Drop vestigial compat.items

[bup.git] / lib / bup / git.py

diff --git a/lib/bup/git.py b/lib/bup/git.py
index 691a5552dd66ae776b988e64d639f1a7a3ee7d3c..5d46cc7b6099245ae8a527227328f0c8c0913b6d 100644 (file)
--- a/lib/bup/git.py
+++ b/lib/bup/git.py
@@ -4,40 +4,39 @@ interact with the Git data structures.
 """
 
 from __future__ import absolute_import, print_function
 """
 
 from __future__ import absolute_import, print_function

-import errno, os, sys, zlib, time, subprocess, struct, stat, re, tempfile, glob
+import os, sys, zlib, subprocess, struct, stat, re, tempfile, glob

 from array import array
 from binascii import hexlify, unhexlify
 from collections import namedtuple
 from itertools import islice
 from array import array
 from binascii import hexlify, unhexlify
 from collections import namedtuple
 from itertools import islice

-from numbers import Integral

 
 

-from bup import _helpers, compat, hashsplit, path, midx, bloom, xstat
+from bup import _helpers, hashsplit, path, midx, bloom, xstat

 from bup.compat import (buffer,
                         byte_int, bytes_from_byte, bytes_from_uint,
                         environ,
 from bup.compat import (buffer,
                         byte_int, bytes_from_byte, bytes_from_uint,
                         environ,

-                        items,
-                        range,
+                        ExitStack,
+                        pending_raise,

                         reraise)
 from bup.io import path_msg
 from bup.helpers import (Sha1, add_error, chunkyreader, debug1, debug2,
                          exo,
                          fdatasync,
                         reraise)
 from bup.io import path_msg
 from bup.helpers import (Sha1, add_error, chunkyreader, debug1, debug2,
                          exo,
                          fdatasync,

-                         hostname, localtime, log,
+                         finalized,
+                         log,

                          merge_dict,
                          merge_iter,
                          mmap_read, mmap_readwrite,
                          merge_dict,
                          merge_iter,
                          mmap_read, mmap_readwrite,

-                         parse_num,
+                         nullcontext_if_not,

                          progress, qprogress, stat_if_exists,
                          unlink,
                          utc_offset_str)
                          progress, qprogress, stat_if_exists,
                          unlink,
                          utc_offset_str)

-from bup.pwdgrp import username, userfullname

 
 
 verbose = 0
 repodir = None  # The default repository, once initialized
 
 _typemap =  {b'blob': 3, b'tree': 2, b'commit': 1, b'tag': 4}
 
 
 verbose = 0
 repodir = None  # The default repository, once initialized
 
 _typemap =  {b'blob': 3, b'tree': 2, b'commit': 1, b'tag': 4}

-_typermap = {v: k for k, v in items(_typemap)}
+_typermap = {v: k for k, v in _typemap.items()}

 
 
 _total_searches = 0
 
 
 _total_searches = 0


@@ -66,14 +65,32 @@ def _git_exo(cmd, **kwargs):
         raise GitError('%r returned %d' % (cmd, proc.returncode))
     return result
 
         raise GitError('%r returned %d' % (cmd, proc.returncode))
     return result
 

-def git_config_get(option, repo_dir=None):
-    cmd = (b'git', b'config', b'--get', option)
-    p = subprocess.Popen(cmd, stdout=subprocess.PIPE,
-                         env=_gitenv(repo_dir=repo_dir),
+def git_config_get(option, repo_dir=None, opttype=None, cfg_file=None):
+    assert not (repo_dir and cfg_file), "repo_dir and cfg_file cannot both be used"
+    cmd = [b'git', b'config', b'--null']
+    if cfg_file:
+        cmd.extend([b'--file', cfg_file])
+    if opttype == 'int':
+        cmd.extend([b'--int'])
+    elif opttype == 'bool':
+        cmd.extend([b'--bool'])
+    else:
+        assert opttype is None
+    cmd.extend([b'--get', option])
+    env=None
+    if repo_dir:
+        env = _gitenv(repo_dir=repo_dir)
+    p = subprocess.Popen(cmd, stdout=subprocess.PIPE, env=env,

                          close_fds=True)
                          close_fds=True)

-    r = p.stdout.read()
+    # with --null, git writes out a trailing \0 after the value
+    r = p.stdout.read()[:-1]

     rc = p.wait()
     if rc == 0:
     rc = p.wait()
     if rc == 0:

+        if opttype == 'int':
+            return int(r)
+        elif opttype == 'bool':
+            # git converts to 'true' or 'false'
+            return r == b'true'

         return r
     if rc != 1:
         raise GitError('%r returned %d' % (cmd, rc))
         return r
     if rc != 1:
         raise GitError('%r returned %d' % (cmd, rc))


@@ -87,9 +104,26 @@ def parse_tz_offset(s):
         return - tz_off
     return tz_off
 
         return - tz_off
     return tz_off
 

+def parse_commit_gpgsig(sig):
+    """Return the original signature bytes.
+
+    i.e. with the "gpgsig " header and the leading space character on
+    each continuation line removed.
+
+    """
+    if not sig:
+        return None
+    assert sig.startswith(b'gpgsig ')
+    sig = sig[7:]
+    return sig.replace(b'\n ', b'\n')

 
 # FIXME: derived from http://git.rsbx.net/Documents/Git_Data_Formats.txt
 # Make sure that's authoritative.
 
 # FIXME: derived from http://git.rsbx.net/Documents/Git_Data_Formats.txt
 # Make sure that's authoritative.

+
+# See also
+# https://github.com/git/git/blob/master/Documentation/technical/signature-format.txt
+# The continuation lines have only one leading space.
+

 _start_end_char = br'[^ .,:;<>"\'\0\n]'
 _content_char = br'[^\0\n<>]'
 _safe_str_rx = br'(?:%s{1,2}|(?:%s%s*%s))' \
 _start_end_char = br'[^ .,:;<>"\'\0\n]'
 _content_char = br'[^\0\n<>]'
 _safe_str_rx = br'(?:%s{1,2}|(?:%s%s*%s))' \


@@ -103,7 +137,7 @@ _mergetag_rx = br'(?:\nmergetag object [abcdefABCDEF0123456789]{40}(?:\n [^\0\n]
 _commit_rx = re.compile(br'''tree (?P<tree>[abcdefABCDEF0123456789]{40})
 (?P<parents>%s*)author (?P<author_name>%s) <(?P<author_mail>%s)> (?P<asec>\d+) (?P<atz>%s)
 committer (?P<committer_name>%s) <(?P<committer_mail>%s)> (?P<csec>\d+) (?P<ctz>%s)(?P<mergetag>%s?)
 _commit_rx = re.compile(br'''tree (?P<tree>[abcdefABCDEF0123456789]{40})
 (?P<parents>%s*)author (?P<author_name>%s) <(?P<author_mail>%s)> (?P<asec>\d+) (?P<atz>%s)
 committer (?P<committer_name>%s) <(?P<committer_mail>%s)> (?P<csec>\d+) (?P<ctz>%s)(?P<mergetag>%s?)

-
+(?P<gpgsig>gpgsig .*\n(?: .*\n)*)?

 (?P<message>(?:.|\n)*)''' % (_parent_rx,
                              _safe_str_rx, _safe_str_rx, _tz_rx,
                              _safe_str_rx, _safe_str_rx, _tz_rx,
 (?P<message>(?:.|\n)*)''' % (_parent_rx,
                              _safe_str_rx, _safe_str_rx, _tz_rx,
                              _safe_str_rx, _safe_str_rx, _tz_rx,


@@ -117,6 +151,7 @@ CommitInfo = namedtuple('CommitInfo', ['tree', 'parents',
                                        'author_sec', 'author_offset',
                                        'committer_name', 'committer_mail',
                                        'committer_sec', 'committer_offset',
                                        'author_sec', 'author_offset',
                                        'committer_name', 'committer_mail',
                                        'committer_sec', 'committer_offset',

+                                       'gpgsig',

                                        'message'])
 
 def parse_commit(content):
                                        'message'])
 
 def parse_commit(content):


@@ -134,6 +169,7 @@ def parse_commit(content):
                       committer_mail=matches['committer_mail'],
                       committer_sec=int(matches['csec']),
                       committer_offset=parse_tz_offset(matches['ctz']),
                       committer_mail=matches['committer_mail'],
                       committer_sec=int(matches['csec']),
                       committer_offset=parse_tz_offset(matches['ctz']),

+                      gpgsig=parse_commit_gpgsig(matches['gpgsig']),

                       message=matches['message'])
 
 
                       message=matches['message'])
 
 


@@ -192,12 +228,6 @@ def repo_rel(path):
     return shorten_hash(path)
 
 
     return shorten_hash(path)
 
 

-def all_packdirs():
-    paths = [repo(b'objects/pack')]
-    paths += glob.glob(repo(b'index-cache/*/.'))
-    return paths
-
-

 def auto_midx(objdir):
     args = [path.exe(), b'midx', b'--auto', b'--dir', objdir]
     try:
 def auto_midx(objdir):
     args = [path.exe(), b'midx', b'--auto', b'--dir', objdir]
     try:


@@ -254,8 +284,7 @@ def demangle_name(name, mode):
     elif name.endswith(b'.bupm'):
         return (name[:-5],
                 BUP_CHUNKED if stat.S_ISDIR(mode) else BUP_NORMAL)
     elif name.endswith(b'.bupm'):
         return (name[:-5],
                 BUP_CHUNKED if stat.S_ISDIR(mode) else BUP_NORMAL)

-    else:
-        return (name, BUP_NORMAL)
+    return (name, BUP_NORMAL)

 
 
 def calc_hash(type, content):
 
 
 def calc_hash(type, content):


@@ -324,27 +353,6 @@ def _encode_packobj(type, content, compression_level=1):
     yield z.flush()
 
 
     yield z.flush()
 
 

-def _encode_looseobj(type, content, compression_level=1):
-    z = zlib.compressobj(compression_level)
-    yield z.compress(b'%s %d\0' % (type, len(content)))
-    yield z.compress(content)
-    yield z.flush()
-
-
-def _decode_looseobj(buf):
-    assert(buf);
-    s = zlib.decompress(buf)
-    i = s.find(b'\0')
-    assert(i > 0)
-    l = s[:i].split(b' ')
-    type = l[0]
-    sz = int(l[1])
-    content = s[i+1:]
-    assert(type in _typemap)
-    assert(sz == len(content))
-    return (type, content)
-
-

 def _decode_packobj(buf):
     assert(buf)
     c = byte_int(buf[0])
 def _decode_packobj(buf):
     assert(buf)
     c = byte_int(buf[0])


@@ -362,10 +370,7 @@ def _decode_packobj(buf):
     return (type, zlib.decompress(buf[i+1:]))
 
 
     return (type, zlib.decompress(buf[i+1:]))
 
 

-class PackIdx:
-    def __init__(self):
-        assert(0)
-
+class PackIdx(object):

     def find_offset(self, hash):
         """Get the offset of an object inside the index file."""
         idx = self._idx_from_hash(hash)
     def find_offset(self, hash):
         """Get the offset of an object inside the index file."""
         idx = self._idx_from_hash(hash)


@@ -404,6 +409,8 @@ class PackIdx:
 class PackIdxV1(PackIdx):
     """Object representation of a Git pack index (version 1) file."""
     def __init__(self, filename, f):
 class PackIdxV1(PackIdx):
     """Object representation of a Git pack index (version 1) file."""
     def __init__(self, filename, f):

+        super(PackIdxV1, self).__init__()
+        self.closed = False

         self.name = filename
         self.idxnames = [self.name]
         self.map = mmap_read(f)
         self.name = filename
         self.idxnames = [self.name]
         self.map = mmap_read(f)


@@ -419,7 +426,8 @@ class PackIdxV1(PackIdx):
         return self
 
     def __exit__(self, type, value, traceback):
         return self
 
     def __exit__(self, type, value, traceback):

-        self.close()
+        with pending_raise(value, rethrow=False):
+            self.close()

 
     def __len__(self):
         return int(self.nsha)  # int() from long for python 2
 
     def __len__(self):
         return int(self.nsha)  # int() from long for python 2


@@ -442,15 +450,21 @@ class PackIdxV1(PackIdx):
             yield self.map[ofs : ofs + 20]
 
     def close(self):
             yield self.map[ofs : ofs + 20]
 
     def close(self):

+        self.closed = True

         if self.map is not None:
             self.shatable = None
             self.map.close()
             self.map = None
 
         if self.map is not None:
             self.shatable = None
             self.map.close()
             self.map = None
 

+    def __del__(self):
+        assert self.closed
+

 
 class PackIdxV2(PackIdx):
     """Object representation of a Git pack index (version 2) file."""
     def __init__(self, filename, f):
 
 class PackIdxV2(PackIdx):
     """Object representation of a Git pack index (version 2) file."""
     def __init__(self, filename, f):

+        super(PackIdxV2, self).__init__()
+        self.closed = False

         self.name = filename
         self.idxnames = [self.name]
         self.map = mmap_read(f)
         self.name = filename
         self.idxnames = [self.name]
         self.map = mmap_read(f)


@@ -469,7 +483,8 @@ class PackIdxV2(PackIdx):
         return self
 
     def __exit__(self, type, value, traceback):
         return self
 
     def __exit__(self, type, value, traceback):

-        self.close()
+        with pending_raise(value, rethrow=False):
+            self.close()

 
     def __len__(self):
         return int(self.nsha)  # int() from long for python 2
 
     def __len__(self):
         return int(self.nsha)  # int() from long for python 2


@@ -497,30 +512,62 @@ class PackIdxV2(PackIdx):
             yield self.map[ofs : ofs + 20]
 
     def close(self):
             yield self.map[ofs : ofs + 20]
 
     def close(self):

+        self.closed = True

         if self.map is not None:
             self.shatable = None
             self.map.close()
             self.map = None
 
         if self.map is not None:
             self.shatable = None
             self.map.close()
             self.map = None
 

+    def __del__(self):
+        assert self.closed
+

 
 _mpi_count = 0
 class PackIdxList:
     def __init__(self, dir, ignore_midx=False):
         global _mpi_count
 
 _mpi_count = 0
 class PackIdxList:
     def __init__(self, dir, ignore_midx=False):
         global _mpi_count

+        # Q: was this also intended to prevent opening multiple repos?

         assert(_mpi_count == 0) # these things suck tons of VM; don't waste it
         _mpi_count += 1
         assert(_mpi_count == 0) # these things suck tons of VM; don't waste it
         _mpi_count += 1

+        self.open = True

         self.dir = dir
         self.also = set()
         self.packs = []
         self.do_bloom = False
         self.bloom = None
         self.ignore_midx = ignore_midx
         self.dir = dir
         self.also = set()
         self.packs = []
         self.do_bloom = False
         self.bloom = None
         self.ignore_midx = ignore_midx

-        self.refresh()
+        try:
+            self.refresh()
+        except BaseException as ex:
+            with pending_raise(ex):
+                self.close()

 
 

-    def __del__(self):
+    def close(self):

         global _mpi_count
         global _mpi_count

+        if not self.open:
+            assert _mpi_count == 0
+            return

         _mpi_count -= 1
         _mpi_count -= 1

-        assert(_mpi_count == 0)
+        assert _mpi_count == 0
+        self.also = None
+        self.bloom, bloom = None, self.bloom
+        self.packs, packs = None, self.packs
+        self.open = False
+        with ExitStack() as stack:
+            for pack in packs:
+                stack.enter_context(pack)
+            if bloom:
+                bloom.close()
+
+    def __enter__(self):
+        return self
+
+    def __exit__(self, type, value, traceback):
+        with pending_raise(value, rethrow=False):
+            self.close()
+
+    def __del__(self):
+        assert not self.open

 
     def __iter__(self):
         return iter(idxmerge(self.packs))
 
     def __iter__(self):
         return iter(idxmerge(self.packs))


@@ -601,7 +648,6 @@ class PackIdxList:
                                 broken = True
                         if broken:
                             mx.close()
                                 broken = True
                         if broken:
                             mx.close()

-                            del mx

                             unlink(full)
                         else:
                             midxl.append(mx)
                             unlink(full)
                         else:
                             midxl.append(mx)


@@ -633,14 +679,27 @@ class PackIdxList:
                         continue
                     d[full] = ix
             bfull = os.path.join(self.dir, b'bup.bloom')
                         continue
                     d[full] = ix
             bfull = os.path.join(self.dir, b'bup.bloom')

+            new_packs = set(d.values())
+            for p in self.packs:
+                if not p in new_packs:
+                    p.close()
+            new_packs = list(new_packs)
+            new_packs.sort(reverse=True, key=lambda x: len(x))
+            self.packs = new_packs

             if self.bloom is None and os.path.exists(bfull):
                 self.bloom = bloom.ShaBloom(bfull)
             if self.bloom is None and os.path.exists(bfull):
                 self.bloom = bloom.ShaBloom(bfull)

-            self.packs = list(set(d.values()))
-            self.packs.sort(reverse=True, key=lambda x: len(x))
-            if self.bloom and self.bloom.valid() and len(self.bloom) >= len(self):
-                self.do_bloom = True
-            else:
-                self.bloom = None
+            try:
+                if self.bloom and self.bloom.valid() and len(self.bloom) >= len(self):
+                    self.do_bloom = True
+                else:
+                    if self.bloom:
+                        self.bloom, bloom_tmp = None, self.bloom
+                        bloom_tmp.close()
+            except BaseException as ex:
+                with pending_raise(ex):
+                    if self.bloom:
+                        self.bloom.close()
+

         debug1('PackIdxList: using %d index%s.\n'
             % (len(self.packs), len(self.packs)!=1 and 'es' or ''))
 
         debug1('PackIdxList: using %d index%s.\n'
             % (len(self.packs), len(self.packs)!=1 and 'es' or ''))
 


@@ -683,17 +742,39 @@ def idxmerge(idxlist, final_progress=True):
     return merge_iter(idxlist, 10024, pfunc, pfinal)
 
 
     return merge_iter(idxlist, 10024, pfunc, pfinal)
 
 

+def create_commit_blob(tree, parent,
+                       author, adate_sec, adate_tz,
+                       committer, cdate_sec, cdate_tz,
+                       msg):
+    if adate_tz is not None:
+        adate_str = _git_date_str(adate_sec, adate_tz)
+    else:
+        adate_str = _local_git_date_str(adate_sec)
+    if cdate_tz is not None:
+        cdate_str = _git_date_str(cdate_sec, cdate_tz)
+    else:
+        cdate_str = _local_git_date_str(cdate_sec)
+    l = []
+    if tree: l.append(b'tree %s' % hexlify(tree))
+    if parent: l.append(b'parent %s' % hexlify(parent))
+    if author: l.append(b'author %s %s' % (author, adate_str))
+    if committer: l.append(b'committer %s %s' % (committer, cdate_str))
+    l.append(b'')
+    l.append(msg)
+    return b'\n'.join(l)
+

 def _make_objcache():
     return PackIdxList(repo(b'objects/pack'))
 
 # bup-gc assumes that it can disable all PackWriter activities
 # (bloom/midx/cache) via the constructor and close() arguments.
 
 def _make_objcache():
     return PackIdxList(repo(b'objects/pack'))
 
 # bup-gc assumes that it can disable all PackWriter activities
 # (bloom/midx/cache) via the constructor and close() arguments.
 

-class PackWriter:
+class PackWriter(object):

     """Writes Git objects inside a pack file."""
     def __init__(self, objcache_maker=_make_objcache, compression_level=1,
                  run_midx=True, on_pack_finish=None,
                  max_pack_size=None, max_pack_objects=None, repo_dir=None):
     """Writes Git objects inside a pack file."""
     def __init__(self, objcache_maker=_make_objcache, compression_level=1,
                  run_midx=True, on_pack_finish=None,
                  max_pack_size=None, max_pack_objects=None, repo_dir=None):

+        self.closed = False

         self.repo_dir = repo_dir or repo()
         self.file = None
         self.parentfd = None
         self.repo_dir = repo_dir or repo()
         self.file = None
         self.parentfd = None


@@ -708,9 +789,8 @@ class PackWriter:
         self.on_pack_finish = on_pack_finish
         if not max_pack_size:
             max_pack_size = git_config_get(b'pack.packSizeLimit',
         self.on_pack_finish = on_pack_finish
         if not max_pack_size:
             max_pack_size = git_config_get(b'pack.packSizeLimit',

-                                           repo_dir=self.repo_dir)
-            if max_pack_size is not None:
-                max_pack_size = parse_num(max_pack_size)
+                                           repo_dir=self.repo_dir,
+                                           opttype='int')

             if not max_pack_size:
                 # larger packs slow down pruning
                 max_pack_size = 1000 * 1000 * 1000
             if not max_pack_size:
                 # larger packs slow down pruning
                 max_pack_size = 1000 * 1000 * 1000


@@ -719,14 +799,12 @@ class PackWriter:
         self.max_pack_objects = max_pack_objects if max_pack_objects \
                                 else max(1, self.max_pack_size // 5000)
 
         self.max_pack_objects = max_pack_objects if max_pack_objects \
                                 else max(1, self.max_pack_size // 5000)
 

-    def __del__(self):
-        self.close()
-

     def __enter__(self):
         return self
 
     def __exit__(self, type, value, traceback):
     def __enter__(self):
         return self
 
     def __exit__(self, type, value, traceback):

-        self.close()
+        with pending_raise(value, rethrow=False):
+            self.close()

 
     def _open(self):
         if not self.file:
 
     def _open(self):
         if not self.file:


@@ -787,12 +865,6 @@ class PackWriter:
             self.breakpoint()
         return sha
 
             self.breakpoint()
         return sha
 

-    def breakpoint(self):
-        """Clear byte and object counts and return the last processed id."""
-        id = self._end(self.run_midx)
-        self.outbytes = self.count = 0
-        return id
-

     def _require_objcache(self):
         if self.objcache is None and self.objcache_maker:
             self.objcache = self.objcache_maker()
     def _require_objcache(self):
         if self.objcache is None and self.objcache_maker:
             self.objcache = self.objcache_maker()


@@ -835,90 +907,81 @@ class PackWriter:
                    msg):
         """Create a commit object in the pack.  The date_sec values must be
         epoch-seconds, and if a tz is None, the local timezone is assumed."""
                    msg):
         """Create a commit object in the pack.  The date_sec values must be
         epoch-seconds, and if a tz is None, the local timezone is assumed."""

-        if adate_tz is not None:
-            adate_str = _git_date_str(adate_sec, adate_tz)
-        else:
-            adate_str = _local_git_date_str(adate_sec)
-        if cdate_tz is not None:
-            cdate_str = _git_date_str(cdate_sec, cdate_tz)
-        else:
-            cdate_str = _local_git_date_str(cdate_sec)
-        l = []
-        if tree: l.append(b'tree %s' % hexlify(tree))
-        if parent: l.append(b'parent %s' % hexlify(parent))
-        if author: l.append(b'author %s %s' % (author, adate_str))
-        if committer: l.append(b'committer %s %s' % (committer, cdate_str))
-        l.append(b'')
-        l.append(msg)
-        return self.maybe_write(b'commit', b'\n'.join(l))
+        content = create_commit_blob(tree, parent,
+                                     author, adate_sec, adate_tz,
+                                     committer, cdate_sec, cdate_tz,
+                                     msg)
+        return self.maybe_write(b'commit', content)
+
+    def _end(self, run_midx=True, abort=False):
+        # Ignores run_midx during abort
+        self.parentfd, pfd, = None, self.parentfd
+        self.file, f = None, self.file
+        self.idx, idx = None, self.idx
+        try:
+            with nullcontext_if_not(self.objcache), \
+                 finalized(pfd, lambda x: x is not None and os.close(x)), \
+                 nullcontext_if_not(f):
+                if not f:
+                    return None

 
 

-    def abort(self):
-        """Remove the pack file from disk."""
-        f = self.file
-        if f:
-            pfd = self.parentfd
-            self.file = None
-            self.parentfd = None
-            self.idx = None
-            try:
-                try:
+                if abort:

                     os.unlink(self.filename + b'.pack')
                     os.unlink(self.filename + b'.pack')

-                finally:
-                    f.close()
-            finally:
-                if pfd is not None:
-                    os.close(pfd)
+                    return None
+
+                # update object count
+                f.seek(8)
+                cp = struct.pack('!i', self.count)
+                assert len(cp) == 4
+                f.write(cp)
+
+                # calculate the pack sha1sum
+                f.seek(0)
+                sum = Sha1()
+                for b in chunkyreader(f):
+                    sum.update(b)
+                packbin = sum.digest()
+                f.write(packbin)
+                f.flush()
+                fdatasync(f.fileno())
+                f.close()

 
 

-    def _end(self, run_midx=True):
-        f = self.file
-        if not f: return None
-        self.file = None
-        try:
-            self.objcache = None
-            idx = self.idx
-            self.idx = None
-
-            # update object count
-            f.seek(8)
-            cp = struct.pack('!i', self.count)
-            assert(len(cp) == 4)
-            f.write(cp)
-
-            # calculate the pack sha1sum
-            f.seek(0)
-            sum = Sha1()
-            for b in chunkyreader(f):
-                sum.update(b)
-            packbin = sum.digest()
-            f.write(packbin)
-            fdatasync(f.fileno())
+                idx.write(self.filename + b'.idx', packbin)
+                nameprefix = os.path.join(self.repo_dir,
+                                          b'objects/pack/pack-' +  hexlify(packbin))
+                if os.path.exists(self.filename + b'.map'):
+                    os.unlink(self.filename + b'.map')
+                os.rename(self.filename + b'.pack', nameprefix + b'.pack')
+                os.rename(self.filename + b'.idx', nameprefix + b'.idx')
+                os.fsync(pfd)
+                if run_midx:
+                    auto_midx(os.path.join(self.repo_dir, b'objects/pack'))
+                if self.on_pack_finish:
+                    self.on_pack_finish(nameprefix)
+                return nameprefix

         finally:
         finally:

-            f.close()
-
-        obj_list_sha = idx.write(self.filename + b'.idx', packbin)
-        nameprefix = os.path.join(self.repo_dir,
-                                  b'objects/pack/pack-' +  obj_list_sha)
-        if os.path.exists(self.filename + b'.map'):
-            os.unlink(self.filename + b'.map')
-        os.rename(self.filename + b'.pack', nameprefix + b'.pack')
-        os.rename(self.filename + b'.idx', nameprefix + b'.idx')
-        try:
-            os.fsync(self.parentfd)
-        finally:
-            os.close(self.parentfd)
-
-        if run_midx:
-            auto_midx(os.path.join(self.repo_dir, b'objects/pack'))
+            # Must be last -- some of the code above depends on it
+            self.objcache = None

 
 

-        if self.on_pack_finish:
-            self.on_pack_finish(nameprefix)
+    def abort(self):
+        """Remove the pack file from disk."""
+        self.closed = True
+        self._end(abort=True)

 
 

-        return nameprefix
+    def breakpoint(self):
+        """Clear byte and object counts and return the last processed id."""
+        id = self._end(self.run_midx)
+        self.outbytes = self.count = 0
+        return id

 
     def close(self, run_midx=True):
         """Close the pack file and move it to its definitive path."""
 
     def close(self, run_midx=True):
         """Close the pack file and move it to its definitive path."""

+        self.closed = True

         return self._end(run_midx=run_midx)
 
         return self._end(run_midx=run_midx)
 

+    def __del__(self):
+        assert self.closed
+

 
 class PackIdxV2Writer:
     def __init__(self):
 
 class PackIdxV2Writer:
     def __init__(self):


@@ -963,17 +1026,13 @@ class PackIdxV2Writer:
             b = idx_f.read(8 + 4*256)
             idx_sum.update(b)
 
             b = idx_f.read(8 + 4*256)
             idx_sum.update(b)
 

-            obj_list_sum = Sha1()

             for b in chunkyreader(idx_f, 20 * self.count):
                 idx_sum.update(b)
             for b in chunkyreader(idx_f, 20 * self.count):
                 idx_sum.update(b)

-                obj_list_sum.update(b)
-            namebase = hexlify(obj_list_sum.digest())

 
             for b in chunkyreader(idx_f):
                 idx_sum.update(b)
             idx_f.write(idx_sum.digest())
             fdatasync(idx_f.fileno())
 
             for b in chunkyreader(idx_f):
                 idx_sum.update(b)
             idx_f.write(idx_sum.digest())
             fdatasync(idx_f.fileno())

-            return namebase

         finally:
             idx_f.close()
 
         finally:
             idx_f.close()
 


@@ -1068,17 +1127,6 @@ def rev_list(ref_or_refs, parse=None, format=None, repo_dir=None):
         raise GitError('git rev-list returned error %d' % rv)
 
 
         raise GitError('git rev-list returned error %d' % rv)
 
 

-def get_commit_dates(refs, repo_dir=None):
-    """Get the dates for the specified commit refs.  For now, every unique
-       string in refs must resolve to a different commit or this
-       function will fail."""
-    result = []
-    for ref in refs:
-        commit = get_commit_items(ref, cp(repo_dir))
-        result.append(commit.author_sec)
-    return result
-
-

 def rev_parse(committish, repo_dir=None):
     """Resolve the full hash for 'committish', if it exists.
 
 def rev_parse(committish, repo_dir=None):
     """Resolve the full hash for 'committish', if it exists.
 


@@ -1092,28 +1140,37 @@ def rev_parse(committish, repo_dir=None):
         debug2("resolved from ref: commit = %s\n" % hexlify(head))
         return head
 
         debug2("resolved from ref: commit = %s\n" % hexlify(head))
         return head
 

-    pL = PackIdxList(repo(b'objects/pack', repo_dir=repo_dir))
-

     if len(committish) == 40:
         try:
             hash = unhexlify(committish)
         except TypeError:
             return None
 
     if len(committish) == 40:
         try:
             hash = unhexlify(committish)
         except TypeError:
             return None
 

-        if pL.exists(hash):
-            return hash
+        with PackIdxList(repo(b'objects/pack', repo_dir=repo_dir)) as pL:
+            if pL.exists(hash):
+                return hash

 
     return None
 
 
 
     return None
 
 

-def update_ref(refname, newval, oldval, repo_dir=None):
-    """Update a repository reference."""
-    if not oldval:
-        oldval = b''
+def update_ref(refname, newval, oldval, repo_dir=None, force=False):
+    """Update a repository reference.
+
+    With force=True, don't care about the previous ref (oldval);
+    with force=False oldval must be either a sha1 or None (for an
+    entirely new branch)
+    """
+    if force:
+        assert oldval is None
+        oldarg = []
+    elif not oldval:
+        oldarg = [b'']
+    else:
+        oldarg = [hexlify(oldval)]

     assert refname.startswith(b'refs/heads/') \
         or refname.startswith(b'refs/tags/')
     p = subprocess.Popen([b'git', b'update-ref', refname,
     assert refname.startswith(b'refs/heads/') \
         or refname.startswith(b'refs/tags/')
     p = subprocess.Popen([b'git', b'update-ref', refname,

-                          hexlify(newval), hexlify(oldval)],
+                          hexlify(newval)] + oldarg,

                          env=_gitenv(repo_dir),
                          close_fds=True)
     _git_wait(b'git update-ref', p)
                          env=_gitenv(repo_dir),
                          close_fds=True)
     _git_wait(b'git update-ref', p)


@@ -1237,38 +1294,6 @@ def require_suitable_git(ver_str=None):
     assert False
 
 
     assert False
 
 

-class _AbortableIter:
-    def __init__(self, it, onabort = None):
-        self.it = it
-        self.onabort = onabort
-        self.done = None
-
-    def __iter__(self):
-        return self
-
-    def __next__(self):
-        try:
-            return next(self.it)
-        except StopIteration as e:
-            self.done = True
-            raise
-        except:
-            self.abort()
-            raise
-
-    next = __next__
-
-    def abort(self):
-        """Abort iteration and call the abortion callback, if needed."""
-        if not self.done:
-            self.done = True
-            if self.onabort:
-                self.onabort()
-
-    def __del__(self):
-        self.abort()
-
-

 class CatPipe:
     """Link to 'git cat-file' that is used to retrieve blob data."""
     def __init__(self, repo_dir = None):
 class CatPipe:
     """Link to 'git cat-file' that is used to retrieve blob data."""
     def __init__(self, repo_dir = None):


@@ -1277,15 +1302,19 @@ class CatPipe:
         self.p = self.inprogress = None
 
     def close(self, wait=False):
         self.p = self.inprogress = None
 
     def close(self, wait=False):

-        p = self.p
-        if p:
-            p.stdout.close()
-            p.stdin.close()
-        self.p = None
+        self.p, p = None, self.p

         self.inprogress = None
         self.inprogress = None

+        if p:
+            try:
+                p.stdout.close()
+            finally:
+                # This will handle pending exceptions correctly once
+                # we drop py2
+                p.stdin.close()

         if wait:
             p.wait()
             return p.returncode
         if wait:
             p.wait()
             return p.returncode

+        return None

 
     def restart(self):
         self.close()
 
     def restart(self):
         self.close()


@@ -1316,6 +1345,9 @@ class CatPipe:
         self.p.stdin.write(ref + b'\n')
         self.p.stdin.flush()
         hdr = self.p.stdout.readline()
         self.p.stdin.write(ref + b'\n')
         self.p.stdin.flush()
         hdr = self.p.stdout.readline()

+        if not hdr:
+            raise GitError('unexpected cat-file EOF (last request: %r, exit: %s)'
+                           % (ref, self.p.poll() or 'none'))

         if hdr.endswith(b' missing\n'):
             self.inprogress = None
             yield None, None, None
         if hdr.endswith(b' missing\n'):
             self.inprogress = None
             yield None, None, None


@@ -1325,18 +1357,17 @@ class CatPipe:
             raise GitError('expected object (id, type, size), got %r' % info)
         oidx, typ, size = info
         size = int(size)
             raise GitError('expected object (id, type, size), got %r' % info)
         oidx, typ, size = info
         size = int(size)

-        it = _AbortableIter(chunkyreader(self.p.stdout, size),
-                            onabort=self.close)

         try:
         try:

+            it = chunkyreader(self.p.stdout, size)

             yield oidx, typ, size
             yield oidx, typ, size

-            for blob in it:
+            for blob in chunkyreader(self.p.stdout, size):

                 yield blob
             readline_result = self.p.stdout.readline()
             assert readline_result == b'\n'
             self.inprogress = None
                 yield blob
             readline_result = self.p.stdout.readline()
             assert readline_result == b'\n'
             self.inprogress = None

-        except Exception as e:
-            it.abort()
-            raise
+        except Exception as ex:
+            with pending_raise(ex):
+                self.close()

 
     def _join(self, it):
         _, typ, _ = next(it)
 
     def _join(self, it):
         _, typ, _ = next(it)