X-Git-Url: https://arthur.barton.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fbup%2Fgit.py;h=69134ac33ae4ba0df4eb6bb691b51bbf6e7572e7;hb=84cb94e1feb584ae7c67d65c27b2ff1a8e01e489;hp=61238bd250f42c72cef19914286a0d4e9ed3d7a6;hpb=bb0e9cbf3900e65d2fddbe888e6cb21c59b308df;p=bup.git diff --git a/lib/bup/git.py b/lib/bup/git.py index 61238bd..69134ac 100644 --- a/lib/bup/git.py +++ b/lib/bup/git.py @@ -4,33 +4,34 @@ interact with the Git data structures. """ 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 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, + ExitStack, items, + pending_raise, range, 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, - parse_num, + nullcontext_if_not, progress, qprogress, stat_if_exists, unlink, utc_offset_str) -from bup.pwdgrp import username, userfullname verbose = 0 @@ -66,14 +67,32 @@ def _git_exo(cmd, **kwargs): 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) - 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: + 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)) @@ -87,9 +106,26 @@ def parse_tz_offset(s): 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. + +# 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))' \ @@ -103,7 +139,7 @@ _mergetag_rx = br'(?:\nmergetag object [abcdefABCDEF0123456789]{40}(?:\n [^\0\n] _commit_rx = re.compile(br'''tree (?P[abcdefABCDEF0123456789]{40}) (?P%s*)author (?P%s) <(?P%s)> (?P\d+) (?P%s) committer (?P%s) <(?P%s)> (?P\d+) (?P%s)(?P%s?) - +(?Pgpgsig .*\n(?: .*\n)*)? (?P(?:.|\n)*)''' % (_parent_rx, _safe_str_rx, _safe_str_rx, _tz_rx, _safe_str_rx, _safe_str_rx, _tz_rx, @@ -117,6 +153,7 @@ CommitInfo = namedtuple('CommitInfo', ['tree', 'parents', 'author_sec', 'author_offset', 'committer_name', 'committer_mail', 'committer_sec', 'committer_offset', + 'gpgsig', 'message']) def parse_commit(content): @@ -134,6 +171,7 @@ def parse_commit(content): 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']) @@ -248,8 +286,7 @@ def demangle_name(name, mode): 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): @@ -377,6 +414,7 @@ class PackIdx: class PackIdxV1(PackIdx): """Object representation of a Git pack index (version 1) file.""" def __init__(self, filename, f): + self.closed = False self.name = filename self.idxnames = [self.name] self.map = mmap_read(f) @@ -392,7 +430,8 @@ class PackIdxV1(PackIdx): 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 @@ -415,15 +454,20 @@ class PackIdxV1(PackIdx): 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 + def __del__(self): + assert self.closed + class PackIdxV2(PackIdx): """Object representation of a Git pack index (version 2) file.""" def __init__(self, filename, f): + self.closed = False self.name = filename self.idxnames = [self.name] self.map = mmap_read(f) @@ -442,7 +486,8 @@ class PackIdxV2(PackIdx): 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 @@ -470,30 +515,62 @@ class PackIdxV2(PackIdx): 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 + def __del__(self): + assert self.closed + _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 + self.open = True 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 + if not self.open: + assert _mpi_count == 0 + return _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)) @@ -574,7 +651,6 @@ class PackIdxList: broken = True if broken: mx.close() - del mx unlink(full) else: midxl.append(mx) @@ -606,14 +682,27 @@ class PackIdxList: 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) - 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 '')) @@ -677,18 +766,18 @@ def create_commit_blob(tree, parent, 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. -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): + self.closed = False self.repo_dir = repo_dir or repo() self.file = None self.parentfd = None @@ -703,9 +792,8 @@ class PackWriter: 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 @@ -714,14 +802,12 @@ class PackWriter: 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): - self.close() + with pending_raise(value, rethrow=False): + self.close() def _open(self): if not self.file: @@ -782,12 +868,6 @@ class PackWriter: 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() @@ -836,73 +916,75 @@ class PackWriter: msg) return self.maybe_write(b'commit', content) - 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: + 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 + + if abort: 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()) - finally: - f.close() - - 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') - try: - os.fsync(self.parentfd) + 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: - 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.""" + self.closed = True return self._end(run_midx=run_midx) + def __del__(self): + assert self.closed + class PackIdxV2Writer: def __init__(self): @@ -1061,28 +1143,37 @@ def rev_parse(committish, repo_dir=None): 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 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 -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, - hexlify(newval), hexlify(oldval)], + hexlify(newval)] + oldarg, env=_gitenv(repo_dir), close_fds=True) _git_wait(b'git update-ref', p) @@ -1206,38 +1297,6 @@ def require_suitable_git(ver_str=None): 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): @@ -1246,15 +1305,19 @@ class CatPipe: 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 + 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 + return None def restart(self): self.close() @@ -1285,6 +1348,9 @@ class CatPipe: 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 @@ -1294,18 +1360,17 @@ class CatPipe: 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: + it = chunkyreader(self.p.stdout, 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 - except Exception as e: - it.abort() - raise + except Exception as ex: + with pending_raise(ex): + self.close() def _join(self, it): _, typ, _ = next(it)