclass trying to write to stdout in color during sourcing.

- commandline.OptionParser now does a shallow copy of all items in
standard_options_list; this protects against class/instance level cycles
inherent in optparse.OptionParser's design.


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

0.5.11.7

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

- use_enable/use_with; make use_enable/use_with 3rd arg form match pms in eapi4,
match long standing portage behaviour for eapi's 0 through 3.

- when combining repository and slot restrictions in an atom, repository is now
always prefixed with ::, not intermixed. sys-apps/portage:0::gentoo for
example specifies slotting 0, repository gentoo.

- fixed a bug in installed pkgs virtual cache staleness detection- this
accounted for a surprisingly hefty ~25% for simple pquery invocations.

- fix typo in env protection code- load the scrubbed env, not the raw source.


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

0.5.11.6

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

- add a bit of a hack to tty detection tests; PlainTextFormatter is valid for
broken terminfo entries.

- fix support for unpacking of xz tarballs.


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

0.5.11.5

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

- fix yet *another* fucking distutils bit of idiocy. Piece Of Shit.


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

Page 16 of 27

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.