pkgcore 0.12.44__tar.gz → 0.12.46__tar.gz

{pkgcore-0.12.44 → pkgcore-0.12.46}/NEWS.rst

@@ -2,6 +2,349 @@
 Release Notes
 =============
 
+----------------------------
+pkgcore 0.12.46 (2026-10-08)
+----------------------------
+
+Fixes
+~~~~~
+
+- ``repo:path`` entries in the ``parent`` file of a profile outside a repo,
+  such as a ``/etc/portage/make.profile`` directory, now work (Arthur Zamarin,
+  #440)
+
+- ``pmaint regen`` no longer deletes the ``Manifest.gz`` files in the
+  metadata cache of rsync checkouts, and reading the cache no longer fails
+  on them (Arthur Zamarin)
+
+- ``pmaint regen`` no longer fails with ``TimeoutError: ebp for pid ...
+  appears dead`` while updating ``use.local.desc`` (Arthur Zamarin)
+
+- Sourcing ebuilds from a repo whose path has non-ASCII characters no
+  longer fails for every package (Arthur Zamarin)
+
+- pkgcore tools no longer hang when terminated while starting ebuild
+  processors, or when an ebuild processor fails during startup (Arthur
+  Zamarin)
+
+- ``pmaint regen`` of several repos no longer crashes with ``signal only
+  works in main thread`` when an ebuild fails to source (Arthur Zamarin)
+
+- pkgcore tools no longer exit silently with status 141 when an ebuild
+  processor dies, e.g. from the OOM killer (Arthur Zamarin)
+
+- An ebuild processor dying while reporting a ``die`` no longer leaves
+  pkgcore looping forever with growing memory (Arthur Zamarin)
+
+- ``pmaint regen`` no longer leaves an ebuild processor running for each
+  ebuild that fails to source without calling ``die`` (Arthur Zamarin)
+
+- ``pmaint regen`` writes ``use.local.desc`` and ``pkg_desc_index`` as UTF-8
+  instead of failing on non-ASCII text in a non-UTF-8 locale (Arthur
+  Zamarin)
+
+- ``pmaint regen`` with threads no longer writes ``pkg_desc_index``,
+  ``use.local.desc`` and ``timestamp.chk`` world-writable, with snakeoil
+  0.11.7 (Arthur Zamarin)
+
+- A ``thirdpartymirrors`` line naming a mirror without URIs is skipped like
+  in portage, instead of failing to load the repo (Arthur Zamarin)
+
+- An ebuild file whose name isn't a valid version of its package, such as
+  ``pkg-1.0-rc1.ebuild``, is skipped with a warning like in portage,
+  instead of aborting ``pmaint regen`` (Arthur Zamarin)
+
+- ``pmaint regen --use-local-desc`` reports a ``metadata.xml`` that isn't
+  valid UTF-8 or can't be read as an error for that package, instead of
+  not writing ``use.local.desc`` at all (Arthur Zamarin)
+
+- ``pmaint regen`` reports an ebuild that confuses or kills its ebuild
+  processor as failing to source, instead of crashing after the regen
+  (Arthur Zamarin)
+
+- ``pmaint regen`` reports and counts a ``RuntimeError`` while regenerating
+  a package, instead of losing it with its worker thread and exiting 0
+  (Arthur Zamarin)
+
+- An ebuild that sets an invalid ``EAPI`` after its header, or adds an
+  unknown eclass to ``INHERITED``, is reported as bad metadata instead of
+  crashing ``pmaint regen`` (Arthur Zamarin)
+
+- A repo whose ``profiles/categories``, ``profiles/repo_name``,
+  ``profiles/thirdpartymirrors`` or ``metadata/layout.conf`` isn't valid
+  UTF-8 loads with the bad bytes replaced like in portage, instead of
+  failing (Arthur Zamarin)
+
+- A category listed in ``profiles/categories`` that's a file or can't be
+  read is logged and treated as empty, instead of crashing ``pmaint regen``
+  (Arthur Zamarin)
+
+- An eclass changing during ``pmaint regen`` no longer makes the next
+  packages sourced by the same processor fail (Arthur Zamarin)
+
+- ``pconfig configurables`` no longer crashes with a ``TypeError``
+  (Arthur Zamarin)
+
+- With ``--sandbox`` confinement and ``TMPDIR`` set, ebuilds using here-strings
+  or here-documents in global scope no longer fail to source (Arthur Zamarin)
+
+- Sync, strip and ``install-info`` binaries are no longer looked up in the
+  current directory when ``PATH`` is empty or unset (Arthur Zamarin)
+
+----------------------------
+pkgcore 0.12.45 (2026-10-03)
+----------------------------
+
+Fixes
+~~~~~
+
+- resolver: resolve a package's IDEPEND instead of resolving its PDEPEND
+  twice (Arthur Zamarin)
+
+- resolver: when a candidate fails to insert, check the next one against its
+  own dependencies rather than the previous candidate's (Arthur Zamarin)
+
+- binpkg: installing or replacing from a binpkg repo no longer fails with a
+  TypeError (Arthur Zamarin)
+
+- unmerging a package with files matching ``UNINSTALL_IGNORE`` no longer
+  crashes (Arthur Zamarin)
+
+- fetch: a download that succeeds on the last allowed attempt is no longer
+  reported as failed, so ``attempts=1`` works (Arthur Zamarin)
+
+- GPG signed Manifests can be read again (Arthur Zamarin)
+
+- ebd: a sandbox violation shows its access summary again instead of hanging
+  the build (Arthur Zamarin)
+
+- ebd: ``nonfatal`` works again for ``newins`` and the other ``new*`` helpers
+  (Arthur Zamarin)
+
+- ebd: a failed ``nonfatal doins`` (and other install helpers) no longer breaks
+  every later call in the same phase (Arthur Zamarin)
+
+- ebd: ``eapply_user`` no longer reuses the options of the last ``eapply``
+  call, such as ``-p0`` (Arthur Zamarin)
+
+- EAPI 9: ``ver_replacing`` can be called (Arthur Zamarin)
+
+- profiles can be looked up by absolute path, which also fixes
+  ``pkgcheck --profiles /path/to/profile`` (Arthur Zamarin)
+
+- ``ACCEPT_KEYWORDS`` wildcards (``*``, ``~*`` and ``**``) work without any
+  package.accept_keywords entries (Arthur Zamarin)
+
+- USE is now stacked in portage's order: make.conf USE, including ``-*`` and
+  USE_EXPAND variables such as ``VIDEO_CARDS``, overrides the profile's
+  package.use, and ``USE`` from the environment overrides package.use
+  (Arthur Zamarin)
+
+- ebd: environments saved by pkgcore or portage are filtered again before
+  the ``pkg_*`` phases load them, so stale internal variables no longer leak
+  in (Arthur Zamarin)
+
+- an unquoted ``EAPI=8 # comment`` line is accepted again (Arthur Zamarin)
+
+- binpkgs containing device nodes can be read (Arthur Zamarin)
+
+- vdb: installing a package that has no ebuild no longer crashes and leaves a
+  half-written entry behind (Arthur Zamarin)
+
+- ebd: ``eapply`` on a directory applies only the patches directly inside it,
+  as PMS says, instead of descending into subdirectories (Arthur Zamarin)
+
+- with ``XDG_CONFIG_HOME``, ``XDG_CACHE_HOME`` or ``XDG_DATA_HOME`` set, pkgcore
+  now uses a ``pkgcore`` subdirectory of it, as it does for the defaults. A
+  config at ``$XDG_CONFIG_HOME/pkgcore.conf`` has to move to
+  ``$XDG_CONFIG_HOME/pkgcore/pkgcore.conf`` (Arthur Zamarin)
+
+- rsync sync: the timestamp check uses the repo's ssh, proxy and rsync
+  settings, and a failed check falls back to a full sync instead of aborting
+  (Arthur Zamarin)
+
+- a malformed line in ``use.desc``, ``use.local.desc`` or ``desc/*.desc`` is
+  logged and skipped instead of making every USE description unavailable
+  (Arthur Zamarin)
+
+- ``FEATURES=unmerge-backup`` saves a binpkg of the package being removed again
+  (Arthur Zamarin)
+
+- ``pkgcore.bugzilla``: searching an empty list of bug ids or packages
+  returns nothing instead of every bug on the server (Arthur Zamarin)
+
+- ``pkgcore.bugzilla``: ``limit`` and ``offset`` apply to a whole search even
+  when a long one is split into several requests, rather than to each request
+  (Arthur Zamarin)
+
+- ``pkgcore.bugzilla``: a 429 response is retried whatever its body
+  (Arthur Zamarin)
+
+- ``pkgcore.bugzilla``: a connection reset while reading the response is
+  retried and raised as ``BugzillaConnectionError``, instead of escaping as a
+  bare ``OSError`` (Arthur Zamarin)
+
+- ``pkgcore.bugzilla``: redirects are refused with a caller supplied opener
+  too, so the api key isn't sent on to another host (Arthur Zamarin)
+
+- ``pkgcore.bugzilla.ListChange``: combining a ``setting()`` with adds or
+  removes keeps it a set of the resulting list, instead of dropping the set
+  (Arthur Zamarin)
+
+- ``pkgcore.bugzilla.BugUpdate``: an update that clears a field, such as
+  ``whiteboard=""``, is no longer falsy (Arthur Zamarin)
+
+- ``--api-key`` is stripped, and a blank one falls back to the environment and
+  dotfiles like an unset one, instead of being sent as the key (Arthur Zamarin)
+
+- a configured binpkg repo passes install, uninstall and replace through to the
+  underlying repo again (Arthur Zamarin)
+
+- looking a repo up by path no longer matches a sibling whose name starts the
+  same, such as ``gentoo-extra`` for ``gentoo`` (Arthur Zamarin)
+
+- ``bugzilla_cassette`` fixture: a test that fails no longer gets an extra
+  "unused recordings" error on top (Arthur Zamarin)
+
+- ``pinspect profile status`` and ``pinspect query get_profiles`` work again
+  (Arthur Zamarin)
+
+- cvs sync: the first checkout names the module to check out (Arthur Zamarin)
+
+- rsync sync: ``metadata/timestamp.chk`` is read as UTC with its offset, rather
+  than as local time, so the "is the tree new enough" check isn't skewed by the
+  local timezone (Arthur Zamarin)
+
+- libtool archive fixing no longer rewrites every ``.la`` file that has an
+  ``inherited_linker_flags`` entry, and moves thread flags into that entry
+  when present, as portage does (Arthur Zamarin)
+
+- an unknown ``profile-formats`` entry no longer adds ``pms`` next to the known
+  ones, so ``portage-2`` repos with an extra format keep directory support
+  (Arthur Zamarin)
+
+- the bash version an EAPI needs is compared numerically, so bash 5.10 won't
+  disable EAPI 9, and a bash that can't be queried no longer fails the import
+  (Arthur Zamarin)
+
+- http, tar and sqfs sync: a DNS or connection failure is reported as a sync
+  error instead of crashing with ``AttributeError`` (Arthur Zamarin)
+
+- a failed fetch through package operations created without an observer
+  reports the fetch failure instead of crashing with ``AttributeError``
+  (Arthur Zamarin)
+
+- the unsafe world writable SetUID/SetGID warning names the right bit, and a
+  world writable setgid directory is left alone, as portage does
+  (Arthur Zamarin)
+
+- ebd: shutting down a daemon that stopped answering kills it instead of
+  hanging forever, as on Ctrl-C during a phase (Arthur Zamarin)
+
+- ebd: a multi-line error from a ``nonfatal`` helper, such as ``install``'s
+  stderr, no longer leaves stray lines that confuse the next helper call
+  (Arthur Zamarin)
+
+- config: a native int given to an int typed setting, as from Python built
+  config sections, is accepted instead of failing to convert (Arthur Zamarin)
+
+- ``pconfig dump`` and ``pconfig dump-uncollapsed`` show int settings
+  (Arthur Zamarin)
+
+- ``pmaint eclass <dir>`` reads the eclasses from that directory rather than
+  the current one (Arthur Zamarin)
+
+- incremental settings from ``/etc/profile.env``, such as ``CONFIG_PROTECT_MASK``,
+  are kept even when the profile doesn't set them (Arthur Zamarin)
+
+- a bad ``parent`` entry in a profile outside a repo, such as a stale
+  ``/etc/portage/make.profile`` directory, is logged instead of crashing with
+  ``AttributeError`` (Arthur Zamarin)
+
+- ebd: ``doman -i18n=<lang>`` installs into that language's man directory
+  instead of failing (Arthur Zamarin)
+
+- merging a symlink over an existing directory resolves the link's target from
+  the link's own directory, and an absolute target under ``ROOT``, rather than
+  checking the wrong path or the host system (Arthur Zamarin)
+
+- ``pmerge -x`` and ``--pdb-intercept`` given more than once keep every value,
+  not only the last one (Arthur Zamarin)
+
+- ``pclean`` removes files even when its output is redirected, as from cron;
+  only ``--pretend`` lists them, one bare path per line when redirected
+  (Arthur Zamarin)
+
+- ``restrictions``: the DNF solutions of a negated ``OrRestriction`` are only
+  the negated terms, without the original ones after them (Arthur Zamarin)
+
+- ``pkgcore.pkgsets.glsa``: an unaffected glob range such as ``1.5*`` no longer
+  marks the vulnerable versions safe and the fixed ones vulnerable, and glob and
+  ``-r0`` revision ranges respect the GLSA's slot (Arthur Zamarin)
+
+- ``pkgcore.pkgsets.glsa``: a GLSA range with a subslot, such as
+  ``slot="0/esr78"``, matches that slot and subslot instead of nothing, and
+  ``slot="*"`` matches any slot (Arthur Zamarin)
+
+- ``pkgcore.ebuild.atom``: an ``=*`` glob only matches up to a version
+  component boundary, so ``=cat/pkg-1*`` no longer matches ``cat/pkg-10``, as
+  in Portage; GLSA glob ranges follow the same rule (Arthur Zamarin)
+
+- ``pkgcore.ebuild.ebd_ipc``: ``dosym`` checks for a directory target in the
+  image instead of on the host, and ``dohard`` links a file from the image
+  instead of the host's copy (Arthur Zamarin)
+
+- ``pkgcore.ebuild.processor``: environment values containing both a quote and
+  a backslash, such as ``it's C:\new``, reach the ebuild unchanged instead of
+  with the backslash sequences expanded (Arthur Zamarin)
+
+- ``pkgcore.ebuild.ebd_ipc``: ``doins`` given only directories without ``-r``
+  fails instead of installing nothing and succeeding, as in Portage
+  (Arthur Zamarin)
+
+- ``pkgcore.ebuild.conditionals``: merging a package with ``:=`` deps that
+  aren't all installed, such as ``|| ( a:= b:= )``, no longer crashes; the
+  unmatched atoms are recorded as written, as in Portage. Recording the slots
+  also no longer changes the same atoms elsewhere in the process
+  (Arthur Zamarin)
+
+- ``pkgcore.ebuild.portageq``: ``pinspect query mass_best_version`` and its
+  ``pinspect portageq`` form print each best version instead of failing with a
+  TypeError (Arthur Zamarin)
+
+- ``pkgcore.merge.triggers.BinaryDebug``: configured extra strip flags are
+  passed once per binary instead of piling up with each one stripped
+  (Arthur Zamarin)
+
+- ``pkgcore.config.basics``: an inline config section of the wrong type is
+  reported as such instead of failing with an ``AttributeError``
+  (Arthur Zamarin)
+
+- ``pquery -`` and ``patom -F FORMAT -`` skip blank lines read from stdin
+  instead of rejecting them as invalid atoms (Arthur Zamarin)
+
+- ``pquery -1`` stops at the first match instead of printing one per repo
+  (Arthur Zamarin)
+
+- ``pkgcore.fs.fs.get_major_minor``: device files with a major or minor number
+  above 255, such as NVMe disks, get their real numbers instead of truncated
+  ones (Arthur Zamarin)
+
+- ``pkgcore.util.commandline.StoreRepoObject``: ``-r``/``--repo`` looks up
+  repos in the domain chosen with ``--domain`` instead of always in the default
+  domain (Arthur Zamarin)
+
+- ``FEATURES=multilib-strict`` without ``MULTILIB_STRICT_DIRS`` or
+  ``MULTILIB_STRICT_DENY`` set skips the check, as in Portage, instead of
+  crashing every merge (Arthur Zamarin)
+
+- profiles: a repo inside a directory named ``profiles`` gets the right
+  profile names and file paths (Arthur Zamarin)
+
+- ``pmaint regen``: ``_eclasses_`` in the metadata cache follows each
+  package's inherit order, so regenerating no longer reorders unchanged
+  entries (Arthur Zamarin)
+
 ----------------------------
 pkgcore 0.12.44 (2029-09-26)
 ----------------------------
@@ -62,6 +405,10 @@ Fixes
 - eclass: separate an ``@SUBSECTION`` heading from the prose above it, so it
   renders as a heading (Arthur Zamarin)
 
+- ``ContainmentMatch``: render its values in sorted order, so the string form of
+  a restriction is stable across runs rather than following set iteration order
+  (Arthur Zamarin)
+
 ----------------------------
 pkgcore 0.12.42 (2026-09-05)
 ----------------------------

{pkgcore-0.12.44 → pkgcore-0.12.46}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pkgcore
-Version: 0.12.44
+Version: 0.12.46
 Summary: package managing framework
 Author-email: Michał Górny <mgorny@gentoo.org>, Tim Harder <radhermit@gmail.com>, Brian Harring <ferringb@gmail.com>, Arthur Zamarin <arthurzam@gentoo.org>, Marien Zwart <marienz@gentoo.org>
 Maintainer-email: Arthur Zamarin <arthurzam@gentoo.org>

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/patom.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PATOM" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PATOM" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 patom \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pclean.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PCLEAN" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PCLEAN" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pclean \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pclonecache.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PCLONECACHE" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PCLONECACHE" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pclonecache \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pconfig.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PCONFIG" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PCONFIG" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pconfig \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pebuild.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PEBUILD" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PEBUILD" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pebuild \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pinspect.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PINSPECT" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PINSPECT" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pinspect \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pmaint.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PMAINT" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PMAINT" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pmaint \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pmerge.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PMERGE" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PMERGE" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pmerge \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/build/sphinx/man/pquery.1

@@ -28,7 +28,7 @@ level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
 ..
-.TH "PQUERY" "1" "Sep 26, 2026" "0.12.44" "pkgcore"
+.TH "PQUERY" "1" "Oct 08, 2026" "0.12.46" "pkgcore"
 .SH NAME
 pquery \- Wrapper for running commandline scripts.
 .SH SYNOPSIS

{pkgcore-0.12.44 → pkgcore-0.12.46}/data/lib/pkgcore/ebd/ebuild-daemon-lib.bash

@@ -25,6 +25,8 @@ __ebd_read_array() {
 
 # read -N usage requires bash-4.1 or so (EAPI 6 requires >= 4.2)
 __ebd_read_size() {
+	# python sends sizes in bytes
+	local LC_ALL=C
 	read -u ${PKGCORE_EBD_READ_FD} -r -N $1 $2
 	local ret=$?
 	[[ ${ret} -ne 0 ]] && \
@@ -94,7 +96,7 @@ __ebd_ipc_cmd() {
 # ask the python side to display sandbox complaints
 __request_sandbox_summary() {
 	local line
-	__ebd_write_line "__request_sandbox_summary ${SANDBOX_LOG}"
+	__ebd_write_line "request_sandbox_summary ${SANDBOX_LOG}"
 	__ebd_read_line line
 	while [[ ${line} != "end_sandbox_summary" ]]; do
 		echo "${line}"

{pkgcore-0.12.44 → pkgcore-0.12.46}/data/lib/pkgcore/ebd/ebuild-daemon.bash

@@ -135,8 +135,8 @@ __ebd_exec_main() {
 
 	# finally, load the master list of pkgcore funcs. fallback to
 	# regenerating it if needed.
-	if [[ -f ${PKGCORE_EBD_PATH}/generated/funcs/global ]]; then
-		PKGCORE_BLACKLIST_FUNCS+=( $(<"${PKGCORE_EBD_PATH}"/generated/funcs/global) )
+	if [[ -f ${PKGCORE_EBD_PATH}/.generated/funcs/global ]]; then
+		PKGCORE_BLACKLIST_FUNCS+=( $(<"${PKGCORE_EBD_PATH}"/.generated/funcs/global) )
 	else
 		PKGCORE_BLACKLIST_FUNCS+=( $("${PKGCORE_EBD_PATH}"/generate_global_func_list 2> /dev/null) )
 	fi
@@ -285,10 +285,8 @@ __ebd_process_ebuild_phases() {
 }
 
 __ebd_process_metadata() {
-	# protect the env.
-	# note the local usage is redundant in light of it, but prefer to write it this
-	# way so that if someone ever drops the (), it'll still not bleed out.
-	(
+	# only run via $(), which protects the env
+	{
 		# Heavy QA checks (IFS, shopt, etc) are suppressed for speed
 		declare -r PKGCORE_QA_SUPPRESSED=false
 		# Wipe __mode; it bleeds from our parent.
@@ -317,7 +315,7 @@ __ebd_process_metadata() {
 		__execute_phases "${2:-depend}" && exit 0
 		__ebd_process_sandbox_results
 		exit 1
-	)
+	}
 }
 
 __make_preloaded_eclass_func() {

{pkgcore-0.12.44 → pkgcore-0.12.46}/data/lib/pkgcore/ebd/ebuild.bash

@@ -86,13 +86,13 @@ alias __load_environ='{
 	PKGCORE_EXISTING_PATH=${PATH};
 	__timed_call __environ_sanitize_saved_env "${PKGCORE_TARGET_ENV}"
 	if [[ -n ${PKGCORE_PERF_DEBUG} ]]; then
-		echo "timing source ${PKGCORE_TARGET_ENV}" >&2
-		time source "${PKGCORE_TARGET_ENV}" >&2
-		echo "timed source ${PKGCORE_TARGET_ENV}" >&2
+		echo "timing source ${T}/.scrubbed-env" >&2
+		time source "${T}"/.scrubbed-env >&2
+		echo "timed source ${T}/.scrubbed-env" >&2
 	else
-		source "${PKGCORE_TARGET_ENV}" >&2
+		source "${T}"/.scrubbed-env >&2
 	fi
-	[[ $? == 0 ]] || die "sourcing saved env failed";
+	[[ $? == 0 ]] || die "sourcing scrubbed env failed";
 	__ensure_PATH "${PKGCORE_EXISTING_PATH}";
 	__timed_call __load_eapi_libs
 	__timed_call __source_bashrcs
@@ -502,26 +502,23 @@ __dump_metadata_keys() {
 	# be invoked after ebuild code has done it's thing, as such we no longer care,
 	# and directly screw w/ it for speed reasons- about 5% speedup in metadata regen.
 	set -f
-	local key phases phase
+	local key phases phase val out
+	local -a words
 	for key in "${PKGCORE_METADATA_KEYS[@]}"; do
 		if [[ ${key} == DEFINED_PHASES ]]; then
 			for phase in "${PKGCORE_EBUILD_PHASES[@]}"; do
 				__is_function "${phase}" && phases+=( ${phase} )
 			done
-			__ebd_write_line "key DEFINED_PHASES=${phases[@]:--}"
-		else
-			# deref the val, if it's not empty/unset, then spit a key command to EBD
-			# after using echo to normalize whitespace (specifically removal of newlines)
-			if [[ ${!key:-unset} != "unset" ]]; then
-				# note that we explicitly bypass the normal functions, and directly
-				# write to the FD. This is done since it's about 25% faster for our usage;
-				# if we used the functions, we'd have to subshell the 'echo ${!key}', which
-				# because of bash behaviour, means the content would be read byte by byte.
-				echo -n "key ${key}=" >&${PKGCORE_EBD_WRITE_FD}
-				echo ${!key} >&${PKGCORE_EBD_WRITE_FD}
-			fi
+			out+="key DEFINED_PHASES=${phases[@]:--}"$'\n'
+		elif [[ ${!key:-unset} != "unset" ]]; then
+			# normalize whitespace (specifically removal of newlines)
+			words=( ${!key} )
+			printf -v val '%s ' "${words[@]}"
+			out+="key ${key}=${val% }"$'\n'
 		fi
 	done
+	# a single write is notably faster than writing each key separately
+	printf '%s' "${out}" >&${PKGCORE_EBD_WRITE_FD}
 	set +f
 }
 

pkgcore-0.12.46/data/lib/pkgcore/ebd/helpers/9/pkg_postinst/ver_replacing

@@ -0,0 +1,10 @@
+#!/usr/bin/env pkgcore-ebuild-helper
+
+source "${PKGCORE_EBD_PATH}"/eapi/7/global.bash || die "failed sourcing eapi/7/global.bash"
+
+[[ $# -eq 2 ]] || die "Usage: ver_replacing <op> <ver>"
+
+for v in ${REPLACING_VERSIONS}; do
+	ver_test "${v}" "$@" && exit 0
+done
+exit 1

{pkgcore-0.12.44 → pkgcore-0.12.46}/data/lib/pkgcore/ebd/helpers/internals/helper-lib.bash

@@ -75,6 +75,5 @@ __helper_failed() {
 
 __helper_check_exit() {
 	[[ $1 == 0 ]] && return
-	shift
 	__helper_exit "$@"
 }

{pkgcore-0.12.44 → pkgcore-0.12.46}/data/lib/pkgcore/ebd/isolated-functions.bash

@@ -217,7 +217,7 @@ __var_push() {
 
 		# If the specified variable currently has a value we save it;
 		# otherwise, just push the variable name onto the stack.
-		if orig_val=$(declare -p ${var} 2>/dev/null); then
+		if [[ -v ${var} ]] && orig_val=$(declare -p ${var} 2>/dev/null); then
 			orig_val=${orig_val/declare *${var}=[\'\"]/${var}=}
 			orig_val=${orig_val%[\'\"]}
 		else

{pkgcore-0.12.44 → pkgcore-0.12.46}/examples/report_pkg_changes.py

@@ -4,7 +4,6 @@ import os
 import sys
 
 from snakeoil.bash import read_bash
-from snakeoil.osutils import listdir_files
 
 from pkgcore.config import load_config
 from pkgcore.ebuild.atom import atom
@@ -76,7 +75,9 @@ def parse_moves(location):
         return tuple(reversed(fname.split("-")))
 
     moves = {}
-    for update_file in sorted(listdir_files(location), key=get_key):
+    for update_file in sorted(
+        [e.name for e in os.scandir(location) if e.is_file()], key=get_key
+    ):
         for line in read_bash(pjoin(location, update_file)):
             line = line.split()
             if line[0] != "move":

{pkgcore-0.12.44 → pkgcore-0.12.46}/src/pkgcore/__init__.py

@@ -1,4 +1,4 @@
 __title__ = "pkgcore"
-__version__ = "0.12.44"
-__version_info__ = (0, 12, 44)
+__version__ = "0.12.46"
+__version_info__ = (0, 12, 46)
 __python_mininum_version__ = (3, 12, 0)

pkgcore-0.12.46/src/pkgcore/_verinfo.py

@@ -0,0 +1 @@
+version_info={'rev': '3bfcc03d4071d8c75ab927fd52fb0ef5e6c47a5f', 'date': 'Thu Oct 08 15:27:42 2026', 'timestamp': 1791473262, 'tag': '0.12.46', 'commits': 0}

{pkgcore-0.12.44 → pkgcore-0.12.46}/src/pkgcore/binpkg/repo_ops.py

@@ -13,9 +13,10 @@ __all__ = ("install", "operations", "replace", "uninstall")
 
 import os
 from os.path import join as pjoin
+from pathlib import Path
 
 from snakeoil.compression import compress_data
-from snakeoil.osutils import ensure_dirs, unlink_if_exists
+from snakeoil.osutils import ensure_dirs
 
 from ..fs import tar
 from ..log import logger
@@ -94,7 +95,7 @@ class install(repo_interfaces.install):
             os.chmod(tmp_path, 0o644)
         except Exception:
             try:
-                unlink_if_exists(tmp_path)
+                Path(tmp_path).unlink(missing_ok=True)
             except OSError as e:
                 logger.warning(f"failed removing {tmp_path!r}: {e}")
             raise