Merge pull request #1596 from ThomasWaldmann/merge-1.0-maint

Merge 1.0 maint

README.rst

@@ -92,7 +92,6 @@ Initialize a new backup repository and create a backup archive::
 Now doing another backup, just to show off the great deduplication:
 
 .. code-block:: none
-   :emphasize-lines: 11
 
     $ borg create -v --stats /path/to/repo::Saturday2 ~/Documents
     -----------------------------------------------------------------------------
@@ -114,6 +113,22 @@ Now doing another backup, just to show off the great deduplication:
 
 For a graphical frontend refer to our complementary project `BorgWeb <https://borgweb.readthedocs.io/>`_.
 
+Checking Release Authenticity and Security Contact
+==================================================
+
+`Releases <https://github.com/borgbackup/borg/releases>`_ are signed with this GPG key,
+please use GPG to verify their authenticity.
+
+In case you discover a security issue, please use this contact for reporting it privately
+and please, if possible, use encrypted E-Mail:
+
+Thomas Waldmann <tw@waldmann-edv.de>
+
+GPG Key Fingerprint: 6D5B EF9A DD20 7580 5747  B70F 9F88 FB52 FAF7 B393
+
+The public key can be fetched from any GPG keyserver, but be careful: you must
+use the **full fingerprint** to check that you got the correct key.
+
 Links
 =====
 
@@ -169,7 +184,7 @@ THIS IS SOFTWARE IN DEVELOPMENT, DECIDE YOURSELF WHETHER IT FITS YOUR NEEDS.
 
 Borg is distributed under a 3-clause BSD license, see `License`_ for the complete license.
 
-|doc| |build| |coverage|
+|doc| |build| |coverage| |bestpractices|
 
 .. |doc| image:: https://readthedocs.org/projects/borgbackup/badge/?version=stable
         :alt: Documentation
@@ -186,3 +201,7 @@ Borg is distributed under a 3-clause BSD license, see `License`_ for the complet
 .. |screencast| image:: https://asciinema.org/a/28691.png
         :alt: BorgBackup Installation and Basic Usage
         :target: https://asciinema.org/a/28691?autoplay=1&speed=2
+
+.. |bestpractices| image:: https://bestpractices.coreinfrastructure.org/projects/271/badge
+        :alt: Best Practices Score
+        :target: https://bestpractices.coreinfrastructure.org/projects/271

Vagrantfile

@@ -387,7 +387,7 @@ Vagrant.configure(2) do |config|
   end
 
   config.vm.define "wheezy32" do |b|
-    b.vm.box = "boxcutter/debian79-i386"
+    b.vm.box = "boxcutter/debian711-i386"
     b.vm.provision "packages prepare wheezy", :type => :shell, :inline => packages_prepare_wheezy
     b.vm.provision "packages debianoid", :type => :shell, :inline => packages_debianoid
     b.vm.provision "install pyenv", :type => :shell, :privileged => false, :inline => install_pyenv("wheezy32")
@@ -400,7 +400,7 @@ Vagrant.configure(2) do |config|
   end
 
   config.vm.define "wheezy64" do |b|
-    b.vm.box = "boxcutter/debian79"
+    b.vm.box = "boxcutter/debian711"
     b.vm.provision "packages prepare wheezy", :type => :shell, :inline => packages_prepare_wheezy
     b.vm.provision "packages debianoid", :type => :shell, :inline => packages_debianoid
     b.vm.provision "install pyenv", :type => :shell, :privileged => false, :inline => install_pyenv("wheezy64")

docs/development.rst

@@ -10,6 +10,45 @@ This chapter will get you started with |project_name| development.
 |project_name| is written in Python (with a little bit of Cython and C for
 the performance critical parts).
 
+Contributions
+-------------
+
+... are welcome!
+
+Some guidance for contributors:
+
+- discuss about changes on github issue tracker, IRC or mailing list
+
+- choose the branch you base your changesets on wisely:
+
+  - choose x.y-maint for stuff that should go into next x.y release
+    (it usually gets merged into master branch later also)
+  - choose master if that does not apply
+
+- do clean changesets:
+
+  - focus on some topic, resist changing anything else.
+  - do not do style changes mixed with functional changes.
+  - try to avoid refactorings mixed with functional changes.
+  - if you need to fix something after commit/push:
+
+    - if there are ongoing reviews: do a fixup commit you can
+      merge into the bad commit later.
+    - if there are no ongoing reviews or you did not push the
+      bad commit yet: edit the commit to include your fix or
+      merge the fixup commit before pushing.
+  - have a nice, clear, typo-free commit comment
+  - if you fixed an issue, refer to it in your commit comment
+  - follow the style guide (see below)
+
+- if you write new code, please add tests and docs for it
+
+- run the tests, fix anything that comes up
+
+- make a pull request on github
+
+- wait for review by other developers
+
 Code and issues
 ---------------
 

docs/faq.rst

@@ -352,6 +352,8 @@ those files are reported as being added when, really, chunks are
 already used.
 
 
+.. _always_chunking:
+
 It always chunks all my files, even unchanged ones!
 ---------------------------------------------------
 

docs/usage.rst

@@ -86,6 +86,7 @@ General:
     BORG_FILES_CACHE_TTL
         When set to a numeric value, this determines the maximum "time to live" for the files cache
         entries (default: 20). The files cache is used to quickly determine whether a file is unchanged.
+        The FAQ explains this more detailled in: :ref:`always_chunking`
     TMPDIR
         where temporary files are stored (might need a lot of temporary space for some operations)
 

src/borg/archive.py

@@ -732,7 +732,8 @@ Number of files: {0.stats.nfiles}'''.format(
             return 'b'  # block device
 
     def process_symlink(self, path, st):
-        source = os.readlink(path)
+        with backup_io():
+            source = os.readlink(path)
         item = Item(path=make_path_safe(path), source=source)
         item.update(self.stat_attrs(st, path))
         self.add_item(item)
@@ -1009,8 +1010,9 @@ class ArchiveChecker:
         """Fetch a list of all object keys from repository
         """
         # Explicitly set the initial hash table capacity to avoid performance issues
-        # due to hash table "resonance"
+        # due to hash table "resonance".
-        capacity = int(len(self.repository) * 1.35 + 1)  # > len * 1.0 / HASH_MAX_LOAD (see _hashindex.c)
+        # Since reconstruction of archive items can add some new chunks, add 10 % headroom
+        capacity = int(len(self.repository) / ChunkIndex.MAX_LOAD_FACTOR * 1.1)
         self.chunks = ChunkIndex(capacity)
         marker = None
         while True:

src/borg/archiver.py

@@ -379,8 +379,13 @@ class Archiver:
                 if not read_special:
                     status = archive.process_symlink(path, st)
                 else:
-                    st_target = os.stat(path)
+                    try:
-                    if is_special(st_target.st_mode):
+                        st_target = os.stat(path)
+                    except OSError:
+                        special = False
+                    else:
+                        special = is_special(st_target.st_mode)
+                    if special:
                         status = archive.process_file(path, st_target, cache)
                     else:
                         status = archive.process_symlink(path, st)
@@ -984,7 +989,8 @@ class Archiver:
                                      dry_run=args.dry_run)
 
         with signal_handler(signal.SIGTERM, interrupt), \
-             signal_handler(signal.SIGINT, interrupt):
+             signal_handler(signal.SIGINT, interrupt), \
+             signal_handler(signal.SIGHUP, interrupt):
             if args.location.archive:
                 name = args.location.archive
                 if recreater.is_temporary_archive(name):
@@ -1865,11 +1871,14 @@ class Archiver:
         info_epilog = textwrap.dedent("""
         This command displays detailed information about the specified archive or repository.
 
-        The "This archive" line refers exclusively to the given archive:
+        Please note that the deduplicated sizes of the individual archives do not add
-        "Deduplicated size" is the size of the unique chunks stored only for the
+        up to the deduplicated size of the repository ("all archives"), because the two
-        given archive.
+        are meaning different things:
 
-        The "All archives" line shows global statistics (all chunks).
+        This archive / deduplicated size = amount of data stored ONLY for this archive
+                                         = unique chunks of this archive.
+        All archives / deduplicated size = amount of data stored in the repo
+                                         = all chunks in the repository.
         """)
         subparser = subparsers.add_parser('info', parents=[common_parser], add_help=False,
                                           description=self.do_info.__doc__,
@@ -2375,6 +2384,14 @@ def sig_term_handler(signum, stack):
     raise SIGTERMReceived
 
 
+class SIGHUPReceived(BaseException):
+    pass
+
+
+def sig_hup_handler(signum, stack):
+    raise SIGHUPReceived
+
+
 def setup_signal_handlers():  # pragma: no cover
     sigs = []
     if hasattr(signal, 'SIGUSR1'):
@@ -2383,7 +2400,12 @@ def setup_signal_handlers():  # pragma: no cover
         sigs.append(signal.SIGINFO)  # kill -INFO pid (or ctrl-t)
     for sig in sigs:
         signal.signal(sig, sig_info_handler)
+    # If we received SIGTERM or SIGHUP, catch them and raise a proper exception
+    # that can be handled for an orderly exit. SIGHUP is important especially
+    # for systemd systems, where logind sends it when a session exits, in
+    # addition to any traditional use.
     signal.signal(signal.SIGTERM, sig_term_handler)
+    signal.signal(signal.SIGHUP, sig_hup_handler)
 
 
 def main():  # pragma: no cover
@@ -2438,6 +2460,9 @@ def main():  # pragma: no cover
         tb_log_level = logging.DEBUG
         tb = '%s\n%s' % (traceback.format_exc(), sysinfo())
         exit_code = EXIT_ERROR
+    except SIGHUPReceived:
+        msg = 'Received SIGHUP.'
+        exit_code = EXIT_ERROR
     if msg:
         logger.error(msg)
     if tb:

src/borg/hashindex.pyx

@@ -28,6 +28,8 @@ cdef extern from "_hashindex.c":
     uint32_t _htole32(uint32_t v)
     uint32_t _le32toh(uint32_t v)
 
+    double HASH_MAX_LOAD
+
 
 cdef _NoDefault = object()
 
@@ -50,7 +52,6 @@ assert UINT32_MAX == 2**32-1
 
 # module-level constant because cdef's in classes can't have default values
 cdef uint32_t _MAX_VALUE = 2**32-1025
-MAX_VALUE = _MAX_VALUE
 
 assert _MAX_VALUE % 2 == 1
 
@@ -60,6 +61,9 @@ cdef class IndexBase:
     cdef HashIndex *index
     cdef int key_size
 
+    MAX_LOAD_FACTOR = HASH_MAX_LOAD
+    MAX_VALUE = _MAX_VALUE
+
     def __cinit__(self, capacity=0, path=None, key_size=32):
         self.key_size = key_size
         if path:
@@ -296,7 +300,7 @@ cdef class ChunkIndex(IndexBase):
             unique_chunks += 1
             values = <uint32_t*> (key + self.key_size)
             refcount = _le32toh(values[0])
-            assert refcount <= MAX_VALUE, "invalid reference count"
+            assert refcount <= _MAX_VALUE, "invalid reference count"
             chunks += refcount
             unique_size += _le32toh(values[1])
             unique_csize += _le32toh(values[2])
@@ -358,5 +362,5 @@ cdef class ChunkKeyIterator:
             raise StopIteration
         cdef uint32_t *value = <uint32_t *>(self.key + self.key_size)
         cdef uint32_t refcount = _le32toh(value[0])
-        assert refcount <= MAX_VALUE, "invalid reference count"
+        assert refcount <= _MAX_VALUE, "invalid reference count"
         return (<char *>self.key)[:self.key_size], ChunkIndexEntry(refcount, _le32toh(value[1]), _le32toh(value[2]))

src/borg/testsuite/archiver.py

@@ -1130,6 +1130,14 @@ class ArchiverTestCase(ArchiverTestCaseBase):
         output = self.cmd('create', '--list', '--filter=AM', self.repository_location + '::test3', 'input')
         self.assert_in('file1', output)
 
+    def test_create_read_special_broken_symlink(self):
+        os.symlink('somewhere doesnt exist', os.path.join(self.input_path, 'link'))
+        self.cmd('init', self.repository_location)
+        archive = self.repository_location + '::test'
+        self.cmd('create', '--read-special', archive, 'input')
+        output = self.cmd('list', archive)
+        assert 'input/link -> somewhere doesnt exist' in output
+
     # def test_cmdline_compatibility(self):
     #    self.create_regular_file('file1', size=1024 * 80)
     #    self.cmd('init', self.repository_location)

src/borg/testsuite/hashindex.py

@@ -140,16 +140,16 @@ class HashIndexSizeTestCase(BaseTestCase):
 class HashIndexRefcountingTestCase(BaseTestCase):
     def test_chunkindex_limit(self):
         idx = ChunkIndex()
-        idx[H(1)] = hashindex.MAX_VALUE - 1, 1, 2
+        idx[H(1)] = ChunkIndex.MAX_VALUE - 1, 1, 2
 
         # 5 is arbitray, any number of incref/decrefs shouldn't move it once it's limited
         for i in range(5):
             # first incref to move it to the limit
             refcount, *_ = idx.incref(H(1))
-            assert refcount == hashindex.MAX_VALUE
+            assert refcount == ChunkIndex.MAX_VALUE
         for i in range(5):
             refcount, *_ = idx.decref(H(1))
-            assert refcount == hashindex.MAX_VALUE
+            assert refcount == ChunkIndex.MAX_VALUE
 
     def _merge(self, refcounta, refcountb):
         def merge(refcount1, refcount2):
@@ -168,23 +168,23 @@ class HashIndexRefcountingTestCase(BaseTestCase):
     def test_chunkindex_merge_limit1(self):
         # Check that it does *not* limit at MAX_VALUE - 1
         # (MAX_VALUE is odd)
-        half = hashindex.MAX_VALUE // 2
+        half = ChunkIndex.MAX_VALUE // 2
-        assert self._merge(half, half) == hashindex.MAX_VALUE - 1
+        assert self._merge(half, half) == ChunkIndex.MAX_VALUE - 1
 
     def test_chunkindex_merge_limit2(self):
         # 3000000000 + 2000000000 > MAX_VALUE
-        assert self._merge(3000000000, 2000000000) == hashindex.MAX_VALUE
+        assert self._merge(3000000000, 2000000000) == ChunkIndex.MAX_VALUE
 
     def test_chunkindex_merge_limit3(self):
         # Crossover point: both addition and limit semantics will yield the same result
-        half = hashindex.MAX_VALUE // 2
+        half = ChunkIndex.MAX_VALUE // 2
-        assert self._merge(half + 1, half) == hashindex.MAX_VALUE
+        assert self._merge(half + 1, half) == ChunkIndex.MAX_VALUE
 
     def test_chunkindex_merge_limit4(self):
         # Beyond crossover, result of addition would be 2**31
-        half = hashindex.MAX_VALUE // 2
+        half = ChunkIndex.MAX_VALUE // 2
-        assert self._merge(half + 2, half) == hashindex.MAX_VALUE
+        assert self._merge(half + 2, half) == ChunkIndex.MAX_VALUE
-        assert self._merge(half + 1, half + 1) == hashindex.MAX_VALUE
+        assert self._merge(half + 1, half + 1) == ChunkIndex.MAX_VALUE
 
     def test_chunkindex_add(self):
         idx1 = ChunkIndex()
@@ -195,17 +195,17 @@ class HashIndexRefcountingTestCase(BaseTestCase):
 
     def test_incref_limit(self):
         idx1 = ChunkIndex()
-        idx1[H(1)] = (hashindex.MAX_VALUE, 6, 7)
+        idx1[H(1)] = (ChunkIndex.MAX_VALUE, 6, 7)
         idx1.incref(H(1))
         refcount, *_ = idx1[H(1)]
-        assert refcount == hashindex.MAX_VALUE
+        assert refcount == ChunkIndex.MAX_VALUE
 
     def test_decref_limit(self):
         idx1 = ChunkIndex()
-        idx1[H(1)] = hashindex.MAX_VALUE, 6, 7
+        idx1[H(1)] = ChunkIndex.MAX_VALUE, 6, 7
         idx1.decref(H(1))
         refcount, *_ = idx1[H(1)]
-        assert refcount == hashindex.MAX_VALUE
+        assert refcount == ChunkIndex.MAX_VALUE
 
     def test_decref_zero(self):
         idx1 = ChunkIndex()
@@ -225,7 +225,7 @@ class HashIndexRefcountingTestCase(BaseTestCase):
     def test_setitem_raises(self):
         idx1 = ChunkIndex()
         with self.assert_raises(AssertionError):
-            idx1[H(1)] = hashindex.MAX_VALUE + 1, 0, 0
+            idx1[H(1)] = ChunkIndex.MAX_VALUE + 1, 0, 0
 
     def test_keyerror(self):
         idx = ChunkIndex()
@@ -282,14 +282,20 @@ class HashIndexDataTestCase(BaseTestCase):
         idx2 = ChunkIndex()
         idx2[H(3)] = 2**32 - 123456, 6, 7
         idx1.merge(idx2)
-        assert idx1[H(3)] == (hashindex.MAX_VALUE, 6, 7)
+        assert idx1[H(3)] == (ChunkIndex.MAX_VALUE, 6, 7)
 
 
 class NSIndexTestCase(BaseTestCase):
     def test_nsindex_segment_limit(self):
         idx = NSIndex()
         with self.assert_raises(AssertionError):
-            idx[H(1)] = hashindex.MAX_VALUE + 1, 0
+            idx[H(1)] = NSIndex.MAX_VALUE + 1, 0
         assert H(1) not in idx
-        idx[H(2)] = hashindex.MAX_VALUE, 0
+        idx[H(2)] = NSIndex.MAX_VALUE, 0
         assert H(2) in idx
+
+
+class AllIndexTestCase(BaseTestCase):
+    def test_max_load_factor(self):
+        assert NSIndex.MAX_LOAD_FACTOR < 1.0
+        assert ChunkIndex.MAX_LOAD_FACTOR < 1.0