diff options
Diffstat (limited to 'git')
-rw-r--r-- | git/cmd.py | 111 | ||||
-rw-r--r-- | git/config.py | 36 | ||||
-rw-r--r-- | git/exc.py | 11 | ||||
-rw-r--r-- | git/objects/submodule/base.py | 652 | ||||
-rw-r--r-- | git/objects/submodule/root.py | 431 | ||||
-rw-r--r-- | git/objects/submodule/util.py | 2 | ||||
-rw-r--r-- | git/repo/base.py | 18 | ||||
-rw-r--r-- | git/repo/fun.py | 3 | ||||
-rw-r--r-- | git/test/test_config.py | 14 | ||||
-rw-r--r-- | git/test/test_docs.py | 26 | ||||
-rw-r--r-- | git/test/test_index.py | 3 | ||||
-rw-r--r-- | git/test/test_repo.py | 6 | ||||
-rw-r--r-- | git/test/test_submodule.py | 295 | ||||
-rw-r--r-- | git/util.py | 17 |
14 files changed, 1119 insertions, 506 deletions
@@ -9,6 +9,9 @@ import sys import select import logging import threading +import errno +import mmap + from subprocess import ( call, Popen, @@ -26,6 +29,7 @@ from git.compat import ( string_types, defenc, PY3, + bchr, # just to satisfy flake8 on py3 unicode ) @@ -41,6 +45,13 @@ __all__ = ('Git', ) if sys.platform != 'win32': WindowsError = OSError +if PY3: + _bchr = bchr +else: + def _bchr(c): + return c +# get custom byte character handling + # ============================================================================== ## @name Utilities @@ -58,52 +69,73 @@ def handle_process_output(process, stdout_handler, stderr_handler, finalizer): :param stdout_handler: f(stdout_line_string), or None :param stderr_hanlder: f(stderr_line_string), or None :param finalizer: f(proc) - wait for proc to finish""" - def read_line_fast(stream): - return stream.readline() - - def read_line_slow(stream): + def parse_lines_from_buffer(fno, buf): line = b'' - while True: - char = stream.read(1) # reads individual single byte strings - if not char: - break + bi = 0 + lb = len(buf) + while bi < lb: + char = _bchr(buf[bi]) + bi += 1 if char in (b'\r', b'\n') and line: - break + yield bi, line + line = b'' else: line += char # END process parsed line # END while file is not done reading - return line # end - def dispatch_line(stream, handler, readline): - # this can possibly block for a while, but since we wake-up with at least one or more lines to handle, - # we are good ... - line = readline(stream).decode(defenc) - if line and handler: - try: - handler(line) - except Exception: - # Keep reading, have to pump the lines empty nontheless - log.error("Line handler exception on line: %s", line, exc_info=True) - # end - return line + def read_lines_from_fno(fno, last_buf_list): + buf = os.read(fno, mmap.PAGESIZE) + buf = last_buf_list[0] + buf + + bi = 0 + for bi, line in parse_lines_from_buffer(fno, buf): + yield line + # for each line to parse from the buffer + + # keep remainder + last_buf_list[0] = buf[bi:] + + def dispatch_single_line(line, handler): + line = line.decode(defenc) + if line and handler: + try: + handler(line) + except Exception: + # Keep reading, have to pump the lines empty nontheless + log.error("Line handler exception on line: %s", line, exc_info=True) + # end # end dispatch helper + # end single line helper + + def dispatch_lines(fno, handler, buf_list): + lc = 0 + for line in read_lines_from_fno(fno, buf_list): + dispatch_single_line(line, handler) + lc += 1 + # for each line + return lc # end - def deplete_buffer(stream, handler, readline, wg=None): + def deplete_buffer(fno, handler, buf_list, wg=None): while True: - line = dispatch_line(stream, handler, readline) - if not line: + line_count = dispatch_lines(fno, handler, buf_list) + if line_count == 0: break # end deplete buffer + + if buf_list[0]: + dispatch_single_line(buf_list[0], handler) + # end + if wg: wg.done() # end - fdmap = {process.stdout.fileno(): (process.stdout, stdout_handler, read_line_fast), - process.stderr.fileno(): (process.stderr, stderr_handler, read_line_slow)} + fdmap = {process.stdout.fileno(): (stdout_handler, [b'']), + process.stderr.fileno(): (stderr_handler, [b''])} if hasattr(select, 'poll'): # poll is preferred, as select is limited to file handles up to 1024 ... . This could otherwise be @@ -118,12 +150,20 @@ def handle_process_output(process, stdout_handler, stderr_handler, finalizer): closed_streams = set() while True: # no timeout - poll_result = poll.poll() + + try: + poll_result = poll.poll() + except select.error as e: + if e.args[0] == errno.EINTR: + continue + raise + # end handle poll exception + for fd, result in poll_result: if result & CLOSED: closed_streams.add(fd) else: - dispatch_line(*fdmap[fd]) + dispatch_lines(fd, *fdmap[fd]) # end handle closed stream # end for each poll-result tuple @@ -133,19 +173,22 @@ def handle_process_output(process, stdout_handler, stderr_handler, finalizer): # end endless loop # Depelete all remaining buffers - for stream, handler, readline in fdmap.values(): - deplete_buffer(stream, handler, readline) + for fno, (handler, buf_list) in fdmap.items(): + deplete_buffer(fno, handler, buf_list) # end for each file handle + + for fno in fdmap.keys(): + poll.unregister(fno) + # end don't forget to unregister ! else: # Oh ... probably we are on windows. select.select() can only handle sockets, we have files # The only reliable way to do this now is to use threads and wait for both to finish # Since the finalizer is expected to wait, we don't have to introduce our own wait primitive # NO: It's not enough unfortunately, and we will have to sync the threads wg = WaitGroup() - for fno in fdmap.keys(): + for fno, (handler, buf_list) in fdmap.items(): wg.add(1) - stream, handler, readline = fdmap[fno] - t = threading.Thread(target=lambda: deplete_buffer(stream, handler, readline, wg)) + t = threading.Thread(target=lambda: deplete_buffer(fno, handler, buf_list, wg)) t.start() # end # NOTE: Just joining threads can possibly fail as there is a gap between .start() and when it's diff --git a/git/config.py b/git/config.py index 4c4cb491..38dd1b44 100644 --- a/git/config.py +++ b/git/config.py @@ -23,6 +23,7 @@ from git.compat import ( string_types, FileType, defenc, + force_text, with_metaclass, PY3 ) @@ -412,7 +413,7 @@ class GitConfigParser(with_metaclass(MetaParserBuilder, cp.RawConfigParser, obje fp.write(("[%s]\n" % name).encode(defenc)) for (key, value) in section_dict.items(): if key != "__name__": - fp.write(("\t%s = %s\n" % (key, str(value).replace('\n', '\n\t'))).encode(defenc)) + fp.write(("\t%s = %s\n" % (key, self._value_to_string(value).replace('\n', '\n\t'))).encode(defenc)) # END if key is not __name__ # END section writing @@ -478,8 +479,6 @@ class GitConfigParser(with_metaclass(MetaParserBuilder, cp.RawConfigParser, obje if self.read_only: raise IOError("Cannot execute non-constant method %s.%s" % (self, method_name)) - @needs_values - @set_dirty_and_flush_changes def add_section(self, section): """Assures added options will stay in order""" return super(GitConfigParser, self).add_section(section) @@ -531,6 +530,11 @@ class GitConfigParser(with_metaclass(MetaParserBuilder, cp.RawConfigParser, obje return valuestr + def _value_to_string(self, value): + if isinstance(value, (int, float, bool)): + return str(value) + return force_text(value) + @needs_values @set_dirty_and_flush_changes def set_value(self, section, option, value): @@ -542,7 +546,29 @@ class GitConfigParser(with_metaclass(MetaParserBuilder, cp.RawConfigParser, obje :param option: Name of the options whose value to set :param value: Value to set the option to. It must be a string or convertible - to a string""" + to a string + :return: this instance""" if not self.has_section(section): self.add_section(section) - self.set(section, option, str(value)) + self.set(section, option, self._value_to_string(value)) + return self + + def rename_section(self, section, new_name): + """rename the given section to new_name + :raise ValueError: if section doesn't exit + :raise ValueError: if a section with new_name does already exist + :return: this instance + """ + if not self.has_section(section): + raise ValueError("Source section '%s' doesn't exist" % section) + if self.has_section(new_name): + raise ValueError("Destination section '%s' already exists" % new_name) + + super(GitConfigParser, self).add_section(new_name) + for k, v in self.items(section): + self.set(new_name, k, self._value_to_string(v)) + # end for each value to copy + + # This call writes back the changes, which is why we don't have the respective decorator + self.remove_section(section) + return self @@ -84,3 +84,14 @@ class HookExecutionError(Exception): def __str__(self): return ("'%s' hook returned with exit code %i\nstdout: '%s'\nstderr: '%s'" % (self.command, self.status, self.stdout, self.stderr)) + + +class RepositoryDirtyError(Exception): + """Thrown whenever an operation on a repository fails as it has uncommited changes that would be overwritten""" + + def __init__(self, repo, message): + self.repo = repo + self.message = message + + def __str__(self): + return "Operation cannot be performed on %r: %s" % (self.repo, self.message) diff --git a/git/objects/submodule/base.py b/git/objects/submodule/base.py index e055cb8c..ebb66495 100644 --- a/git/objects/submodule/base.py +++ b/git/objects/submodule/base.py @@ -19,19 +19,25 @@ from git.util import ( from git.config import ( SectionConstraint, + GitConfigParser, cp ) from git.exc import ( InvalidGitRepositoryError, - NoSuchPathError + NoSuchPathError, + RepositoryDirtyError +) +from git.compat import ( + string_types, + defenc ) -from git.compat import string_types import stat import git import os import logging +import tempfile __all__ = ["Submodule", "UpdateProgress"] @@ -92,8 +98,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): :param branch_path: full (relative) path to ref to checkout when cloning the remote repository""" super(Submodule, self).__init__(repo, binsha, mode, path) self.size = 0 - if parent_commit is not None: - self._parent_commit = parent_commit + self._parent_commit = parent_commit if url is not None: self._url = url if branch_path is not None: @@ -103,13 +108,15 @@ class Submodule(util.IndexObject, Iterable, Traversable): self._name = name def _set_cache_(self, attr): - if attr == '_parent_commit': - # set a default value, which is the root tree of the current head - self._parent_commit = self.repo.commit() - elif attr in ('path', '_url', '_branch_path'): + if attr in ('path', '_url', '_branch_path'): reader = self.config_reader() # default submodule values - self.path = reader.get_value('path') + try: + self.path = reader.get_value('path') + except cp.NoSectionError: + raise ValueError("This submodule instance does not exist anymore in '%s' file" + % os.path.join(self.repo.working_tree_dir, '.gitmodules')) + # end self._url = reader.get_value('url') # git-python extension values - optional self._branch_path = reader.get_value(self.k_head_option, git.Head.to_full_path(self.k_head_default)) @@ -127,6 +134,10 @@ class Submodule(util.IndexObject, Iterable, Traversable): return list() # END handle intermeditate items + @classmethod + def _need_gitfile_submodules(cls, git): + return git.version_info[:3] >= (1, 8, 0) + def __eq__(self, other): """Compare with another submodule""" # we may only compare by name as this should be the ID they are hashed with @@ -155,12 +166,19 @@ class Submodule(util.IndexObject, Iterable, Traversable): :raise IOError: If the .gitmodules file cannot be found, either locally or in the repository at the given parent commit. Otherwise the exception would be delayed until the first access of the config parser""" - parent_matches_head = repo.head.commit == parent_commit + parent_matches_head = True + if parent_commit is not None: + try: + parent_matches_head = repo.head.commit == parent_commit + except ValueError: + # We are most likely in an empty repository, so the HEAD doesn't point to a valid ref + pass + # end hanlde parent_commit + if not repo.bare and parent_matches_head: - fp_module = cls.k_modules_file - fp_module_path = os.path.join(repo.working_tree_dir, fp_module) - fp_module = fp_module_path + fp_module = os.path.join(repo.working_tree_dir, cls.k_modules_file) else: + assert parent_commit is not None, "need valid parent_commit in bare repositories" try: fp_module = cls._sio_modules(parent_commit) except KeyError: @@ -194,10 +212,90 @@ class Submodule(util.IndexObject, Iterable, Traversable): def _config_parser_constrained(self, read_only): """:return: Config Parser constrained to our submodule in read or write mode""" - parser = self._config_parser(self.repo, self._parent_commit, read_only) + try: + pc = self.parent_commit + except ValueError: + pc = None + # end hande empty parent repository + parser = self._config_parser(self.repo, pc, read_only) parser.set_submodule(self) return SectionConstraint(parser, sm_section(self.name)) + @classmethod + def _module_abspath(cls, parent_repo, path, name): + if cls._need_gitfile_submodules(parent_repo.git): + return os.path.join(parent_repo.git_dir, 'modules', name) + else: + return os.path.join(parent_repo.working_tree_dir, path) + # end + + @classmethod + def _clone_repo(cls, repo, url, path, name, **kwargs): + """:return: Repo instance of newly cloned repository + :param repo: our parent repository + :param url: url to clone from + :param path: repository-relative path to the submodule checkout location + :param name: canonical of the submodule + :param kwrags: additinoal arguments given to git.clone""" + module_abspath = cls._module_abspath(repo, path, name) + module_checkout_path = module_abspath + if cls._need_gitfile_submodules(repo.git): + kwargs['separate_git_dir'] = module_abspath + module_abspath_dir = os.path.dirname(module_abspath) + if not os.path.isdir(module_abspath_dir): + os.makedirs(module_abspath_dir) + module_checkout_path = os.path.join(repo.working_tree_dir, path) + # end + + clone = git.Repo.clone_from(url, module_checkout_path, **kwargs) + if cls._need_gitfile_submodules(repo.git): + cls._write_git_file_and_module_config(module_checkout_path, module_abspath) + # end + return clone + + @classmethod + def _to_relative_path(cls, parent_repo, path): + """:return: a path guaranteed to be relative to the given parent-repository + :raise ValueError: if path is not contained in the parent repository's working tree""" + path = to_native_path_linux(path) + if path.endswith('/'): + path = path[:-1] + # END handle trailing slash + + if os.path.isabs(path): + working_tree_linux = to_native_path_linux(parent_repo.working_tree_dir) + if not path.startswith(working_tree_linux): + raise ValueError("Submodule checkout path '%s' needs to be within the parents repository at '%s'" + % (working_tree_linux, path)) + path = path[len(working_tree_linux) + 1:] + if not path: + raise ValueError("Absolute submodule path '%s' didn't yield a valid relative path" % path) + # end verify converted relative path makes sense + # end convert to a relative path + + return path + + @classmethod + def _write_git_file_and_module_config(cls, working_tree_dir, module_abspath): + """Writes a .git file containing a (preferably) relative path to the actual git module repository. + It is an error if the module_abspath cannot be made into a relative path, relative to the working_tree_dir + :note: will overwrite existing files ! + :note: as we rewrite both the git file as well as the module configuration, we might fail on the configuration + and will not roll back changes done to the git file. This should be a non-issue, but may easily be fixed + if it becomes one + :param working_tree_dir: directory to write the .git file into + :param module_abspath: absolute path to the bare repository + """ + git_file = os.path.join(working_tree_dir, '.git') + rela_path = os.path.relpath(module_abspath, start=working_tree_dir) + fp = open(git_file, 'wb') + fp.write(("gitdir: %s" % rela_path).encode(defenc)) + fp.close() + + writer = GitConfigParser(os.path.join(module_abspath, 'config'), read_only=False, merge_includes=False) + writer.set_value('core', 'worktree', os.path.relpath(working_tree_dir, start=module_abspath)) + writer.release() + #{ Edit Interface @classmethod @@ -233,21 +331,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): raise InvalidGitRepositoryError("Cannot add submodules to bare repositories") # END handle bare repos - path = to_native_path_linux(path) - if path.endswith('/'): - path = path[:-1] - # END handle trailing slash - - if os.path.isabs(path): - working_tree_linux = to_native_path_linux(repo.working_tree_dir) - if not path.startswith(working_tree_linux): - raise ValueError("Submodule checkout path '%s' needs to be within the parents repository at '%s'" - % (working_tree_linux, path)) - path = path[len(working_tree_linux) + 1:] - if not path: - raise ValueError("Absolute submodule path '%s' didn't yield a valid relative path" % path) - # end verify converted relative path makes sense - # end convert to a relative path + path = cls._to_relative_path(repo, path) # assure we never put backslashes into the url, as some operating systems # like it ... @@ -298,9 +382,19 @@ class Submodule(util.IndexObject, Iterable, Traversable): if not branch_is_default: kwargs['b'] = br.name # END setup checkout-branch - mrepo = git.Repo.clone_from(url, os.path.join(repo.working_tree_dir, path), **kwargs) + + # _clone_repo(cls, repo, url, path, name, **kwargs): + mrepo = cls._clone_repo(repo, url, path, name, **kwargs) # END verify url + # It's important to add the URL to the parent config, to let `git submodule` know. + # otherwise there is a '-' character in front of the submodule listing + # a38efa84daef914e4de58d1905a500d8d14aaf45 mymodule (v0.9.0-1-ga38efa8) + # -a38efa84daef914e4de58d1905a500d8d14aaf45 submodules/intermediate/one + writer = sm.repo.config_writer() + writer.set_value(sm_section(name), 'url', url) + writer.release() + # update configuration and index index = sm.repo.index writer = sm.config_writer(index=index, write=False) @@ -317,15 +411,13 @@ class Submodule(util.IndexObject, Iterable, Traversable): del(writer) # we deliberatly assume that our head matches our index ! - pcommit = repo.head.commit - sm._parent_commit = pcommit sm.binsha = mrepo.head.commit.binsha index.add([sm], write=True) return sm - def update(self, recursive=False, init=True, to_latest_revision=False, progress=None, - dry_run=False): + def update(self, recursive=False, init=True, to_latest_revision=False, progress=None, dry_run=False, + force=False, keep_going=False): """Update the repository of this submodule to point to the checkout we point at with the binsha of this instance. @@ -340,6 +432,16 @@ class Submodule(util.IndexObject, Iterable, Traversable): :param progress: UpdateProgress instance or None of no progress should be shown :param dry_run: if True, the operation will only be simulated, but not performed. All performed operations are read-only + :param force: + If True, we may reset heads even if the repository in question is dirty. Additinoally we will be allowed + to set a tracking branch which is ahead of its remote branch back into the past or the location of the + remote branch. This will essentially 'forget' commits. + If False, local tracking branches that are in the future of their respective remote branches will simply + not be moved. + :param keep_going: if True, we will ignore but log all errors, and keep going recursively. + Unless dry_run is set as well, keep_going could cause subsequent/inherited errors you wouldn't see + otherwise. + In conjunction with dry_run, it can be useful to anticipate all errors when updating submodules :note: does nothing in bare repositories :note: method is definitely not atomic if recurisve is True :return: self""" @@ -360,130 +462,160 @@ class Submodule(util.IndexObject, Iterable, Traversable): mrepo = None # END init mrepo - # ASSURE REPO IS PRESENT AND UPTODATE - ##################################### try: - mrepo = self.module() - rmts = mrepo.remotes - len_rmts = len(rmts) - for i, remote in enumerate(rmts): - op = FETCH - if i == 0: - op |= BEGIN - # END handle start - - progress.update(op, i, len_rmts, prefix + "Fetching remote %s of submodule %r" % (remote, self.name)) - #=============================== + # ASSURE REPO IS PRESENT AND UPTODATE + ##################################### + try: + mrepo = self.module() + rmts = mrepo.remotes + len_rmts = len(rmts) + for i, remote in enumerate(rmts): + op = FETCH + if i == 0: + op |= BEGIN + # END handle start + + progress.update(op, i, len_rmts, prefix + "Fetching remote %s of submodule %r" + % (remote, self.name)) + #=============================== + if not dry_run: + remote.fetch(progress=progress) + # END handle dry-run + #=============================== + if i == len_rmts - 1: + op |= END + # END handle end + progress.update(op, i, len_rmts, prefix + "Done fetching remote of submodule %r" % self.name) + # END fetch new data + except InvalidGitRepositoryError: + if not init: + return self + # END early abort if init is not allowed + + # there is no git-repository yet - but delete empty paths + checkout_module_abspath = self.abspath + if not dry_run and os.path.isdir(checkout_module_abspath): + try: + os.rmdir(checkout_module_abspath) + except OSError: + raise OSError("Module directory at %r does already exist and is non-empty" + % checkout_module_abspath) + # END handle OSError + # END handle directory removal + + # don't check it out at first - nonetheless it will create a local + # branch according to the remote-HEAD if possible + progress.update(BEGIN | CLONE, 0, 1, prefix + "Cloning url '%s' to '%s' in submodule %r" % + (self.url, checkout_module_abspath, self.name)) if not dry_run: - remote.fetch(progress=progress) + mrepo = self._clone_repo(self.repo, self.url, self.path, self.name, n=True) # END handle dry-run - #=============================== - if i == len_rmts - 1: - op |= END - # END handle end - progress.update(op, i, len_rmts, prefix + "Done fetching remote of submodule %r" % self.name) - # END fetch new data - except InvalidGitRepositoryError: - if not init: - return self - # END early abort if init is not allowed - import git - - # there is no git-repository yet - but delete empty paths - module_path = join_path_native(self.repo.working_tree_dir, self.path) - if not dry_run and os.path.isdir(module_path): - try: - os.rmdir(module_path) - except OSError: - raise OSError("Module directory at %r does already exist and is non-empty" % module_path) - # END handle OSError - # END handle directory removal - - # don't check it out at first - nonetheless it will create a local - # branch according to the remote-HEAD if possible - progress.update(BEGIN | CLONE, 0, 1, prefix + "Cloning %s to %s in submodule %r" % - (self.url, module_path, self.name)) - if not dry_run: - mrepo = git.Repo.clone_from(self.url, module_path, n=True) - # END handle dry-run - progress.update(END | CLONE, 0, 1, prefix + "Done cloning to %s" % module_path) - - if not dry_run: - # see whether we have a valid branch to checkout - try: - # find a remote which has our branch - we try to be flexible - remote_branch = find_first_remote_branch(mrepo.remotes, self.branch_name) - local_branch = mkhead(mrepo, self.branch_path) - - # have a valid branch, but no checkout - make sure we can figure - # that out by marking the commit with a null_sha - local_branch.set_object(util.Object(mrepo, self.NULL_BIN_SHA)) - # END initial checkout + branch creation - - # make sure HEAD is not detached - mrepo.head.set_reference(local_branch, logmsg="submodule: attaching head to %s" % local_branch) - mrepo.head.ref.set_tracking_branch(remote_branch) - except IndexError: - log.warn("Failed to checkout tracking branch %s", self.branch_path) - # END handle tracking branch - - # NOTE: Have to write the repo config file as well, otherwise - # the default implementation will be offended and not update the repository - # Maybe this is a good way to assure it doesn't get into our way, but - # we want to stay backwards compatible too ... . Its so redundant ! - writer = self.repo.config_writer() - writer.set_value(sm_section(self.name), 'url', self.url) - writer.release() + progress.update(END | CLONE, 0, 1, prefix + "Done cloning to %s" % checkout_module_abspath) + + if not dry_run: + # see whether we have a valid branch to checkout + try: + # find a remote which has our branch - we try to be flexible + remote_branch = find_first_remote_branch(mrepo.remotes, self.branch_name) + local_branch = mkhead(mrepo, self.branch_path) + + # have a valid branch, but no checkout - make sure we can figure + # that out by marking the commit with a null_sha + local_branch.set_object(util.Object(mrepo, self.NULL_BIN_SHA)) + # END initial checkout + branch creation + + # make sure HEAD is not detached + mrepo.head.set_reference(local_branch, logmsg="submodule: attaching head to %s" % local_branch) + mrepo.head.ref.set_tracking_branch(remote_branch) + except IndexError: + log.warn("Failed to checkout tracking branch %s", self.branch_path) + # END handle tracking branch + + # NOTE: Have to write the repo config file as well, otherwise + # the default implementation will be offended and not update the repository + # Maybe this is a good way to assure it doesn't get into our way, but + # we want to stay backwards compatible too ... . Its so redundant ! + writer = self.repo.config_writer() + writer.set_value(sm_section(self.name), 'url', self.url) + writer.release() + # END handle dry_run + # END handle initalization + + # DETERMINE SHAS TO CHECKOUT + ############################ + binsha = self.binsha + hexsha = self.hexsha + if mrepo is not None: + # mrepo is only set if we are not in dry-run mode or if the module existed + is_detached = mrepo.head.is_detached # END handle dry_run - # END handle initalization - - # DETERMINE SHAS TO CHECKOUT - ############################ - binsha = self.binsha - hexsha = self.hexsha - if mrepo is not None: - # mrepo is only set if we are not in dry-run mode or if the module existed - is_detached = mrepo.head.is_detached - # END handle dry_run - - if mrepo is not None and to_latest_revision: - msg_base = "Cannot update to latest revision in repository at %r as " % mrepo.working_dir - if not is_detached: - rref = mrepo.head.ref.tracking_branch() - if rref is not None: - rcommit = rref.commit - binsha = rcommit.binsha - hexsha = rcommit.hexsha - else: - log.error("%s a tracking branch was not set for local branch '%s'", msg_base, mrepo.head.ref) - # END handle remote ref - else: - log.error("%s there was no local tracking branch", msg_base) - # END handle detached head - # END handle to_latest_revision option - - # update the working tree - # handles dry_run - if mrepo is not None and mrepo.head.commit.binsha != binsha: - progress.update(BEGIN | UPDWKTREE, 0, 1, prefix + - "Updating working tree at %s for submodule %r to revision %s" - % (self.path, self.name, hexsha)) - if not dry_run: - if is_detached: - # NOTE: for now we force, the user is no supposed to change detached - # submodules anyway. Maybe at some point this becomes an option, to - # properly handle user modifications - see below for future options - # regarding rebase and merge. - mrepo.git.checkout(hexsha, force=True) + + if mrepo is not None and to_latest_revision: + msg_base = "Cannot update to latest revision in repository at %r as " % mrepo.working_dir + if not is_detached: + rref = mrepo.head.ref.tracking_branch() + if rref is not None: + rcommit = rref.commit + binsha = rcommit.binsha + hexsha = rcommit.hexsha + else: + log.error("%s a tracking branch was not set for local branch '%s'", msg_base, mrepo.head.ref) + # END handle remote ref else: - # TODO: allow to specify a rebase, merge, or reset - # TODO: Warn if the hexsha forces the tracking branch off the remote - # branch - this should be prevented when setting the branch option - mrepo.head.reset(hexsha, index=True, working_tree=True) - # END handle checkout - # END handle dry_run - progress.update(END | UPDWKTREE, 0, 1, prefix + "Done updating working tree for submodule %r" % self.name) - # END update to new commit only if needed + log.error("%s there was no local tracking branch", msg_base) + # END handle detached head + # END handle to_latest_revision option + + # update the working tree + # handles dry_run + if mrepo is not None and mrepo.head.commit.binsha != binsha: + # We must assure that our destination sha (the one to point to) is in the future of our current head. + # Otherwise, we will reset changes that might have been done on the submodule, but were not yet pushed + # We also handle the case that history has been rewritten, leaving no merge-base. In that case + # we behave conservatively, protecting possible changes the user had done + may_reset = True + if mrepo.head.commit.binsha != self.NULL_BIN_SHA: + base_commit = mrepo.merge_base(mrepo.head.commit, hexsha) + if len(base_commit) == 0 or base_commit[0].hexsha == hexsha: + if force: + log.debug("Will force checkout or reset on local branch that is possibly in the future of" + + "the commit it will be checked out to, effectively 'forgetting' new commits") + else: + log.info("Skipping %s on branch '%s' of submodule repo '%s' as it contains " + + "un-pushed commits", is_detached and "checkout" or "reset", mrepo.head, mrepo) + may_reset = False + # end handle force + # end handle if we are in the future + + if may_reset and not force and mrepo.is_dirty(index=True, working_tree=True, untracked_files=True): + raise RepositoryDirtyError(mrepo, "Cannot reset a dirty repository") + # end handle force and dirty state + # end handle empty repo + + # end verify future/past + progress.update(BEGIN | UPDWKTREE, 0, 1, prefix + + "Updating working tree at %s for submodule %r to revision %s" + % (self.path, self.name, hexsha)) + + if not dry_run and may_reset: + if is_detached: + # NOTE: for now we force, the user is no supposed to change detached + # submodules anyway. Maybe at some point this becomes an option, to + # properly handle user modifications - see below for future options + # regarding rebase and merge. + mrepo.git.checkout(hexsha, force=force) + else: + mrepo.head.reset(hexsha, index=True, working_tree=True) + # END handle checkout + # if we may reset/checkout + progress.update(END | UPDWKTREE, 0, 1, prefix + "Done updating working tree for submodule %r" + % self.name) + # END update to new commit only if needed + except Exception as err: + if not keep_going: + raise + log.error(str(err)) + # end handle keep_going # HANDLE RECURSION ################## @@ -491,7 +623,8 @@ class Submodule(util.IndexObject, Iterable, Traversable): # in dry_run mode, the module might not exist if mrepo is not None: for submodule in self.iter_items(self.module()): - submodule.update(recursive, init, to_latest_revision, progress=progress, dry_run=dry_run) + submodule.update(recursive, init, to_latest_revision, progress=progress, dry_run=dry_run, + force=force, keep_going=keep_going) # END handle recursive update # END handle dry run # END for each submodule @@ -504,16 +637,15 @@ class Submodule(util.IndexObject, Iterable, Traversable): the repository at our current path, changing the configuration, as well as adjusting our index entry accordingly. - :param module_path: the path to which to move our module, given as - repository-relative path. Intermediate directories will be created + :param module_path: the path to which to move our module in the parent repostory's working tree, + given as repository-relative or absolute path. Intermediate directories will be created accordingly. If the path already exists, it must be empty. - Trailling (back)slashes are removed automatically + Trailing (back)slashes are removed automatically :param configuration: if True, the configuration will be adjusted to let the submodule point to the given path. :param module: if True, the repository managed by this submodule - will be moved, not the configuration. This will effectively - leave your repository in an inconsistent state unless the configuration - and index already point to the target location. + will be moved as well. If False, we don't move the submodule's checkout, which may leave + the parent repository in an inconsistent state. :return: self :raise ValueError: if the module path existed and was not empty, or was a file :note: Currently the method is not atomic, and it could leave the repository @@ -523,64 +655,67 @@ class Submodule(util.IndexObject, Iterable, Traversable): raise ValueError("You must specify to move at least the module or the configuration of the submodule") # END handle input - module_path = to_native_path_linux(module_path) - if module_path.endswith('/'): - module_path = module_path[:-1] - # END handle trailing slash + module_checkout_path = self._to_relative_path(self.repo, module_path) # VERIFY DESTINATION - if module_path == self.path: + if module_checkout_path == self.path: return self # END handle no change - dest_path = join_path_native(self.repo.working_tree_dir, module_path) - if os.path.isfile(dest_path): - raise ValueError("Cannot move repository onto a file: %s" % dest_path) + module_checkout_abspath = join_path_native(self.repo.working_tree_dir, module_checkout_path) + if os.path.isfile(module_checkout_abspath): + raise ValueError("Cannot move repository onto a file: %s" % module_checkout_abspath) # END handle target files index = self.repo.index - tekey = index.entry_key(module_path, 0) + tekey = index.entry_key(module_checkout_path, 0) # if the target item already exists, fail if configuration and tekey in index.entries: - raise ValueError("Index entry for target path did alredy exist") + raise ValueError("Index entry for target path did already exist") # END handle index key already there # remove existing destination if module: - if os.path.exists(dest_path): - if len(os.listdir(dest_path)): + if os.path.exists(module_checkout_abspath): + if len(os.listdir(module_checkout_abspath)): raise ValueError("Destination module directory was not empty") - # END handle non-emptyness + # END handle non-emptiness - if os.path.islink(dest_path): - os.remove(dest_path) + if os.path.islink(module_checkout_abspath): + os.remove(module_checkout_abspath) else: - os.rmdir(dest_path) + os.rmdir(module_checkout_abspath) # END handle link else: # recreate parent directories # NOTE: renames() does that now pass - # END handle existance + # END handle existence # END handle module # move the module into place if possible cur_path = self.abspath renamed_module = False if module and os.path.exists(cur_path): - os.renames(cur_path, dest_path) + os.renames(cur_path, module_checkout_abspath) renamed_module = True + + if os.path.isfile(os.path.join(module_checkout_abspath, '.git')): + module_abspath = self._module_abspath(self.repo, self.path, self.name) + self._write_git_file_and_module_config(module_checkout_abspath, module_abspath) + # end handle git file rewrite # END move physical module # rename the index entry - have to manipulate the index directly as # git-mv cannot be used on submodules ... yeah + previous_sm_path = self.path try: if configuration: try: ekey = index.entry_key(self.path, 0) entry = index.entries[ekey] del(index.entries[ekey]) - nentry = git.IndexEntry(entry[:3] + (module_path,) + entry[4:]) + nentry = git.IndexEntry(entry[:3] + (module_checkout_path,) + entry[4:]) index.entries[tekey] = nentry except KeyError: raise InvalidGitRepositoryError("Submodule's entry at %r did not exist" % (self.path)) @@ -588,18 +723,23 @@ class Submodule(util.IndexObject, Iterable, Traversable): # update configuration writer = self.config_writer(index=index) # auto-write - writer.set_value('path', module_path) - self.path = module_path + writer.set_value('path', module_checkout_path) + self.path = module_checkout_path writer.release() del(writer) # END handle configuration flag except Exception: if renamed_module: - os.renames(dest_path, cur_path) + os.renames(module_checkout_abspath, cur_path) # END undo module renaming raise # END handle undo rename + # Auto-rename submodule if it's name was 'default', that is, the checkout directory + if previous_sm_path == self.name: + self.rename(module_checkout_path) + # end + return self @unbare_repo @@ -607,7 +747,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): """Remove this submodule from the repository. This will remove our entry from the .gitmodules file and the entry in the .git/config file. - :param module: If True, the module we point to will be deleted + :param module: If True, the module checkout we point to will be deleted as well. If the module is currently on a commit which is not part of any branch in the remote, if the currently checked out branch working tree, or untracked files, @@ -626,15 +766,32 @@ class Submodule(util.IndexObject, Iterable, Traversable): we would usually throw :return: self :note: doesn't work in bare repositories + :note: doesn't work atomically, as failure to remove any part of the submodule will leave + an inconsistent state :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted :raise OSError: if directories or files could not be removed""" if not (module + configuration): raise ValueError("Need to specify to delete at least the module, or the configuration") - # END handle params - - # DELETE MODULE REPOSITORY - ########################## + # END handle parameters + + # Recursively remove children of this submodule + nc = 0 + for csm in self.children(): + nc += 1 + csm.remove(module, force, configuration, dry_run) + del(csm) + # end + if not dry_run and nc > 0: + # Assure we don't leave the parent repository in a dirty state, and commit our changes + # It's important for recursive, unforced, deletions to work as expected + self.module().index.commit("Removed submodule '%s'" % self.name) + # end handle recursion + + # DELETE REPOSITORY WORKING TREE + ################################ if module and self.module_exists(): + mod = self.module() + git_dir = mod.git_dir if force: # take the fast lane and just delete everything in our module path # TODO: If we run into permission problems, we have a highly inconsistent @@ -654,8 +811,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): # END apply deletion method else: # verify we may delete our module - mod = self.module() - if mod.is_dirty(untracked_files=True): + if mod.is_dirty(index=True, working_tree=True, untracked_files=True): raise InvalidGitRepositoryError( "Cannot delete module at %s with any modifications, unless force is specified" % mod.working_tree_dir) @@ -665,7 +821,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): # NOTE: If the user pulled all the time, the remote heads might # not have been updated, so commits coming from the remote look # as if they come from us. But we stay strictly read-only and - # don't fetch beforhand. + # don't fetch beforehand. for remote in mod.remotes: num_branches_with_new_commits = 0 rrefs = remote.refs @@ -686,12 +842,6 @@ class Submodule(util.IndexObject, Iterable, Traversable): del(remote) # END for each remote - # gently remove all submodule repositories - for sm in self.children(): - sm.remove(module=True, force=False, configuration=False, dry_run=dry_run) - del(sm) - # END for each child-submodule - # finally delete our own submodule if not dry_run: wtd = mod.working_tree_dir @@ -699,46 +849,59 @@ class Submodule(util.IndexObject, Iterable, Traversable): rmtree(wtd) # END delete tree if possible # END handle force + + if not dry_run and os.path.isdir(git_dir): + rmtree(git_dir) + # end handle separate bare repository # END handle module deletion + # void our data not to delay invalid access + if not dry_run: + self._clear_cache() + # DELETE CONFIGURATION ###################### if configuration and not dry_run: # first the index-entry - index = self.repo.index + parent_index = self.repo.index try: - del(index.entries[index.entry_key(self.path, 0)]) + del(parent_index.entries[parent_index.entry_key(self.path, 0)]) except KeyError: pass # END delete entry - index.write() + parent_index.write() # now git config - need the config intact, otherwise we can't query - # inforamtion anymore + # information anymore writer = self.repo.config_writer() writer.remove_section(sm_section(self.name)) writer.release() + writer = self.config_writer() writer.remove_section() writer.release() # END delete configuration - # void our data not to delay invalid access - self._clear_cache() - return self def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. - :param commit: Commit'ish reference pointing at the root_tree - :param check: if True, relatively expensive checks will be performed to verify + :param commit: + Commit'ish reference pointing at the root_tree, or None to always point to the + most recent commit + :param check: + if True, relatively expensive checks will be performed to verify validity of the submodule. :raise ValueError: if the commit's tree didn't contain the .gitmodules blob. - :raise ValueError: if the parent commit didn't store this submodule under the - current path + :raise ValueError: + if the parent commit didn't store this submodule under the current path :return: self""" + if commit is None: + self._parent_commit = None + return self + # end handle None pcommit = self.repo.commit(commit) pctree = pcommit.tree if self.k_modules_file not in pctree: @@ -757,10 +920,15 @@ class Submodule(util.IndexObject, Iterable, Traversable): # END handle checking mode # update our sha, it could have changed - self.binsha = pctree[self.path].binsha + # If check is False, we might see a parent-commit that doens't even contain the submodule anymore. + # in that case, mark our sha as being NULL + try: + self.binsha = pctree[self.path].binsha + except KeyError: + self.binsha = self.NULL_BIN_SHA + # end self._clear_cache() - return self @unbare_repo @@ -784,6 +952,56 @@ class Submodule(util.IndexObject, Iterable, Traversable): writer.config._auto_write = write return writer + @unbare_repo + def rename(self, new_name): + """Rename this submodule + :note: This method takes care of renaming the submodule in various places, such as + + * $parent_git_dir/config + * $working_tree_dir/.gitmodules + * (git >=v1.8.0: move submodule repository to new name) + + As .gitmodules will be changed, you would need to make a commit afterwards. The changed .gitmodules file + will already be added to the index + + :return: this submodule instance + """ + if self.name == new_name: + return self + + # .git/config + pw = self.repo.config_writer() + # As we ourselves didn't write anything about submodules into the parent .git/config, we will not require + # it to exist, and just ignore missing entries + if pw.has_section(sm_section(self.name)): + pw.rename_section(sm_section(self.name), sm_section(new_name)) + # end + pw.release() + + # .gitmodules + cw = self.config_writer(write=True).config + cw.rename_section(sm_section(self.name), sm_section(new_name)) + cw.release() + + self._name = new_name + + # .git/modules + mod = self.module() + if mod.has_separate_working_tree(): + destination_module_abspath = self._module_abspath(self.repo, self.path, new_name) + source_dir = mod.git_dir + # Let's be sure the submodule name is not so obviously tied to a directory + if destination_module_abspath.startswith(mod.git_dir): + tmp_dir = self._module_abspath(self.repo, self.path, os.path.basename(tempfile.mkdtemp())) + os.renames(source_dir, tmp_dir) + source_dir = tmp_dir + # end handle self-containment + os.renames(source_dir, destination_module_abspath) + self._write_git_file_and_module_config(mod.working_tree_dir, destination_module_abspath) + # end move separate git repository + + return self + #} END edit interface #{ Query Interface @@ -794,16 +1012,16 @@ class Submodule(util.IndexObject, Iterable, Traversable): :raise InvalidGitRepositoryError: if a repository was not available. This could also mean that it was not yet initialized""" # late import to workaround circular dependencies - module_path = self.abspath + module_checkout_abspath = self.abspath try: - repo = git.Repo(module_path) + repo = git.Repo(module_checkout_abspath) if repo != self.repo: return repo # END handle repo uninitialized except (InvalidGitRepositoryError, NoSuchPathError): - raise InvalidGitRepositoryError("No valid repository at %s" % self.path) + raise InvalidGitRepositoryError("No valid repository at %s" % module_checkout_abspath) else: - raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_path) + raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_checkout_abspath) # END handle exceptions def module_exists(self): @@ -828,7 +1046,7 @@ class Submodule(util.IndexObject, Iterable, Traversable): if hasattr(self, attr): loc[attr] = getattr(self, attr) # END if we have the attribute cache - except cp.NoSectionError: + except (cp.NoSectionError, ValueError): # on PY3, this can happen apparently ... don't know why this doesn't happen on PY2 pass # END for each attr @@ -878,6 +1096,8 @@ class Submodule(util.IndexObject, Iterable, Traversable): def parent_commit(self): """:return: Commit instance with the tree containing the .gitmodules file :note: will always point to the current head's commit if it was not set explicitly""" + if self._parent_commit is None: + return self.repo.commit() return self._parent_commit @property @@ -949,7 +1169,9 @@ class Submodule(util.IndexObject, Iterable, Traversable): # fill in remaining info - saves time as it doesn't have to be parsed again sm._name = n - sm._parent_commit = pc + if pc != repo.commit(): + sm._parent_commit = pc + # end set only if not most recent ! sm._branch_path = git.Head.to_full_path(b) sm._url = u diff --git a/git/objects/submodule/root.py b/git/objects/submodule/root.py index 8c9afff1..1c863f6f 100644 --- a/git/objects/submodule/root.py +++ b/git/objects/submodule/root.py @@ -61,7 +61,8 @@ class RootModule(Submodule): #{ Interface def update(self, previous_commit=None, recursive=True, force_remove=False, init=True, - to_latest_revision=False, progress=None, dry_run=False): + to_latest_revision=False, progress=None, dry_run=False, force_reset=False, + keep_going=False): """Update the submodules of this repository to the current HEAD commit. This method behaves smartly by determining changes of the path of a submodules repository, next to changes to the to-be-checked-out commit or the branch to be @@ -80,10 +81,20 @@ class RootModule(Submodule): :param init: If we encounter a new module which would need to be initialized, then do it. :param to_latest_revision: If True, instead of checking out the revision pointed to by this submodule's sha, the checked out tracking branch will be merged with the - newest remote branch fetched from the repository's origin + latest remote branch fetched from the repository's origin. + Unless force_reset is specified, a local tracking branch will never be reset into its past, therefore + the remote branch must be in the future for this to have an effect. + :param force_reset: if True, submodules may checkout or reset their branch even if the repository has + pending changes that would be overwritten, or if the local tracking branch is in the future of the + remote tracking branch and would be reset into its past. :param progress: RootUpdateProgress instance or None if no progress should be sent :param dry_run: if True, operations will not actually be performed. Progress messages - will change accordingly to indicate the WOULD DO state of the operation.""" + will change accordingly to indicate the WOULD DO state of the operation. + :param keep_going: if True, we will ignore but log all errors, and keep going recursively. + Unless dry_run is set as well, keep_going could cause subsequent/inherited errors you wouldn't see + otherwise. + In conjunction with dry_run, it can be useful to anticipate all errors when updating submodules + :return: self""" if self.repo.bare: raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") # END handle bare @@ -98,220 +109,221 @@ class RootModule(Submodule): repo = self.repo - # SETUP BASE COMMIT - ################### - cur_commit = repo.head.commit - if previous_commit is None: - try: - previous_commit = repo.commit(repo.head.log_entry(-1).oldhexsha) - if previous_commit.binsha == previous_commit.NULL_BIN_SHA: - raise IndexError - # END handle initial commit - except IndexError: - # in new repositories, there is no previous commit - previous_commit = cur_commit - # END exception handling - else: - previous_commit = repo.commit(previous_commit) # obtain commit object - # END handle previous commit - - psms = self.list_items(repo, parent_commit=previous_commit) - sms = self.list_items(repo) - spsms = set(psms) - ssms = set(sms) - - # HANDLE REMOVALS - ################### - rrsm = (spsms - ssms) - len_rrsm = len(rrsm) - for i, rsm in enumerate(rrsm): - op = REMOVE - if i == 0: - op |= BEGIN - # END handle begin - - # fake it into thinking its at the current commit to allow deletion - # of previous module. Trigger the cache to be updated before that - progress.update(op, i, len_rrsm, prefix + "Removing submodule %r at %s" % (rsm.name, rsm.abspath)) - rsm._parent_commit = repo.head.commit - if not dry_run: - rsm.remove(configuration=False, module=True, force=force_remove) - # END handle dry-run - - if i == len_rrsm - 1: - op |= END - # END handle end - progress.update(op, i, len_rrsm, prefix + "Done removing submodule %r" % rsm.name) - # END for each removed submodule - - # HANDLE PATH RENAMES - ##################### - # url changes + branch changes - csms = (spsms & ssms) - len_csms = len(csms) - for i, csm in enumerate(csms): - psm = psms[csm.name] - sm = sms[csm.name] - - # PATH CHANGES - ############## - if sm.path != psm.path and psm.module_exists(): - progress.update(BEGIN | PATHCHANGE, i, len_csms, prefix + - "Moving repository of submodule %r from %s to %s" % (sm.name, psm.abspath, sm.abspath)) - # move the module to the new path - if not dry_run: - psm.move(sm.path, module=True, configuration=False) - # END handle dry_run - progress.update( - END | PATHCHANGE, i, len_csms, prefix + "Done moving repository of submodule %r" % sm.name) - # END handle path changes - - if sm.module_exists(): - # HANDLE URL CHANGE - ################### - if sm.url != psm.url: - # Add the new remote, remove the old one - # This way, if the url just changes, the commits will not - # have to be re-retrieved - nn = '__new_origin__' - smm = sm.module() - rmts = smm.remotes - - # don't do anything if we already have the url we search in place - if len([r for r in rmts if r.url == sm.url]) == 0: - progress.update(BEGIN | URLCHANGE, i, len_csms, prefix + - "Changing url of submodule %r from %s to %s" % (sm.name, psm.url, sm.url)) + try: + # SETUP BASE COMMIT + ################### + cur_commit = repo.head.commit + if previous_commit is None: + try: + previous_commit = repo.commit(repo.head.log_entry(-1).oldhexsha) + if previous_commit.binsha == previous_commit.NULL_BIN_SHA: + raise IndexError + # END handle initial commit + except IndexError: + # in new repositories, there is no previous commit + previous_commit = cur_commit + # END exception handling + else: + previous_commit = repo.commit(previous_commit) # obtain commit object + # END handle previous commit + + psms = self.list_items(repo, parent_commit=previous_commit) + sms = self.list_items(repo) + spsms = set(psms) + ssms = set(sms) + + # HANDLE REMOVALS + ################### + rrsm = (spsms - ssms) + len_rrsm = len(rrsm) + + for i, rsm in enumerate(rrsm): + op = REMOVE + if i == 0: + op |= BEGIN + # END handle begin + + # fake it into thinking its at the current commit to allow deletion + # of previous module. Trigger the cache to be updated before that + progress.update(op, i, len_rrsm, prefix + "Removing submodule %r at %s" % (rsm.name, rsm.abspath)) + rsm._parent_commit = repo.head.commit + rsm.remove(configuration=False, module=True, force=force_remove, dry_run=dry_run) + + if i == len_rrsm - 1: + op |= END + # END handle end + progress.update(op, i, len_rrsm, prefix + "Done removing submodule %r" % rsm.name) + # END for each removed submodule + + # HANDLE PATH RENAMES + ##################### + # url changes + branch changes + csms = (spsms & ssms) + len_csms = len(csms) + for i, csm in enumerate(csms): + psm = psms[csm.name] + sm = sms[csm.name] + + # PATH CHANGES + ############## + if sm.path != psm.path and psm.module_exists(): + progress.update(BEGIN | PATHCHANGE, i, len_csms, prefix + + "Moving repository of submodule %r from %s to %s" + % (sm.name, psm.abspath, sm.abspath)) + # move the module to the new path + if not dry_run: + psm.move(sm.path, module=True, configuration=False) + # END handle dry_run + progress.update( + END | PATHCHANGE, i, len_csms, prefix + "Done moving repository of submodule %r" % sm.name) + # END handle path changes + if sm.module_exists(): + # HANDLE URL CHANGE + ################### + if sm.url != psm.url: + # Add the new remote, remove the old one + # This way, if the url just changes, the commits will not + # have to be re-retrieved + nn = '__new_origin__' + smm = sm.module() + rmts = smm.remotes + + # don't do anything if we already have the url we search in place + if len([r for r in rmts if r.url == sm.url]) == 0: + progress.update(BEGIN | URLCHANGE, i, len_csms, prefix + + "Changing url of submodule %r from %s to %s" % (sm.name, psm.url, sm.url)) + + if not dry_run: + assert nn not in [r.name for r in rmts] + smr = smm.create_remote(nn, sm.url) + smr.fetch(progress=progress) + + # If we have a tracking branch, it should be available + # in the new remote as well. + if len([r for r in smr.refs if r.remote_head == sm.branch_name]) == 0: + raise ValueError( + "Submodule branch named %r was not available in new submodule remote at %r" + % (sm.branch_name, sm.url) + ) + # END head is not detached + + # now delete the changed one + rmt_for_deletion = None + for remote in rmts: + if remote.url == psm.url: + rmt_for_deletion = remote + break + # END if urls match + # END for each remote + + # if we didn't find a matching remote, but have exactly one, + # we can safely use this one + if rmt_for_deletion is None: + if len(rmts) == 1: + rmt_for_deletion = rmts[0] + else: + # if we have not found any remote with the original url + # we may not have a name. This is a special case, + # and its okay to fail here + # Alternatively we could just generate a unique name and leave all + # existing ones in place + raise InvalidGitRepositoryError( + "Couldn't find original remote-repo at url %r" % psm.url) + # END handle one single remote + # END handle check we found a remote + + orig_name = rmt_for_deletion.name + smm.delete_remote(rmt_for_deletion) + # NOTE: Currently we leave tags from the deleted remotes + # as well as separate tracking branches in the possibly totally + # changed repository ( someone could have changed the url to + # another project ). At some point, one might want to clean + # it up, but the danger is high to remove stuff the user + # has added explicitly + + # rename the new remote back to what it was + smr.rename(orig_name) + + # early on, we verified that the our current tracking branch + # exists in the remote. Now we have to assure that the + # sha we point to is still contained in the new remote + # tracking branch. + smsha = sm.binsha + found = False + rref = smr.refs[self.branch_name] + for c in rref.commit.traverse(): + if c.binsha == smsha: + found = True + break + # END traverse all commits in search for sha + # END for each commit + + if not found: + # adjust our internal binsha to use the one of the remote + # this way, it will be checked out in the next step + # This will change the submodule relative to us, so + # the user will be able to commit the change easily + log.warn("Current sha %s was not contained in the tracking\ + branch at the new remote, setting it the the remote's tracking branch", sm.hexsha) + sm.binsha = rref.commit.binsha + # END reset binsha + + # NOTE: All checkout is performed by the base implementation of update + # END handle dry_run + progress.update( + END | URLCHANGE, i, len_csms, prefix + "Done adjusting url of submodule %r" % (sm.name)) + # END skip remote handling if new url already exists in module + # END handle url + + # HANDLE PATH CHANGES + ##################### + if sm.branch_path != psm.branch_path: + # finally, create a new tracking branch which tracks the + # new remote branch + progress.update(BEGIN | BRANCHCHANGE, i, len_csms, prefix + + "Changing branch of submodule %r from %s to %s" + % (sm.name, psm.branch_path, sm.branch_path)) if not dry_run: - assert nn not in [r.name for r in rmts] - smr = smm.create_remote(nn, sm.url) - smr.fetch(progress=progress) - - # If we have a tracking branch, it should be available - # in the new remote as well. - if len([r for r in smr.refs if r.remote_head == sm.branch_name]) == 0: - raise ValueError( - "Submodule branch named %r was not available in new submodule remote at %r" - % (sm.branch_name, sm.url) - ) - # END head is not detached - - # now delete the changed one - rmt_for_deletion = None - for remote in rmts: - if remote.url == psm.url: - rmt_for_deletion = remote - break - # END if urls match - # END for each remote - - # if we didn't find a matching remote, but have exactly one, - # we can safely use this one - if rmt_for_deletion is None: - if len(rmts) == 1: - rmt_for_deletion = rmts[0] - else: - # if we have not found any remote with the original url - # we may not have a name. This is a special case, - # and its okay to fail here - # Alternatively we could just generate a unique name and leave all - # existing ones in place - raise InvalidGitRepositoryError( - "Couldn't find original remote-repo at url %r" % psm.url) - # END handle one single remote - # END handle check we found a remote - - orig_name = rmt_for_deletion.name - smm.delete_remote(rmt_for_deletion) - # NOTE: Currently we leave tags from the deleted remotes - # as well as separate tracking branches in the possibly totally - # changed repository ( someone could have changed the url to - # another project ). At some point, one might want to clean - # it up, but the danger is high to remove stuff the user - # has added explicitly - - # rename the new remote back to what it was - smr.rename(orig_name) - - # early on, we verified that the our current tracking branch - # exists in the remote. Now we have to assure that the - # sha we point to is still contained in the new remote - # tracking branch. - smsha = sm.binsha - found = False - rref = smr.refs[self.branch_name] - for c in rref.commit.traverse(): - if c.binsha == smsha: - found = True - break - # END traverse all commits in search for sha - # END for each commit - - if not found: - # adjust our internal binsha to use the one of the remote - # this way, it will be checked out in the next step - # This will change the submodule relative to us, so - # the user will be able to commit the change easily - log.warn("Current sha %s was not contained in the tracking\ - branch at the new remote, setting it the the remote's tracking branch", sm.hexsha) - sm.binsha = rref.commit.binsha - # END reset binsha - - # NOTE: All checkout is performed by the base implementation of update + smm = sm.module() + smmr = smm.remotes + # As the branch might not exist yet, we will have to fetch all remotes to be sure ... . + for remote in smmr: + remote.fetch(progress=progress) + # end for each remote + + try: + tbr = git.Head.create(smm, sm.branch_name, logmsg='branch: Created from HEAD') + except OSError: + # ... or reuse the existing one + tbr = git.Head(smm, sm.branch_path) + # END assure tracking branch exists + + tbr.set_tracking_branch(find_first_remote_branch(smmr, sm.branch_name)) + # NOTE: All head-resetting is done in the base implementation of update + # but we will have to checkout the new branch here. As it still points to the currently + # checkout out commit, we don't do any harm. + # As we don't want to update working-tree or index, changing the ref is all there is to do + smm.head.reference = tbr # END handle dry_run - progress.update( - END | URLCHANGE, i, len_csms, prefix + "Done adjusting url of submodule %r" % (sm.name)) - # END skip remote handling if new url already exists in module - # END handle url - - # HANDLE PATH CHANGES - ##################### - if sm.branch_path != psm.branch_path: - # finally, create a new tracking branch which tracks the - # new remote branch - progress.update(BEGIN | BRANCHCHANGE, i, len_csms, prefix + - "Changing branch of submodule %r from %s to %s" - % (sm.name, psm.branch_path, sm.branch_path)) - if not dry_run: - smm = sm.module() - smmr = smm.remotes - try: - tbr = git.Head.create(smm, sm.branch_name, logmsg='branch: Created from HEAD') - except OSError: - # ... or reuse the existing one - tbr = git.Head(smm, sm.branch_path) - # END assure tracking branch exists - - tbr.set_tracking_branch(find_first_remote_branch(smmr, sm.branch_name)) - # figure out whether the previous tracking branch contains - # new commits compared to the other one, if not we can - # delete it. - try: - tbr = find_first_remote_branch(smmr, psm.branch_name) - if len(smm.git.cherry(tbr, psm.branch)) == 0: - psm.branch.delete(smm, psm.branch) - # END delete original tracking branch if there are no changes - except InvalidGitRepositoryError: - # ignore it if the previous branch couldn't be found in the - # current remotes, this just means we can't handle it - pass - # END exception handling - - # NOTE: All checkout is done in the base implementation of update - # END handle dry_run - progress.update( - END | BRANCHCHANGE, i, len_csms, prefix + "Done changing branch of submodule %r" % sm.name) - # END handle branch - # END handle - # END for each common submodule + progress.update( + END | BRANCHCHANGE, i, len_csms, prefix + "Done changing branch of submodule %r" % sm.name) + # END handle branch + # END handle + # END for each common submodule + except Exception as err: + if not keep_going: + raise + log.error(str(err)) + # end handle keep_going # FINALLY UPDATE ALL ACTUAL SUBMODULES ###################################### for sm in sms: # update the submodule using the default method sm.update(recursive=False, init=init, to_latest_revision=to_latest_revision, - progress=progress, dry_run=dry_run) + progress=progress, dry_run=dry_run, force=force_reset, keep_going=keep_going) # update recursively depth first - question is which inconsitent # state will be better in case it fails somewhere. Defective branch @@ -322,11 +334,14 @@ class RootModule(Submodule): if sm.module_exists(): type(self)(sm.module()).update(recursive=True, force_remove=force_remove, init=init, to_latest_revision=to_latest_revision, - progress=progress, dry_run=dry_run) + progress=progress, dry_run=dry_run, force_reset=force_reset, + keep_going=keep_going) # END handle dry_run # END handle recursive # END for each submodule to update + return self + def module(self): """:return: the actual repository containing the submodules""" return self.repo diff --git a/git/objects/submodule/util.py b/git/objects/submodule/util.py index 5604dec7..8b9873fc 100644 --- a/git/objects/submodule/util.py +++ b/git/objects/submodule/util.py @@ -49,7 +49,7 @@ def find_first_remote_branch(remotes, branch_name): continue # END exception handling # END for remote - raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch_name) + raise InvalidGitRepositoryError("Didn't find remote branch '%r' in any of the given remotes" % branch_name) #} END utilities diff --git a/git/repo/base.py b/git/repo/base.py index 6d9af6d4..ef12473b 100644 --- a/git/repo/base.py +++ b/git/repo/base.py @@ -131,14 +131,18 @@ class Repo(object): # walk up the path to find the .git dir while curpath: + # ABOUT os.path.NORMPATH + # It's important to normalize the paths, as submodules will otherwise initialize their + # repo instances with paths that depend on path-portions that will not exist after being + # removed. It's just cleaner. if is_git_dir(curpath): - self.git_dir = curpath + self.git_dir = os.path.normpath(curpath) self._working_tree_dir = os.path.dirname(self.git_dir) break gitpath = find_git_dir(join(curpath, '.git')) if gitpath is not None: - self.git_dir = gitpath + self.git_dir = os.path.normpath(gitpath) self._working_tree_dir = curpath break @@ -864,6 +868,16 @@ class Repo(object): self.git.archive(treeish, *path, **kwargs) return self + def has_separate_working_tree(self): + """ + :return: True if our git_dir is not at the root of our working_tree_dir, but a .git file with a + platform agnositic symbolic link. Our git_dir will be whereever the .git file points to + :note: bare repositories will always return False here + """ + if self.bare: + return False + return os.path.isfile(os.path.join(self.working_tree_dir, '.git')) + rev_parse = rev_parse def __repr__(self): diff --git a/git/repo/fun.py b/git/repo/fun.py index 1ee11ffc..2321dbc8 100644 --- a/git/repo/fun.py +++ b/git/repo/fun.py @@ -26,6 +26,7 @@ __all__ = ('rev_parse', 'is_git_dir', 'touch', 'find_git_dir', 'name_to_object', def touch(filename): fp = open(filename, "ab") fp.close() + return filename def is_git_dir(d): @@ -296,7 +297,7 @@ def rev_parse(repo, rev): raise ValueError("Invalid token: %r" % token) # END end handle tag except (IndexError, AttributeError): - raise BadObject("Invalid Revision in %s" % rev) + raise BadName("Invalid revision spec '%s' - not enough parent commits to reach '%s%i'" % (rev, token, num)) # END exception handling # END parse loop diff --git a/git/test/test_config.py b/git/test/test_config.py index 9a44d9e3..fc2b87b6 100644 --- a/git/test/test_config.py +++ b/git/test/test_config.py @@ -73,6 +73,7 @@ class TestBase(TestCase): assert r_config.has_section(sname) assert r_config.has_option(sname, oname) assert r_config.get(sname, oname) == val + w_config.release() # END for each filename def test_multi_line_config(self): @@ -193,3 +194,16 @@ class TestBase(TestCase): cr = GitConfigParser(fpa, read_only=True) check_test_value(cr, tv) cr.release() + + def test_rename(self): + file_obj = self._to_memcache(fixture_path('git_config')) + cw = GitConfigParser(file_obj, read_only=False, merge_includes=False) + + self.failUnlessRaises(ValueError, cw.rename_section, "doesntexist", "foo") + self.failUnlessRaises(ValueError, cw.rename_section, "core", "include") + + nn = "bee" + assert cw.rename_section('core', nn) is cw + assert not cw.has_section('core') + assert len(cw.items(nn)) == 4 + cw.release() diff --git a/git/test/test_docs.py b/git/test/test_docs.py new file mode 100644 index 00000000..5ebae513 --- /dev/null +++ b/git/test/test_docs.py @@ -0,0 +1,26 @@ +#-*-coding:utf-8-*- +# test_git.py +# Copyright (C) 2008, 2009 Michael Trier (mtrier@gmail.com) and contributors +# +# This module is part of GitPython and is released under +# the BSD License: http://www.opensource.org/licenses/bsd-license.php +import os + +import git +from git.test.lib import TestBase +from gitdb.test.lib import with_rw_directory +from git.repo.fun import touch + + +class TestGit(TestBase): + + @with_rw_directory + def test_add_file_and_commit(self, rw_dir): + repo_dir = os.path.join(rw_dir, 'my-new-repo') + file_name = os.path.join(repo_dir, 'new-file') + + r = git.Repo.init(repo_dir) + # This function just creates an empty file ... + touch(file_name) + r.index.add([file_name]) + r.index.commit("initial commit") diff --git a/git/test/test_index.py b/git/test/test_index.py index 2be776cd..0569f40f 100644 --- a/git/test/test_index.py +++ b/git/test/test_index.py @@ -381,12 +381,13 @@ class TestIndex(TestBase): num_entries = len(index.entries) cur_head = rw_repo.head - uname = "Some Developer" + uname = u"Thomas Müller" umail = "sd@company.com" writer = rw_repo.config_writer() writer.set_value("user", "name", uname) writer.set_value("user", "email", umail) writer.release() + assert writer.get_value("user", "name") == uname # remove all of the files, provide a wild mix of paths, BaseIndexEntries, # IndexEntries diff --git a/git/test/test_repo.py b/git/test/test_repo.py index 226c1d26..c32dbdbf 100644 --- a/git/test/test_repo.py +++ b/git/test/test_repo.py @@ -164,6 +164,7 @@ class TestRepo(TestBase): r = Repo.init(path=path, bare=True) assert isinstance(r, Repo) assert r.bare is True + assert not r.has_separate_working_tree() assert os.path.isdir(r.git_dir) self._assert_empty_repo(r) @@ -200,6 +201,7 @@ class TestRepo(TestBase): os.chdir(git_dir_rela) r = Repo.init(bare=False) assert r.bare is False + assert not r.has_separate_working_tree() self._assert_empty_repo(r) finally: @@ -589,8 +591,8 @@ class TestRepo(TestBase): commit = rev_parse(first_rev) assert len(commit.parents) == 0 assert commit.hexsha == first_rev - self.failUnlessRaises(BadObject, rev_parse, first_rev + "~") - self.failUnlessRaises(BadObject, rev_parse, first_rev + "^") + self.failUnlessRaises(BadName, rev_parse, first_rev + "~") + self.failUnlessRaises(BadName, rev_parse, first_rev + "^") # short SHA1 commit2 = rev_parse(first_rev[:20]) diff --git a/git/test/test_submodule.py b/git/test/test_submodule.py index 3e3a97d8..49ab2586 100644 --- a/git/test/test_submodule.py +++ b/git/test/test_submodule.py @@ -1,6 +1,5 @@ # This module is part of GitPython and is released under # the BSD License: http://www.opensource.org/licenses/bsd-license.php -import shutil import sys import os @@ -11,13 +10,18 @@ from git.test.lib import ( with_rw_repo ) from gitdb.test.lib import with_rw_directory -from git.exc import InvalidGitRepositoryError +from git.exc import ( + InvalidGitRepositoryError, + RepositoryDirtyError +) from git.objects.submodule.base import Submodule from git.objects.submodule.root import RootModule, RootUpdateProgress from git.util import to_native_path_linux, join_path_native from git.compat import string_types - -from nose import SkipTest +from git.repo.fun import ( + find_git_dir, + touch +) # Change the configuration if possible to prevent the underlying memory manager # to keep file handles open. On windows we get problems as they are not properly @@ -37,15 +41,15 @@ class TestRootProgress(RootUpdateProgress): """Just prints messages, for now without checking the correctness of the states""" - def update(self, op, index, max_count, message=''): - print(message) + def update(self, op, cur_count, max_count, message=''): + print(op, cur_count, max_count, message) prog = TestRootProgress() class TestSubmodule(TestBase): - k_subm_current = "468cad66ff1f80ddaeee4123c24e4d53a032c00d" + k_subm_current = "c15a6e1923a14bc760851913858a3942a4193cdb" k_subm_changed = "394ed7006ee5dc8bddfd132b64001d5dfc0ffdd3" k_no_subm_tag = "0.1.6" @@ -66,7 +70,7 @@ class TestSubmodule(TestBase): assert sm.path == 'git/ext/gitdb' assert sm.path != sm.name # in our case, we have ids there, which don't equal the path - assert sm.url == 'git://github.com/gitpython-developers/gitdb.git' + assert sm.url.endswith('github.com/gitpython-developers/gitdb.git') assert sm.branch_path == 'refs/heads/master' # the default ... assert sm.branch_name == 'master' assert sm.parent_commit == rwrepo.head.commit @@ -183,7 +187,9 @@ class TestSubmodule(TestBase): assert sm.module().head.ref.tracking_branch() is not None # delete the whole directory and re-initialize - shutil.rmtree(sm.abspath) + assert len(sm.children()) != 0 + # shutil.rmtree(sm.abspath) + sm.remove(force=True, configuration=False) assert len(sm.children()) == 0 # dry-run does nothing sm.update(dry_run=True, recursive=False, progress=prog) @@ -225,12 +231,14 @@ class TestSubmodule(TestBase): # END for each repo to reset # dry run does nothing - sm.update(recursive=True, dry_run=True, progress=prog) + self.failUnlessRaises(RepositoryDirtyError, sm.update, recursive=True, dry_run=True, progress=prog) + sm.update(recursive=True, dry_run=True, progress=prog, force=True) for repo in smods: assert repo.head.commit != repo.head.ref.tracking_branch().commit # END for each repo to check - sm.update(recursive=True, to_latest_revision=True) + self.failUnlessRaises(RepositoryDirtyError, sm.update, recursive=True, to_latest_revision=True) + sm.update(recursive=True, to_latest_revision=True, force=True) for repo in smods: assert repo.head.commit == repo.head.ref.tracking_branch().commit # END for each repo to check @@ -276,9 +284,12 @@ class TestSubmodule(TestBase): # enforce the submodule to be checked out at the right spot as well. csm.update() + assert csm.module_exists() + assert csm.exists() + assert os.path.isdir(csm.module().working_tree_dir) # this would work - assert sm.remove(dry_run=True) is sm + assert sm.remove(force=True, dry_run=True) is sm assert sm.module_exists() sm.remove(force=True, dry_run=True) assert sm.module_exists() @@ -290,25 +301,30 @@ class TestSubmodule(TestBase): # forcibly delete the child repository prev_count = len(sm.children()) - assert csm.remove(force=True) is csm + self.failUnlessRaises(ValueError, csm.remove, force=True) + # We removed sm, which removed all submodules. Howver, the instance we have + # still points to the commit prior to that, where it still existed + csm.set_parent_commit(csm.repo.commit(), check=False) assert not csm.exists() assert not csm.module_exists() - assert len(sm.children()) == prev_count - 1 + assert len(sm.children()) == prev_count # now we have a changed index, as configuration was altered. # fix this sm.module().index.reset(working_tree=True) # now delete only the module of the main submodule assert sm.module_exists() - sm.remove(configuration=False) + sm.remove(configuration=False, force=True) assert sm.exists() assert not sm.module_exists() assert sm.config_reader().get_value('url') # delete the rest + sm_path = sm.path sm.remove() assert not sm.exists() assert not sm.module_exists() + self.failUnlessRaises(ValueError, getattr, sm, 'path') assert len(rwrepo.submodules) == 0 @@ -345,11 +361,11 @@ class TestSubmodule(TestBase): # MOVE MODULE ############# - # invalid inptu + # invalid input self.failUnlessRaises(ValueError, nsm.move, 'doesntmatter', module=False, configuration=False) # renaming to the same path does nothing - assert nsm.move(sm.path) is nsm + assert nsm.move(sm_path) is nsm # rename a module nmp = join_path_native("new", "module", "dir") + "/" # new module path @@ -371,8 +387,6 @@ class TestSubmodule(TestBase): assert nsm.path == pmp assert rwrepo.submodules[0].path == pmp - # TODO lowprio: test remaining exceptions ... for now its okay, the code looks right - # REMOVE 'EM ALL ################ # if a submodule's repo has no remotes, it can't be added without an explicit url @@ -390,7 +404,6 @@ class TestSubmodule(TestBase): @with_rw_repo(k_subm_current) def test_base_rw(self, rwrepo): - raise SkipTest("Disabled as long as it fails and submodule support wasn't overhauled") self._do_base_tests(rwrepo) @with_rw_repo(k_subm_current, bare=True) @@ -477,8 +490,8 @@ class TestSubmodule(TestBase): #================ nsmn = "newsubmodule" nsmp = "submrepo" - async_url = to_native_path_linux(join_path_native(self.rorepo.working_tree_dir, rsmsp[0], rsmsp[1])) - nsm = Submodule.add(rwrepo, nsmn, nsmp, url=async_url) + subrepo_url = to_native_path_linux(join_path_native(self.rorepo.working_tree_dir, rsmsp[0], rsmsp[1])) + nsm = Submodule.add(rwrepo, nsmn, nsmp, url=subrepo_url) csmadded = rwrepo.index.commit("Added submodule").hexsha # make sure we don't keep the repo reference nsm.set_parent_commit(csmadded) assert nsm.module_exists() @@ -504,15 +517,38 @@ class TestSubmodule(TestBase): # an update will remove the module # not in dry_run - rm.update(recursive=False, dry_run=True) + rm.update(recursive=False, dry_run=True, force_remove=True) assert os.path.isdir(smp) - rm.update(recursive=False) + # when removing submodules, we may get new commits as nested submodules are auto-committing changes + # to allow deletions without force, as the index would be dirty otherwise. + # QUESTION: Why does this seem to work in test_git_submodule_compatibility() ? + self.failUnlessRaises(InvalidGitRepositoryError, rm.update, recursive=False, force_remove=False) + rm.update(recursive=False, force_remove=True) assert not os.path.isdir(smp) - # change url - #============= - # to the first repository, this way we have a fast checkout, and a completely different + # 'apply work' to the nested submodule and assure this is not removed/altered during updates + # Need to commit first, otherwise submodule.update wouldn't have a reason to change the head + touch(os.path.join(nsm.module().working_tree_dir, 'new-file')) + # We cannot expect is_dirty to even run as we wouldn't reset a head to the same location + assert nsm.module().head.commit.hexsha == nsm.hexsha + nsm.module().index.add([nsm]) + nsm.module().index.commit("added new file") + rm.update(recursive=False, dry_run=True, progress=prog) # would not change head, and thus doens't fail + # Everything we can do from now on will trigger the 'future' check, so no is_dirty() check will even run + # This would only run if our local branch is in the past and we have uncommitted changes + + prev_commit = nsm.module().head.commit + rm.update(recursive=False, dry_run=False, progress=prog) + assert prev_commit == nsm.module().head.commit, "head shouldn't change, as it is in future of remote branch" + + # this kills the new file + rm.update(recursive=True, progress=prog, force_reset=True) + assert prev_commit != nsm.module().head.commit, "head changed, as the remote url and its commit changed" + + # change url ... + #=============== + # ... to the first repository, this way we have a fast checkout, and a completely different # repository at the different url nsm.set_parent_commit(csmremoved) nsmurl = to_native_path_linux(join_path_native(self.rorepo.working_tree_dir, rsmsp[0])) @@ -522,15 +558,17 @@ class TestSubmodule(TestBase): csmpathchange = rwrepo.index.commit("changed url") nsm.set_parent_commit(csmpathchange) + # Now nsm head is in the future of the tracked remote branch prev_commit = nsm.module().head.commit # dry-run does nothing rm.update(recursive=False, dry_run=True, progress=prog) assert nsm.module().remotes.origin.url != nsmurl - rm.update(recursive=False, progress=prog) + rm.update(recursive=False, progress=prog, force_reset=True) assert nsm.module().remotes.origin.url == nsmurl - # head changed, as the remote url and its commit changed - assert prev_commit != nsm.module().head.commit + assert prev_commit != nsm.module().head.commit, "Should now point to gitdb" + assert len(rwrepo.submodules) == 1 + assert not rwrepo.submodules[0].children()[0].module_exists(), "nested submodule should not be checked out" # add the submodule's changed commit to the index, which is what the # user would do @@ -577,7 +615,7 @@ class TestSubmodule(TestBase): # assure we pull locally only nsmc = nsm.children()[0] writer = nsmc.config_writer() - writer.set_value('url', async_url) + writer.set_value('url', subrepo_url) writer.release() rm.update(recursive=True, progress=prog, dry_run=True) # just to run the code rm.update(recursive=True, progress=prog) @@ -602,11 +640,9 @@ class TestSubmodule(TestBase): @with_rw_directory def test_add_empty_repo(self, rwdir): - parent_dir = os.path.join(rwdir, 'parent') - os.mkdir(parent_dir) empty_repo_dir = os.path.join(rwdir, 'empty-repo') - parent = git.Repo.init(parent_dir) + parent = git.Repo.init(os.path.join(rwdir, 'parent')) git.Repo.init(empty_repo_dir) for checkout_mode in range(2): @@ -614,3 +650,194 @@ class TestSubmodule(TestBase): self.failUnlessRaises(ValueError, parent.create_submodule, name, name, url=empty_repo_dir, no_checkout=checkout_mode and True or False) # end for each checkout mode + + def _submodule_url(self): + return os.path.join(self.rorepo.working_tree_dir, 'git/ext/gitdb/gitdb/ext/smmap') + + @with_rw_directory + def test_git_submodules(self, rwdir): + parent = git.Repo.init(os.path.join(rwdir, 'parent')) + parent.git.submodule('add', self._submodule_url(), 'module') + parent.index.commit("added submodule") + + assert len(parent.submodules) == 1 + sm = parent.submodules[0] + + assert sm.exists() and sm.module_exists() + + clone = git.Repo.clone_from(self._submodule_url(), + os.path.join(parent.working_tree_dir, 'existing-subrepository')) + sm2 = parent.create_submodule('nongit-file-submodule', clone.working_tree_dir) + assert len(parent.submodules) == 2 + + for _ in range(2): + for init in (False, True): + sm.update(init=init) + sm2.update(init=init) + # end for each init state + # end for each iteration + + sm.move(sm.path + '_moved') + sm2.move(sm2.path + '_moved') + + @with_rw_directory + def test_git_submodule_compatibility(self, rwdir): + parent = git.Repo.init(os.path.join(rwdir, 'parent')) + sm_path = 'submodules/intermediate/one' + sm = parent.create_submodule('mymodules/myname', sm_path, url=self._submodule_url()) + parent.index.commit("added submodule") + + def assert_exists(sm, value=True): + assert sm.exists() == value + assert sm.module_exists() == value + # end + + # As git is backwards compatible itself, it would still recognize what we do here ... unless we really + # muss it up. That's the only reason why the test is still here ... . + assert len(parent.git.submodule().splitlines()) == 1 + + module_repo_path = os.path.join(sm.module().working_tree_dir, '.git') + assert module_repo_path.startswith(os.path.join(parent.working_tree_dir, sm_path)) + if not sm._need_gitfile_submodules(parent.git): + assert os.path.isdir(module_repo_path) + assert not sm.module().has_separate_working_tree() + else: + assert os.path.isfile(module_repo_path) + assert sm.module().has_separate_working_tree() + assert find_git_dir(module_repo_path) is not None, "module pointed to by .git file must be valid" + # end verify submodule 'style' + + # test move + new_sm_path = 'submodules/one' + sm.move(new_sm_path) + assert_exists(sm) + + # Add additional submodule level + csm = sm.module().create_submodule('nested-submodule', 'nested-submodule/working-tree', + url=self._submodule_url()) + sm.module().index.commit("added nested submodule") + sm_head_commit = sm.module().commit() + assert_exists(csm) + + # Fails because there are new commits, compared to the remote we cloned from + self.failUnlessRaises(InvalidGitRepositoryError, sm.remove, dry_run=True) + assert_exists(sm) + assert sm.module().commit() == sm_head_commit + assert_exists(csm) + + # rename nested submodule + # This name would move itself one level deeper - needs special handling internally + new_name = csm.name + '/mine' + assert csm.rename(new_name).name == new_name + assert_exists(csm) + assert csm.repo.is_dirty(index=True, working_tree=False), "index must contain changed .gitmodules file" + csm.repo.index.commit("renamed module") + + # keep_going evaluation + rsm = parent.submodule_update() + assert_exists(sm) + assert_exists(csm) + csm_writer = csm.config_writer().set_value('url', 'bar') + csm_writer.release() + csm.repo.index.commit("Have to commit submodule change for algorithm to pick it up") + assert csm.url == 'bar' + + self.failUnlessRaises(Exception, rsm.update, recursive=True, to_latest_revision=True, progress=prog) + assert_exists(csm) + rsm.update(recursive=True, to_latest_revision=True, progress=prog, keep_going=True) + + # remove + sm_module_path = sm.module().git_dir + + for dry_run in (True, False): + sm.remove(dry_run=dry_run, force=True) + assert_exists(sm, value=dry_run) + assert os.path.isdir(sm_module_path) == dry_run + # end for each dry-run mode + + @with_rw_directory + def test_rename(self, rwdir): + parent = git.Repo.init(os.path.join(rwdir, 'parent')) + sm_name = 'mymodules/myname' + sm = parent.create_submodule(sm_name, sm_name, url=self._submodule_url()) + parent.index.commit("Added submodule") + + assert sm.rename(sm_name) is sm and sm.name == sm_name + assert not sm.repo.is_dirty(index=True, working_tree=False, untracked_files=False) + + new_path = 'renamed/myname' + assert sm.move(new_path).name == new_path + + new_sm_name = "shortname" + assert sm.rename(new_sm_name) is sm + assert sm.repo.is_dirty(index=True, working_tree=False, untracked_files=False) + assert sm.exists() + + sm_mod = sm.module() + if os.path.isfile(os.path.join(sm_mod.working_tree_dir, '.git')) == sm._need_gitfile_submodules(parent.git): + assert sm_mod.git_dir.endswith(".git/modules/" + new_sm_name) + # end + + @with_rw_directory + def test_branch_renames(self, rw_dir): + # Setup initial sandbox: + # parent repo has one submodule, which has all the latest changes + source_url = self._submodule_url() + sm_source_repo = git.Repo.clone_from(source_url, os.path.join(rw_dir, 'sm-source')) + parent_repo = git.Repo.init(os.path.join(rw_dir, 'parent')) + sm = parent_repo.create_submodule('mysubmodule', 'subdir/submodule', + sm_source_repo.working_tree_dir, branch='master') + parent_repo.index.commit('added submodule') + assert sm.exists() + + # Create feature branch with one new commit in submodule source + sm_fb = sm_source_repo.create_head('feature') + sm_fb.checkout() + new_file = touch(os.path.join(sm_source_repo.working_tree_dir, 'new-file')) + sm_source_repo.index.add([new_file]) + sm.repo.index.commit("added new file") + + # change designated submodule checkout branch to the new upstream feature branch + smcw = sm.config_writer() + smcw.set_value('branch', sm_fb.name) + smcw.release() + assert sm.repo.is_dirty(index=True, working_tree=False) + sm.repo.index.commit("changed submodule branch to '%s'" % sm_fb) + + # verify submodule update with feature branch that leaves currently checked out branch in it's past + sm_mod = sm.module() + prev_commit = sm_mod.commit() + assert sm_mod.head.ref.name == 'master' + assert parent_repo.submodule_update() + assert sm_mod.head.ref.name == sm_fb.name + assert sm_mod.commit() == prev_commit, "Without to_latest_revision, we don't change the commit" + + assert parent_repo.submodule_update(to_latest_revision=True) + assert sm_mod.head.ref.name == sm_fb.name + assert sm_mod.commit() == sm_fb.commit + + # Create new branch which is in our past, and thus seemingly unrelated to the currently checked out one + # To make it even 'harder', we shall fork and create a new commit + sm_pfb = sm_source_repo.create_head('past-feature', commit='HEAD~20') + sm_pfb.checkout() + sm_source_repo.index.add([touch(os.path.join(sm_source_repo.working_tree_dir, 'new-file'))]) + sm_source_repo.index.commit("new file added, to past of '%r'" % sm_fb) + + # Change designated submodule checkout branch to a new commit in its own past + smcw = sm.config_writer() + smcw.set_value('branch', sm_pfb.path) + smcw.release() + sm.repo.index.commit("changed submodule branch to '%s'" % sm_pfb) + + # Test submodule updates - must fail if submodule is dirty + touch(os.path.join(sm_mod.working_tree_dir, 'unstaged file')) + # This doesn't fail as our own submodule binsha didn't change, and the reset is only triggered if + # to latest revision is True. + parent_repo.submodule_update(to_latest_revision=False) + sm_mod.head.ref.name == sm_pfb.name, "should have been switched to past head" + sm_mod.commit() == sm_fb.commit, "Head wasn't reset" + + self.failUnlessRaises(RepositoryDirtyError, parent_repo.submodule_update, to_latest_revision=True) + parent_repo.submodule_update(to_latest_revision=True, force_reset=True) + assert sm_mod.commit() == sm_pfb.commit, "Now head should have been reset" + assert sm_mod.head.ref.name == sm_pfb.name diff --git a/git/util.py b/git/util.py index 010130cb..06fefcc3 100644 --- a/git/util.py +++ b/git/util.py @@ -17,7 +17,11 @@ import threading # NOTE: Some of the unused imports might be used/imported by others. # Handle once test-cases are back up and running. from .exc import GitCommandError -from .compat import MAXSIZE +from .compat import ( + MAXSIZE, + defenc, + PY3 +) # Most of these are unused here, but are for use by git-python modules so these # don't see gitdb all the time. Flake of course doesn't like it. @@ -247,7 +251,10 @@ class RemoteProgress(object): message = message[:-len(done_token)] # END end message handling - self.update(op_code, cur_count, max_count, message) + self.update(op_code, + cur_count and float(cur_count), + max_count and float(max_count), + message) # END for each sub line return failed_lines @@ -364,7 +371,11 @@ class Actor(object): for attr, evar, cvar, default in (('name', env_name, cls.conf_name, default_name), ('email', env_email, cls.conf_email, default_email)): try: - setattr(actor, attr, os.environ[evar]) + val = os.environ[evar] + if not PY3: + val = val.decode(defenc) + # end assure we don't get 'invalid strings' + setattr(actor, attr, val) except KeyError: if config_reader is not None: setattr(actor, attr, config_reader.get_value('user', cvar, default)) |