sysroot: Fix leak/double free of keyfile origin

Use autoptr rather than manual cleanup.  The double free isn't a
security problem, since we trust origin files.

Closes: #598
Approved by: jlebon

Gbp-Pq: Name ASAN-sysroot-Fix-leak-double-free-of-keyfile-origin.patch

traverse: Use g_hash_table_add

And "move semantics" via `g_steal_pointer()`.  Just a minor code
cleanup I noticed when I was hunting for a leak, which ended up being
elsewhere.

Closes: #598
Approved by: jlebon

Gbp-Pq: Name traverse-Use-g_hash_table_add.patch

cmdline: Fix minor leak in delta cmdline entrypoint

Small, but it's important to stay clean.

Closes: #598
Approved by: jlebon

Gbp-Pq: Name ASAN-cmdline-Fix-minor-leak-in-delta-cmdline-entrypoint.patch

deltas: Fix minor memory leak

We were leaking the checksum, ensure we free it in both normal and
error paths.

Closes: #598
Approved by: jlebon

Gbp-Pq: Name ASAN-deltas-Fix-minor-memory-leak.patch

delta compilation: More leak fixes

Now that I remembered to do `env G_SLICE=always-malloc`, lots more
leaks become apparent.  Nothing major.

Closes: #598
Approved by: jlebon

Gbp-Pq: Name ASAN-delta-compilation-More-leak-fixes.patch

delta compilation: Fix leak

We need to ref-sik the new varian for g_autoptr to work

Closes: #597
Approved by: cgwalters
Origin: upstream, 2016.15, commit:d57036f

Gbp-Pq: Name delta-compilation-Fix-leak.patch

pull: Don't leak delta superblock variants

Closes: #596
Approved by: cgwalters
Origin: upstream, 2016.15, commit:36f7824

Gbp-Pq: Name pull-Don-t-leak-delta-superblock-variants.patch

pull_with_options: Don't leak csum_v

Closes: #596
Approved by: cgwalters
Origin: upstream, 2016.15, commit:239cb14

Gbp-Pq: Name pull_with_options-Don-t-leak-csum_v.patch

ostree-repo-traverse: Don't leak floating GVariant

ostree_object_name_serialize returns a floating ref, so we need
to sink it before putting in the hashtable.

Closes: #595
Approved by: cgwalters
Origin: upstream, 2016.15, commit:fc107e5

Gbp-Pq: Name ostree-repo-traverse-Don-t-leak-floating-GVariant.patch

pull: scan_commit_object() - don't load variant twice

ostree_repo_load_commit already loaded the object, no need
to load it twice.

Closes: #595
Approved by: cgwalters
Origin: upstream, 2016.15, commit:c9b158e

Gbp-Pq: Name pull-scan_commit_object-don-t-load-variant-twice.patch

ostree-repo-traverse: Remove an accidental print statement

Closes: #594
Approved by: jlebon
Origin: upstream, 2016.15, commit:70d140d

Gbp-Pq: Name ostree-repo-traverse-Remove-an-accidental-print-statement.patch

ostree (2016.14-2) unstable; urgency=medium

  * Make build-time test failures non-fatal, as long as at least
    3 out of 5 attempts succeed.

    There are several upstream bugs that cause intermittent test
    failures, and can intermittently be reproduced in real use.
    However, these are not regressions, so we should not FTBFS just
    because we happen to have been unlucky during build.

  * d/p/Terminate-individual-tests-after-10-minutes.patch:
    replace d/p/debian/Terminate-individual-tests-after-half-an-hour.patch
    with the version that I sent upstream, which uses SIGABRT and
    terminates the tests sooner
  * d/p/*.patch: import more memory leak fixes from upstream

[dgit import unpatched ostree 2016.14-2]

Import ostree_2016.14-2.debian.tar.xz

[dgit import tarball ostree 2016.14-2 ostree_2016.14-2.debian.tar.xz]

Import ostree_2016.14.orig.tar.xz

[dgit import orig ostree_2016.14.orig.tar.xz]