Apertis CI robot authored 2 months ago and Dylan Aïssi committed 2 months ago

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Apertis CI robot authored 4 months ago and Dylan Aïssi committed 4 months ago

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Apertis CI robot authored 8 months ago and Dylan Aïssi committed 8 months ago

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Apertis CI robot authored 10 months ago and Dylan Aïssi committed 10 months ago

Signed-off-by: Apertis CI robot <noreply+not_a_robot@apertis.org>

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

In contrary to Debian, we use by default -Wformat-overflow=2
and -Wformat-truncation=2 in Apertis, but due to the use of -Werror this
package FTBFS with: "cc1: all warnings being treated as errors".
In order to avoid this error, we don't treat these warnings as errors.

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Signed-off-by: Dylan Aïssi <dylan.aissi@collabora.com>

Signed-off-by: Ariel D'Alessandro <ariel.dalessandro@collabora.com>

Signed-off-by: Ariel D'Alessandro <ariel.dalessandro@collabora.com>

Signed-off-by: Ariel D'Alessandro <ariel.dalessandro@collabora.com>

Signed-off-by: Emanuele Aina <emanuele.aina@collabora.com>

glib2.0 (2.66.7-1) unstable; urgency=high

  * New upstream release
    - Fix another regression caused by the GHSL-2021-045 fixes in 2.66.6
    - Warn and fail on integer overflow in g_byte_array_new_take()
      for arrays larger than G_MAXUINT (similar to GHSL-2021-045)
    - Disallow using currently-undefined D-Bus connection or server flags,
      to prevent forward-compatibility problems with new security-sensitive
      flags that are likely to be introduced in GLib 2.68
  * Drop previous patches for GHSL-2021-045 regressions, applied upstream

glib2.0 (2.66.6-2) unstable; urgency=high

  * d/patches: Add proposed fixes for regressions in 2.66.6.
    Two functions that took either a positive length, or -1 to indicate
    strlen(), had assertions with the wrong sense in 2.66.6, causing some
    valid uses of those functions to regress.
  * d/p/debian/61_glib-compile-binaries-path.patch: Remove.
    This patch turns out to be unnecessary, and is harmful for
    cross-compiling. Thanks to Helmut Grohne (Closes: #982213)
  * Set high urgency to get the regression fixes into bullseye

glib2.0 (2.66.6-1) unstable; urgency=high

  * New upstream release
    - Fix various integer overflows, some of them potentially exploitable
      (GHSL-2021-045)

glib2.0 (2.66.5-1) unstable; urgency=medium

  * New upstream release, equivalent to 2.66.4-27-g0051c0635
  * Drop patches that were applied upstream

glib2.0 (2.66.4-4) unstable; urgency=medium

  * d/patches: Update patch series to upstream commit 2.66.4-27-g0051c0635
    - Improve test coverage for #977961
    - Stop valgrind reporting memory leaks in GSpawn in most cases
    - Partially revert security hardening from 2.66.4-2: allow
      DBUS_SESSION_BUS_ADDRESS to be taken from the environment by
      setcap executables (to avoid regressing gnome-keyring) and by
      setgid executables (to avoid regressing msmtp).
      (Closes: #981420, #981555)
      Note that this is likely to be reverted in GLib 2.70.x to provide
      better hardening. The D-Bus session bus is not designed to be used
      by processes that have elevated privileges.

glib2.0 (2.66.4-3) unstable; urgency=medium

  * Improve patch for #977961, and add basic test coverage

glib2.0 (2.66.4-2) unstable; urgency=medium

  * d/patches: Update patch series to upstream commit 2.66.4-18-g872181c4f
    (excluding Windows-specific changes)
    - Security hardening: in GIO, ignore various environment variables
      if GIO is (inadvisably) used in a setuid process without sanitizing
      the environment first, similar to CVE-2012-3524
    - Reject very long date strings early, instead of spending time
      normalizing and parsing them
    - Fix recursion in GPrivate
  * d/p/spawn-Don-t-set-a-search-path-if-we-don-t-want-to-search-.patch:
    Make the g_spawn family only search PATH if G_SPAWN_SEARCH_PATH is used.
    Previously, they would sometimes search /usr/bin:/bin:. for an
    executable they should have only loaded from the current working
    directory. In particular, this made gtk+3.0 fail its build-time tests
    if ImageMagick display(1) happened to be installed. (Closes: #977961)

glib2.0 (2.66.4-1) unstable; urgency=medium

  * New upstream release

glib2.0 (2.66.3-2) unstable; urgency=medium

  * Apply packaging changes from experimental to unstable:
    - postinst: Clean up outdated copies of GLib if present, to avoid
      infrequent upgrade issues on non-merged-/usr systems.
      See #911225 and #949395 for more information.
      (Closes: #896019, #954960, #955331)
  * Add myself to Uploaders
  * Standards-Version: 4.5.1 (no changes required)
  * Swap Homepage field to something more GLib-specific
  * d/gbp.conf: Change upstream branch to upstream/2.66.x.
    2.67.0 was already released, so it's inaccurate to say that 2.66.x
    is the latest.

glib2.0 (2.66.3-1+exp1) experimental; urgency=medium

  * Merge from unstable

glib2.0 (2.66.3-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - Improve performance of processing files hidden via ./.hidden
    - All other changes were already included in 2.66.2-1
  * Drop patches that were cherry-picked from upstream
  * Stop reverting gtk-doc dependency version.
    We now have a suitable gtk-doc in Debian.
  * Drop a patch that was not applied upstream.
    This was hoped to be a workaround for intermittent test failures, but
    doesn't seem to have had the desired effect in practice.
  * Mark the DBUS_COOKIE_SHA1 parts of gdbus-server-auth test as flaky.
    This is not reliable enough to always pass on buildds, but is too
    intermittent to be able to reproduce the failure in a development
    environment, and DBUS_COOKIE_SHA1 is not an important enough feature
    to justify failing the build for this.
    As with other flaky tests, we still run this as an autopkgtest in an
    attempt to get more useful information, but we ignore failure.

glib2.0 (2.66.2-1+exp1) experimental; urgency=medium

  * Branch for experimental
  * postinst: Clean up outdated copies of GLib to avoid infrequent
    upgrade issues on non-merged-/usr systems
    (Closes: #896019, #954960, #955331)

glib2.0 (2.66.2-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - Add some missing (nullable) and (not nullable) annotations
  * Drop patches that were cherry-picked from upstream
  * Update patch series to upstream 2.66.2-9-g4daaf303a
    - Fix race in socketclient-slow test
    - Cope with sending fds in a D-Bus message that takes multiple writes
    - Don't skip updating polled fd sources
    - Add G_GNUC_PRINTF annotation to g_trace_mark()
  * d/p/glib-tests-fileutils-Make-more-use-of-g_assert_no_errno.patch,
    d/p/glib-tests-fileutils-Fix-expectations-when-running-as-roo.patch:
    Add proposed patch to fix a test failure when running as root
    (Closes: #973271)
  * d/rules: Remove migration path from legacy -dbg package.
    This was most recently shipped in Debian 9, and we don't support
    upgrades from anything older than Debian 10.
  * Drop obsolete workaround for #887629.
    We don't support upgrades from versions older than Debian 10, so we can
    drop workarounds that were only relevant for the upgrade from 9 to 10.

glib2.0 (2.66.1-2) unstable; urgency=medium

  * Cherry-pick patches from the glib-2-66 branch upstream
    - Fixes the regression called out in 2.66.1-1's changelog.
  * Add-a-test-for-the-6-days-until-EOM-bug.patch,
    Fix-the-6-days-until-the-end-of-the-month-bug.patch: Cherry-pick upstream
    mr!1705 to not break on timezones built with `zic -b slim`

glib2.0 (2.66.1-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - A performance problem where timezones were reloaded from disk
      every time a GTimeZone was created has been fixed (upstream issue
      #2204), but this means that changes to /etc/localtime will not take
      effect until a process restarts. Future changes in a subsequent
      2.66.x release will improve this.
    - Security fix for incorrect scope/zone ID parsing in URIs
    - Fix invalid Pointer Arithmetic in g_path_get_basename
    - Fix cookie lifetimes in GDBus DBUS_COOKIE_SHA1 mechanism
    - Fix faulty logic in DNS TXT record parsing
    - trash portal: Handle portal failures
    - gio-tool-trash: Prevent recursion to speed up emptying trash
    - glist: Clarify that g_list_free() and friends only free an entire list
    - gdatetime: Avoid integer overflow creating dates too far in the past
    - Translation updates
  * d/p/glocalfile-Never-require-G_LOCAL_FILE_STAT_FIELD_ATIME.patch,
    d/p/gdbusauthmechanismsha1-Use-the-same-timeouts-as-libdbus.patch:
    Drop patches that were applied upstream

glib2.0 (2.66.0-2) unstable; urgency=medium

  * Team upload
  * d/p/glocalfile-Never-require-G_LOCAL_FILE_STAT_FIELD_ATIME.patch:
    Add proposed patch to fix file copying on ZFS and CIFS (Closes: #970228)
  * d/p/gdbus-server-auth-Don-t-usually-test-non-EXTERNAL-repeate.patch:
    Add proposed patch to work around DBUS_COOKIE_SHA1 test failures
  * d/p/Revert-gtk-doc-dependency-to-1.32.patch: Move to debian subdirectory.
    This patch is not intended to go upstream.

glib2.0 (2.66.0-1) unstable; urgency=medium

  * Team upload
  * New upstream stable release
   - Fix missing tab in makefile rule
   - guri: Fix user passed to g_uri_split_with_user() not being NULL'd
   - Translation updates:
  * d/watch: Only watch for stable releases
  * d/p/gdbusauthmechanismsha1-Use-the-same-timeouts-as-libdbus.patch:
    Add patch to fix intermittent test failures on slower architectures.
    This narrowly missed the upstream code freeze, and should be in 2.66.1.

glib2.0 (2.65.3-1) experimental; urgency=medium

  * New upstream release
    + Fixes to the new `statx()` calls — note that since GLib 2.65.2 uses
      `statx()` (if available) instead of
      `stat()`/`fstat()`/`lstat()`/`fstatat()`, syscall sandboxing for third
      party applications might need to be updated
    + Also includes "Fix splice behavior on cancellation", a fix for a bug
      which was affecting tracker - particularly its autopkgtests.

glib2.0 (2.65.2-1) experimental; urgency=medium

  * Team upload
  * New upstream development release
  * d/rules: Run gtk-doc checks, even if building indep-only.
    Previously we would only run the gtk-doc checks if building
    architecture-dependent and -independent packages in the same build,
    which is done on Ubuntu amd64 buildds, but not on any Debian buildds.
  * Reduce dependency to the version of gtk-doc-tools from unstable.
    Instead of being some random snapshot from upstream git, this is the
    last release plus some selected patches. In particular, it has enough
    fixes to make the gtk-doc tests pass (Closes: #968975).
  * d/libglib2.0-tests.lintian-overrides: Update

glib2.0 (2.65.1-1) experimental; urgency=medium

  [ Sebastien Bacher ]
  * debian/control.in:
    - let libglib2.0-tests Depends on libglib2.0-0 (= ${binary:Version}),
      otherwise we can end up with failures due to out of sync versions

  [ Simon McVittie ]
  * d/shlibs.local: Upgrade all binary packages in lockstep.
    Like many projects where one source package builds multiple binary
    packages, GLib has private headers that share non-public interfaces
    between its binary packages. Instead of setting this up for individual
    binary packages, we can tell dpkg-shlibdeps to generate lockstep
    dependencies whenever one of our binary packages depends on our shared
    libraries.
  * d/watch, d/control.in, d/gbp.conf: Branch for experimental
  * New upstream development release
    - Require the experimental version of gtk-doc-tools.
      GLib 2.65.x requires a version that hasn't been released yet.
    - Update symbols file
    - Drop patches that were applied upstream

glib2.0 (2.64.4-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - Improve async-signal-safety
  * d/tests/build: Don't exercise static linking for GIO.
    libmount will no longer support being linked statically from 2.35.2-8
    onwards. For now I'm continuing to test that the other libraries can
    still be statically linked, but please consider them to be "at risk".
    (Closes: #963933)
  * Re-enable libmount support.
    libmount no longer depends on libcryptsetup, avoiding the various
    crashes that we are working around. Future versions will dlopen it
    on-demand, which should also avoid those crashes. Bump the
    build-dependency to a suitable version.
  * d/p/tests-Use-g_assert_-in-cancellable-test-rather-than-g_ass.patch,
    d/p/gcancellable-Fix-minor-race-between-GCancellable-and-GCan.patch:
    Split combined d/p/git_gsource_segfault.patch into its two component
    upstream commits, and add metadata
  * d/p/glib-compile-resources-Fix-exporting-on-Visual-Studio.patch,
    d/p/gdesktopappinfo-Fix-unnecessarily-copied-and-leaked-URI-l.patch:
    Add post-release bugfixes from upstream

glib2.0 (2.64.3-2) unstable; urgency=medium

  * Team upload
  * Temporarily disable libmount support.
    Recent Debian revisions of libmount pull in libcryptsetup as a
    dependency, for dm-verity support. libcryptsetup depends on json-c
    and OpenSSL, causing crashes due to symbol conflicts with other
    JSON libraries (jansson and json-glib, for example in firewalld and
    virt-manager) and with statically-linked copies of OpenSSL (for
    example in Steam and Minecraft). Until this is resolved in some
    other way, disable libmount and parse /etc/fstab and /proc/mounts
    ourselves, as we do in libglib2.0-udeb.
    Mitigates: #963933, #963932, #963525, #963721

glib2.0 (2.64.3-1) unstable; urgency=medium

  * Team upload

  [ Laurent Bigonville ]
  * Drop the libgio-fam package, and install the fam GIO plugin in
    libglib2.0-0 on Hurd ports. See: #885011 (Closes: #875915)
  * Stop building the libgio-fam package on kFreeBSD ports.
    It is no longer necessary now that gkqueuefilemonitor is available.

  [ Simon McVittie ]
  * Clarify changelog entry regarding Hurd and kFreeBSD
  * New upstream stable release

glib2.0 (2.64.2-1) unstable; urgency=medium

  [ Simon McVittie ]
  * Add Breaks on older versions of gimp, which used a syntactically
    invalid property name in a plugin, and would crash when GObject
    rejects syntactically invalid property names

  [ Sebastien Bacher ]
  * New upstream release
  * debian/patches/git_gsource_segfault.patch:
    - backport an upstream git change to fix a signal handler disconnect
      segfault situation (lp: #1872153)

glib2.0 (2.64.1-1) unstable; urgency=medium

  * Team upload
  * New upstream stable release
  * d/p/tests-Skip-MemoryMonitor-test-if-GObject-Introspection-is.patch:
    Drop patch, applied upstream
  * Add Breaks on glib-networking-tests older than 2.63.2.
    Those versions had a test that relied on TLS version fallback
    behaviour that has now been removed. (Closes: #953766)

glib2.0 (2.64.0-2) unstable; urgency=medium

  * Team upload
  * Merge packaging changes from unstable with new upstream release from
    experimental
  * d/control.in: Add Breaks on libgladeui-2-6 before 3.22.2.
    Older versions used a syntactically invalid property name
    "support warning", which GObject used to canonicalize to
    "support-warning". GLib 2.64 made this check more strict (see #953010).

glib2.0 (2.64.0-1) experimental; urgency=medium

  * Team upload
  * New upstream release
    - Fixes a vulnerability where GSocketClient sometimes forgot to use
      a configured proxy (CVE-2020-6750, Closes: #948554)
    - Stop installing gio-launch-desktop, which no longer exists
    - d/p/docs-Don-t-install-object-manager-example-separately.patch:
      Drop, applied upstream
  * d/p/debian/testfilemonitor-Skip-if-we-are-avoiding-flaky-tests.patch:
    Treat testfilemonitor as a flaky test
  * Standards-Version: 4.5.0 (no changes required)
  * New upstream release
  * d/p/tests-Skip-MemoryMonitor-test-if-GObject-Introspection-is.patch:
    Skip MemoryMonitor test if GObject-Introspection is too old to know it
  * Install a shell script implementation of the old gio-launch-desktop
    executable. While not required for *this* GLib, it is required by
    old processes that already had the old GLib (2.57.2 to 2.63.5) in
    memory before the upgrade. This can be removed after Ubuntu 20.04
    and Debian 11 are both released.

glib2.0 (2.63.5-2) experimental; urgency=medium

  * Skip-unreliable-gdbus-threading-tests--by-default.patch: Skip all of
    gdbus-threading test_method_calls_in_thread() has become (more?)
    unreliable too. When skipped, the test bus doesn't get torn down properly
    - it times out. Let's stop running these tests for now, until they are
    made reliable.

glib2.0 (2.63.5-1) experimental; urgency=medium

  [ Iain Lane ]
  * New upstream release

  [ Philip Withnall ]
  * Rework 01_gettext-desktopfiles.patch to not add new public API.
    Downstreams should not be adding new public API to GLib. From some code
    searching, this doesn’t appear to be used in more than one or two places,
    so won’t be too inconvenient to drop.  The original patch should either be
    upstreamed (I’d be open to some form of it, if there’s still evidence it’s
    useful) or dropped. If it’s upstreamed, the new keys should be
    standardised.  The alternative to this was to document the added public
    API; its addition was causing the new gtk-doc tests in GLib to fail.
  * Bump gtk-doc-tools dependency to >= 1.32-4 as GLib requires some fixes
    pushed to gtk-doc after its 1.32 release.
  * control: Bump Meson dependency to >= 0.52.0 for building the documentation
  * Drop PKCS#11 APIs added in 2.63.1 (not stable yet)
      - g_tls_certificate_new_from_pkcs11_uris()
  * Remove patches applied upstream:
    - tests-Skip-GMemoryMonitor-tests-if-the-dbusmock-template-.patch
    - tests-optional-portal.patch
  * d/p/docs-Don-t-install-object-manager-example-separately.patch: Add patch
    from upstream to disable incorrect installation of some example
    documentation
  * Rework libmount Meson argument as it’s now a feature; see !1344 upstream

glib2.0 (2.63.3-3) experimental; urgency=medium

  * debian/control.in:
    - lower the libglib2.0-tests Depends on xdg-desktop-portal to a
      Recommends since the portal is not available on some architectures
  * debian/patches/tests-optional-portal.patch:
    - skip the new memory monitor tests if the portal is not available,
      that allows the tests to be still successful on architectures were
      the portal is not available (e.g Ubuntu/i386)

glib2.0 (2.63.3-2) experimental; urgency=medium

  * debian/control.in:
    - libglib2.0-tests Depends on xdg-desktop-portal, it's required by the
      new low memory tests (and got enabled by the new python-dbusmock)

glib2.0 (2.63.3-1) experimental; urgency=medium

  [ Iain Lane ]
  * New upstream release
    + Add a `--glib-min-version` argument to `gdbus-codegen` which controls
      breaks in the API of generated code
    + Add `g_clear_list()` API to clear `GList`s to `NULL`
    + Add a `GMemoryMonitor` API to be notified of memory pressure situations
      using the low-memory-monitor project
    + Add support for dispose functions for `GSource` implementations
    + Tighten up validation of GObject signal and property names, allowing
      performance improvements
  * debian/tests/build: Style fixes, thanks to shellcheck.
  * d/p/d/Disable-some-tests-on-slow-architectures-which-keep-faili.patch:
    Rebase. Upstream have disabled these tests by default too (unless slow
    mode is enabled), so we don't need to add a patch to do a similar thing.
  * debian/libglib2.0-0.symbols: New symbols for 2.63.3
  * d/p/tests-Skip-GMemoryMonitor-tests-if-the-dbusmock-template-.patch: Add.
    We don't have a new enough dbusmock in Debian at the minute (one is not
    released yet). Skip the test if the required template isn't available.
  * control: Add Depends for the new memory-monitor tests.
    There are new tests, written in python, for GMemoryMonitor. They require
    dbus-python, pygobject, and the GI bindings for GLib and GIO.

  [ Steve Langasek ]
  * debian/tests/build: Make cross-test friendly
    autopkgtest is soon to get a `-a ARCHITECTURE` switch, which will
    cross-test autopkgtests. This is to be detected by the presence of the
    `dpkg-architecture`-style family of variables being set in the
    environment.
    For build tests like `glib2.0`'s `build` test, this means that we should
    test "${DEB_HOST_ARCH}" and invoke the cross toolchain as necessary.
    (Closes: #946355)

glib2.0 (2.63.2-1) experimental; urgency=medium

  * Team upload
  * Merge packaging changes from unstable
    - Support for pkg.glib2.0.noinsttest build profile
  * d/control.in: Refer to debian/experimental branch.
    This avoids false-positive warnings from vcswatch.
  * New upstream release
    - Drop patches that were applied upstream
  * Rename pkg.glib2.0.noinsttest build profile to noinsttest.
    This is now registered on <https://wiki.debian.org/BuildProfileSpec>.
  * Update symbols file

glib2.0 (2.63.1-2) experimental; urgency=medium

  * d/p/Revert-glocalfileinfo-Only-return-file-mode-not-type-as-U.patch:
    Revert "glocalfileinfo: Only return file mode, not type, as UNIX_MODE
    attribute" This reverts commit bfdc5fc4fc84ef8518d2d1a328c8482cf5a38e98.
    This commit changes the semantics of the `unix::mode` attribute, which
    some things (the one we've noticed is ostree) rely on.
  * d/p/test_copy_preserve_mode-Adjust-for-revert-semantics.patch:
    test_copy_preserve_mode: Adjust for revert semantics. Now we're returning
    the file type again, we need to mask it out to compare with the mode.

glib2.0 (2.63.1-1) experimental; urgency=medium

  * New upstream release
    - Add `g_array_steal()`, `g_ptr_array_steal()` and `g_byte_array_steal()`
      APIs
    - Add `g_get_os_info()` API
    - Add `GMainContextPusher` API
    - Add `g_warning_once()` API
    - Allow passing empty `GValue`s to `g_param_value_set_default()`
    - Always resolve `localhost` to loopback address in `GResolver`
    - Escape header guards generated by `gdbus-codegen` better
    - Fix crash in `g_spawn()` with high FD numbers due to use of `select()`
      rather than `poll()`
    - Limit recursion in `g_variant_parse()`
    - Several usability improvements to command line `gio` tool
  * debian/libglib2.0-0.symbols: Add new symbols for this release
  * debian/patches/*:
    - Drop backports we had which are in this release.
    - Update to upstream master at cc1b53f74. There are several test fixes
      that we might as well grab now.

glib2.0 (2.62.5-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - Fixes a vulnerability where GSocketClient sometimes forgot to use
      a configured proxy (CVE-2020-6750, Closes: #948554)
  * Build-depend on libnss-myhostname | netbase if running tests.
    This is an attempt to work around localhost not being a resolvable
    name in some build environments, notably reproducible-builds.
    (See #948834)
  * Put the result of `getent ahosts localhost` and
    `getent ahosts $(hostname)` in the build log, to check whether those
    names are resolvable in the build environment.
  * d/p/debian/testfilemonitor-Skip-if-we-are-avoiding-flaky-tests.patch:
    Treat testfilemonitor as a flaky test
  * Standards-Version: 4.5.0 (no changes required)

glib2.0 (2.62.4-2) unstable; urgency=medium

  * Team upload
  * Adjust dependencies to avoid broken partial upgrades on arm64 during
    libffi7 transition:
    - Bump versioned Depends on libffi-dev to get a guarantee that we'll
      depend on libffi7
    - Add Breaks on libgirepository-1.0-1 (<< 1.62.0-4~) so we cannot
      get a GObject built with libffi7 but a libgirepository built with
      libffi6

glib2.0 (2.62.4-1) unstable; urgency=medium

  * Team upload

  [ Steve Langasek ]
  * debian/tests/build: Make cross-test friendly (Closes: #946355)

  [ Iain Lane ]
  * debian/tests/build: Style fixes

  [ Simon McVittie ]
  * New upstream release

glib2.0 (2.62.3-2) unstable; urgency=medium

  * Team upload
  * Rename pkg.glib2.0.noinsttest build profile to noinsttest.
    This is now registered on <https://wiki.debian.org/BuildProfileSpec>.

glib2.0 (2.62.3-1) unstable; urgency=medium

  * Team upload
  * New upstream release
    - Drop patches that were applied upstream
  * Don't build libglib2.0-tests under pkg.glib2.0.noinsttest build profile.
    This is a prototype of the proposed standard build profile noinsttest.
    If the build profiles include both nocheck and pkg.glib2.0.noinsttest,
    we can drop the libdbus-1-dev build-dependency without harming test
    coverage or altering the contents of binary packages.
  * d/gbp.conf: Use upstream/2.62.x branch

glib2.0 (2.62.2-3) unstable; urgency=medium

  * Team upload

  [ Iain Lane ]
  * control: Drop `debian/experimental` from Vcs-*

  [ Simon McVittie ]
  * Build-depend on libdbus-1-dev for better test coverage
  * Update to upstream commit 2.62.2-28-g3cf25070e:
    - d/p/goption-Relax-assertion-to-avoid-being-broken-by-kdeinit5.patch:
      Fix assertion failure when called from a process that overwrites its
      argv, such as kdeinit5
    - d/p/gdbus-peer-Specifically-listen-on-127.0.0.1.patch:
      Improve reliability of gdbus-peer test in some container environments
    - d/p/gdbusserver-Delete-socket-and-nonce-file-when-stopping-se.patch,
      d/p/gdbusserver-Keep-a-strong-reference-to-the-server-in-call.patch,
      d/p/gdbusauthmechanismsha1-Remove-unnecessary-g_warning-calls.patch,
      d/p/gdbusauthmechanismsha1-Create-.dbus-keyrings-directory-re.patch,
      d/p/tests-Move-main-loop-and-test-GUID-into-test-functions-in.patch,
      d/p/tests-Isolate-directories-in-gdbus-peer-test.patch,
      d/p/gdbus-peer-test-Improve-diagnostics-if-g_rmdir-fails.patch,
      d/p/gdbus-peer-test-Stop-GDBusServer-before-tearing-down-temp.patch,
      d/p/gdbus-peer-test-Use-unix-dir-address-if-exact-format-does.patch,
      d/p/gdbus-server-auth-test-Create-temporary-directory-for-Uni.patch:
      Mark as applied upstream in 2.62.x branch
  * d/p/gdbus-server-auth-test-Include-gcredentialsprivate.h.patch:
    Apply patch from 2.63.x to fix missing coverage in test for #941018
  * d/p/Make-ld-executable-configurable.patch:
    Apply patch from 2.63.x to use cross ld where necessary
  * d/p/gdbus-server-auth-test-Create-temporary-directory-for-Uni.patch:
    Mark as applied upstream in 2.63.x branch
  * Improve patch metadata: use more URLs for bug references

glib2.0 (2.62.2-2) unstable; urgency=medium

  * Team upload
  * Update to upstream commit 2.62.2-14-gfcbb88823:
    - d/p/gdesktopappinfo-Allocate-DesktopFileDir-structs-dynamical.patch,
      d/p/gdesktopappinfo-Cancel-file-monitor-when-resetting-a-Desk.patch,
      d/p/glocalfilemonitor-Keep-a-weak-ref-to-the-monitor-in-GFile.patch:
      Fix intermittent test failures for GDesktopAppInfo (Closes: #941550)
    - d/p/gvariant-Limit-recursion-in-g_variant_parse.patch:
      Ensure that parsing a text-format GVariant does not run out of stack
      space
    - d/p/tests-Use-objcopy-from-the-cross-compilation-file-if-conf.patch,
      d/p/docs-Add-objcopy-to-example-cross-compilation-file.patch:
      Use the appropriate architecture's objcopy when cross-compiling
    - d/p/gtestutils-Add-additional-non-NULL-check-in-g_assert_cmpm.patch:
      Avoid false positive NULL dereference warnings in g_assert_cmpmem()
    - d/p/gspawn-Port-to-g_poll-from-select.patch:
      Fix launching subprocesses when a very large number of fds are open
    - d/p/gcredentialsprivate-Document-the-various-private-macros.patch,
      d/p/credentials-Invalid-Linux-struct-ucred-means-no-informati.patch,
      d/p/GDBus-prefer-getsockopt-style-credentials-passing-APIs.patch:
      Ensure libdbus clients can authenticate with a GDBusServer like the
      one in ibus (Closes: #941018)
  * d/p/gdbusserver-Delete-socket-and-nonce-file-when-stopping-se.patch,
    d/p/gdbusserver-Keep-a-strong-reference-to-the-server-in-call.patch,
    d/p/Add-a-test-for-GDBusServer-authentication.patch:
    Backport regression test for #941018 from upstream git master
  * d/p/gdbusauthmechanismsha1-Remove-unnecessary-g_warning-calls.patch,
    d/p/gdbusauthmechanismsha1-Create-.dbus-keyrings-directory-re.patch,
    d/p/tests-Move-main-loop-and-test-GUID-into-test-functions-in.patch,
    d/p/tests-Isolate-directories-in-gdbus-peer-test.patch:
    Backport reliability fixes for gdbus-peer test from upstream git master
  * d/p/gdbus-peer-test-Improve-diagnostics-if-g_rmdir-fails.patch,
    d/p/gdbus-peer-test-Stop-GDBusServer-before-tearing-down-temp.patch,
    d/p/gdbus-peer-test-Use-unix-dir-address-if-exact-format-does.patch,
    d/p/gdbus-server-auth-test-Create-temporary-directory-for-Uni.patch:
    Add some proposed patches to improve GDBus unit tests
  * d/p/debian/mimeapps-test-Mark-as-flaky.patch:
    Drop patch, hopefully no longer needed with #941550 fixed
  * d/p/debian/taptestrunner-Stop-looking-like-an-executable-script.patch:
    Make taptestrunner non-executable to avoid a Lintian warning

glib2.0 (2.62.2-1) unstable; urgency=medium

  * New upstream release
    + Fixes use after free when calling g_dbus_connection_flush_sync() in a
      dedicated thread (LP: #1848202)

glib2.0 (2.62.1-1) unstable; urgency=medium

  * Team upload
  * d/watch: Only watch for even-numbered (stable) releases
  * New upstream release
    - Fix regression that made G_FILE_COPY_TARGET_DEFAULT_PERMS result in
      private permissions rather than respecting umask (Closes: #505398)
    - d/p/g_file_info_get_modification_date_time-Calculate-in-integ.patch,
      d/p/Always-build-tests-if-we-enabled-installed-tests.patch:
      Drop patches that were applied upstream
  * d/p/debian/mimeapps-test-Mark-as-flaky.patch:
    Mark mimeapps test as flaky (see #941550)

glib2.0 (2.62.0-3) unstable; urgency=medium

  * Team upload
  * Merge packaging from 2.60.x branch previously in unstable
    - No changes since 2.62.0-2, except in d/changelog
    - d/p/debian/Disable-an-optimization-when-building-with-gcc-9.patch:
      Remove workaround for #931921, which turned out to be a clutter bug
  * d/p/Always-build-tests-if-we-enabled-installed-tests.patch:
    Add patch to fix installation of installed-tests in cross-builds
    (Closes: #941509)
  * d/p/g_file_info_get_modification_date_time-Calculate-in-integ.patch:
    Add patch to fix intermittent g-file-info test failures on i386
    (Closes: #941547)
  * libglib2.0-dev: Suggest libgirepository1.0-dev, for the GIR files
    (Closes: #914152)
  * d/gbp.conf: Use debian/master branch
  * Standards-Version: 4.4.1 (no changes required)

glib2.0 (2.60.6-2) unstable; urgency=medium

  * Team upload
  * d/rules: Edit debcrossgen output instead of using a modified version.
    This fixes use of CFLAGS, etc. during cross-compilation.
    (Closes: #933560)
  * Remove obsolete permissions fixing.
    Issue 1539 was fixed upstream.
  * d/p/debian/Disable-an-optimization-when-building-with-gcc-9.patch:
    Disable an optimization when building with gcc-9, instead of forcing
    gcc-8. This avoids depending on an old gcc, and should be easier to
    deal with for cross-compilation. (Workaround for #931921)
  * d/p/gmessages-Only-use-structured-logs-if-GLIB_VERSION_MAX_AL.patch:
    Update to upstream glib-2-60 branch at commit 2.60.6-2-ga365528f6
    - Don't use structured logging if GLIB_VERSION_MAX_ALLOWED < 2.56

glib2.0 (2.62.0-2) unstable; urgency=medium

  * Team upload.
  * Upload to unstable. (Closes: #940161)

glib2.0 (2.62.0-1) experimental; urgency=medium

  * New upstream release
    + Fix new `GFileInfo` APIs to work when
      `G_FILE_ATTRIBUTE_TIME_MODIFIED_USEC` was not queried

glib2.0 (2.61.3-1) experimental; urgency=medium

  * New upstream release
  * d/p: Drop cherry-picks from upstream branch which we now have
  * d/p/d/Disable-an-optimization-when-building-with-gcc-9.patch: Drop,
    clutter has been fixed now (thanks Simon)
  * d/p/*: Refresh via gbp-pq as necessary

glib2.0 (2.61.2-2) experimental; urgency=medium

  * Team upload
  * d/p/cond-test-Don-t-make-assumptions-about-struct-sigaction-m.patch:
    Add proposed patch to fix FTBFS due to a test failure on mips*

glib2.0 (2.61.2-1) experimental; urgency=medium

  * Team upload
  * New upstream release
  * d/patches: Update to upstream git master, commit 2.61.2-23-g870b30bd7
    - Fix regression in g_mkdir_with_permissions()
    - Fix a memory leak
    - Update translations: es, id, ro
  * Merge changes from unstable
  * Refresh patch series
  * d/p/debian/06_thread_test_ignore_prctl_fail.patch:
    Use g_test_skip() when skipping test
  * d/p/GIO-tests-Don-t-do-clever-tricks-with-objcopy.patch:
    Drop workaround for #932287, and build-depend on fixed binutils on
    mips64el instead
  * d/rules: Edit debcrossgen output instead of using a modified version.
    This fixes use of CFLAGS, etc. during cross-compilation.
    (Closes: #933560)
  * d/libglib2.0-0.symbols: Update
  * Remove obsolete permissions fixing.
    Issue 1539 was fixed upstream.
  * libglib2.0-tests: Depend on libglib2.0-dev-bin.
    This is required for the new mkenums and genmarshal tests.
  * d/p/debian/Disable-an-optimization-when-building-with-gcc-9.patch:
    Disable an optimization when building with gcc-9, instead of forcing
    gcc-8. This avoids depending on an old gcc, and should be easier to
    deal with for cross-compilation. (Workaround for #931921)

glib2.0 (2.60.6-1) unstable; urgency=medium

  * Team upload
  * New upstream release, functionally equivalent to 2.60.5 with the
    patches we were already applying
    - d/p/portal-Add-a-getter-for-dconf-access.patch,
      d/p/settings-Tweak-priorities-for-keyfile-backend.patch,
      d/p/key-file-Handle-filename-being-NULL.patch:
      Drop, applied upstream
  * d/p/tests-Fix-data-race-in-gmenumodel-test.patch,
    d/p/tests-Fix-data-race-in-task-test.patch:
    Add patches from upstream git master to fix data races in tests.
    In particular, the one for gmenumodel might solve an unreproducible
    test failure on i386 (see #932678).
  * d/p/debian/gmenumodel-test-Mark-as-flaky.patch,
    d/p/debian/gvariant-test-Don-t-run-at-build-time-on-mips.patch:
    Skip more tests at build-time and during the non-flaky autopkgtest.
    The unreproducible gmenumodel test failure on i386 might in fact be
    fixed by d/p/tests-Fix-data-race-in-gmenumodel-test.patch, but it's
    hard to be sure about that. The gvariant fuzz test is catastrophically
    slow on certain mips CPUs and so is impractical to run there.
    (Closes: #932678)
  * Standards-Version: 4.4.0 (no changes required)
  * Use debhelper compat level 12
    - Stop explicitly passing -V to dh_makeshlibs, it is now the default
    - Disable dh_dwz for libglib2.0-udeb.
      This avoids an apparent debhelper bug in which dh_dwz generates
      multifiles for udebs, but dh_strip does not remove them from the
      udeb's staging directory. (Workaround for #933212)
  * Stop overriding libexecdir.
    Since FHS 3.0 (Policy 4.1.5), /usr/libexec is considered valid,
    and since debhelper compat level 12 it is the default.
    In this particular package this only affects the installed-tests.
  * Remove an obsolete Lintian override

glib2.0 (2.60.5-1) unstable; urgency=medium

  * Team upload
  * Prepare GLib 2.60.x stable branch for unstable
  * New upstream release
  * d/p/portal-Add-a-getter-for-dconf-access.patch,
    d/p/settings-Tweak-priorities-for-keyfile-backend.patch,
    d/p/key-file-Handle-filename-being-NULL.patch:
    Add post-release fixes from upstream glib-2-60 branch
  * d/p/GIO-tests-Don-t-do-clever-tricks-with-objcopy.patch:
    Don't do strange things with objcopy while testing GResource,
    while we work out what is going on in mips64el builds.
    Mitigates: #932287

glib2.0 (2.61.1-2) experimental; urgency=medium

  * control, rules: Build with gcc-8. See #931921 - when we're built with
    gcc-9, some applications that use GLib might start hanging.

glib2.0 (2.61.1-1) experimental; urgency=medium

  * New upstream release
    + `g_unichar_isxdigit()` and `g_unichar_xdigit_value()` now handle
       full-width characters (U+FF21–U+FF26 and U+FF41–U+FF46)
    + Deprecate `gtester` utility and its test reporting format and enable TAP
      output by default instead — the `--tap` option to tests is now a no-op
    + Add `g_test_summary()` to allow test authors to programmatically summarise
      what each unit test in a test suite does
    * Upgrade to Unicode Character Database v12.1
    + More IPv6 Happy Eyeballs fixes to `GNetworkAddress` and `GSocketClient`
    + Fix valgrind and gdb support for the new `GHashTable` changes
    + Fix GTask wait times growing faster than the number of task threads
    + Add `g_autoqueue()` helper macros, similar to `g_autolist()`
    + Add pre-allocated link helpers for `GList` and `GQueue`:
      - `g_list_insert_before_link()`
      - `g_queue_insert_before_link()`
      - `g_queue_insert_after_link()`
    + Improve network availability detection with NetworkManager to treat lower
      levels of connectivity as having reduced availability
    + Add `g_clear_signal_handler()` to allow disconnecting from a `GObject`
      signal and clearing the signal handler ID to zero in a single call
    + Add `g_autoptr()` support for `GRWLock`
  * 81-skip-monitor-test-on-non-linux.patch: Drop. This test was removed
    upstream.
  * debian/libglib2.0-0.symbols: Update

glib2.0 (2.60.4-1) experimental; urgency=medium

  * New upstream release (LP: #1832457)
  +  Fixes to improved network status detection with NetworkManager (#1788)
  + Leak fixes to some `glib-genmarshal` generated code
  + Further fixes to the Happy Eyeballs (RFC 8305) implementation
  + File system permissions fix to clamp down permissions in a small time window
    when copying files (CVE-2019-12450)
  + Bugs fixed:
   - Please revert #535 gmacros: Try to use the standard __func__ first in
     G_STRFUNC
  * gfile-Limit-access-to-files-when-copying.patch: Drop. It's in this version
    upstream.
  * d/p/*: Refresh through gbp pq

glib2.0 (2.60.3-2) experimental; urgency=medium

  * Team upload
  * d/p/gfile-Limit-access-to-files-when-copying.patch:
    Backport patch from upstream to ensure files don't temporarily have
    less restrictive permissions during copying
    (Closes: #929753, CVE-2019-12450)
  * Drop d/p/debian/04_homedir_env.patch:
    Remove legacy support for Debian-specific G_HOME environment variable.
    GLib has respected the HOME environment variable in the conventional
    way since wheezy. Only two packages (pygtk and ruby-graphviz) still
    set G_HOME without also setting HOME.
  * Drop d/p/debian/90_gio-modules-multiarch-compat.patch:
    Remove legacy support for loading modules from /usr/lib/gio/modules.
    All the known packages containing GIO modules (dconf, glib-networking
    and gvfs) migrated to /usr/lib/${DEB_HOST_MULTIARCH}/gio/modules
    before wheezy.
  * Register documentation in doc-base via symlinks in /usr/share/doc.
    The doc-base specification requires this, presumably for the benefit
    of tools that export /usr/share/doc via HTTP (see #925200), and
    Lintian 2.12 added a warning for not doing so.

glib2.0 (2.60.3-1) experimental; urgency=medium

  * New upstream releases 2.60.1 2.60.2 2.60.3, fixing many bugs:
    + build: Fix a typo in the test whether _NL_ABALTMON_n is supported
    + Critical in g_socket_client_async_connect_complete
    + Fix crash when displaying notifications on macOS (!786)
    + Fix documentation for `gdbus-tool wait` to use correct units
    + Fix typo in German translation
    + glib/date test fails
    + glib/gconstructor.h: Include stdlib.h for MSVC builds
    + GNetworkAddressAddressEnumerator unsafely modifies cache in
      GNetworkAddress
    + gnetworkaddress: Fix parallel enumerations interfering with eachother
    + gnetworkmonitornm: Fix network available detection
    + GResource generation test incompatible with stable LLVM on Linux
    + gschema.dtd: Add target attribute to alias
    + gsocketclient: Fix a leak in the connection code
    + Improve network status detection with NetworkManager
    + Leaks in gsocketclient.c connection code
    + test_month_names: assertion failed
    + tests: Update month name check for Greek locale
    + Update gdb pretty-printer for GHashTable
    + Various fixes to small key/value support in `GHashTable`
    + New GHashTable implementation confuses valgrind
    + more GHashTable fixes
    + GDB pretty-printer for GHashTable no longer works
    + ghash: Disable small-arrays under valgrind

glib2.0 (2.60.0-1) experimental; urgency=medium

  * New upstream release
    + Further fixes to the Happy Eyeballs (RFC 8305) implementation
    + Add support for the XDG trash portal
  * Call dh_python3 for all packages and add Depends to libglib2.0-tests.
    The tests now ship a "static-link.py" test. While it early-exits
    currently, we still need the right dependencies to be able to do that.
    Call dh_python3 for libglib2.0-test's private directory, and while we're
    here make sure that it is called for all packages too.
  * flaky autopkgtest: `closures` has been renamed to `closure-refcount`
  * Refresh and drop patches.
    - d/p/flaky-socket-service-test/*: Drop, applied upstream.
    - d/p/debian/closures-test-Skip-on-arm-unless-flaky-tests-are-allowed.patch:
      Refresh.

glib2.0 (2.59.3-2) experimental; urgency=medium

  [ Philip Withnall ]
  * debian: Drop GVariant alignment patches which are no longer needed.
    The upstream 2.59.0 release contained commit
    0f2a6c61c9c5e34d57293fb6987b21f3d1feb1cb, which automatically aligns
    GVariants at construction time. The realignment in the tests was a
    workaround for this.
    See upstream issue https://gitlab.gnome.org/GNOME/glib/issues/1342.

  [ Iain Lane ]
  * Cherry-pick fixes for upstream bug #1679 to fix a flaky test.
    This is apparently more flaky on the autopkgtest machines than elsewhere
    - it's quite consistently failing there.

glib2.0 (2.59.3-1) experimental; urgency=medium

  * New upstream release
    + Fix support for g_get_user_special_dir() on macOS, including support for
      the Downloads directory
    + Ensure that cancelling a GTask cannot cause its callback to be called
      synchronously (in the same call chain as the original *_async() call)
    + Further fixes to the Happy Eyeballs (RFC 8305) implementation
    + Various fixes for installation of installed tests
  * Refresh and drop patches.
    Drop gio-tests-Install-test1.overlay-file-when-building-instal.patch,
         tests-Install-the-slow-connect-preload.so-library-and-use.patch,
    fixed upstream.

glib2.0 (2.59.2-2) experimental; urgency=medium

  * Install `test1.overlay' file, which is needed for the installed tests.
  * tests: Install the slow-connect-preload.so library and use it

glib2.0 (2.59.2-1) experimental; urgency=medium

  * New upstream release
    + Fix check on GDBusMessage size when reading it.
    + Add async GIO API: g_file_query_default_handler_async(),
      g_app_info_launch_uris_async()
    + Fix some bugs in the Happy Eyeballs implementation.
    + Install a new generated header with enum types for Unicode enums.
    + Support the XDG trash portal.
    + Autotools support is gone.
    + g_format_size() now uses a no-break space to separate digits and units;
      translations will need to be updated accordingly.
    + New g_queue_clear_full() API.
    + Fix argument quoting on win32 when spawning subprocesses.
    + Allow polling more than 64 handles on win32 using g_poll().
    + Tag various tests as ‘flaky’. These are no longer run routinely on our
      upstream CI machines, and downstream packagers may want to not run them
      (or not treat those test failures as package build failures) on their
      test machines either. They are in the `flaky` test suite.
    + Add overlay support to g_resources_get_info().
    + Support defaults and locks in the keyfile GSettings backend. This will
      be used for flatpaks.
    + Accept unquoted strings in the keyfile GSettings backend to simplify
      things for sysadmins.
    + Update our contribution guidelines (`CONTRIBUTING.md`).
    + Add writev() and writev_all() APIs to GOutputStream and
      GPollableOutputStream, and provide implementations of them for many
      subclasses.
  * Refresh patches, drop trash test patches which are upstream now
  * build-time tests: Run "flaky" suite separately and non fatally
  * debian/libglib2.0-0.symbols: Add new symbols from 2.59.{1,2}

glib2.0 (2.59.0-1) experimental; urgency=medium

  * New upstream release
  * Update patches and drop upstreamed ones
  * debian/control: Update meson BD to ≥ 0.48.0 per upstream
  * debian/rules: enable-selinux takes "enabled" or "disabled" now.
    Not true/false.
  * debian/*: Update for experimental
  * debian/libglib2.0-0.symbols: Update with new symbols for this release

glib2.0 (2.58.3-3) unstable; urgency=medium

  * control, rules: Build with gcc-8. See #931921 - when we're built with
    gcc-9, some applications that use GLib might start hanging.