Merge branch 'master' into il/vcs-helper
authorJunio C Hamano <gitster@pobox.com>
Mon, 7 Dec 2009 06:40:16 +0000 (22:40 -0800)
committerJunio C Hamano <gitster@pobox.com>
Mon, 7 Dec 2009 06:40:16 +0000 (22:40 -0800)
* master: (334 commits)
bash: update 'git commit' completion
Git 1.6.5.5
Fix diff -B/--dirstat miscounting of newly added contents
reset: improve worktree safety valves
Documentation: Avoid use of xmlto --stringparam
archive: clarify description of path parameter
rerere: don't segfault on failure to open rr-cache
Prepare for 1.6.5.5
gitweb: Describe (possible) gitweb.js minification in gitweb/README
Documentation: xmlto 0.0.18 does not know --stringparam
Fix crasher on encountering SHA1-like non-note in notes tree
t9001: use older Getopt::Long boolean prefix '--no' rather than '--no-'
t4201: use ISO8859-1 rather than ISO-8859-1
Git 1.6.5.4
Unconditionally set man.base.url.for.relative.links
Documentation/Makefile: allow man.base.url.for.relative.link to be set from Make
Git 1.6.6-rc1
git-pull.sh: Fix call to git-merge for new command format
Prepare for 1.6.5.4
merge: do not add standard message when message is given with -m option
...

Conflicts:
Documentation/git-remote-helpers.txt
Makefile
builtin-ls-remote.c
builtin-push.c
transport-helper.c

Signed-off-by: Junio C Hamano <gitster@pobox.com>
21 files changed:
Documentation/config.txt
Documentation/git-remote-helpers.txt
Makefile
builtin-clone.c
builtin-fetch.c
builtin-ls-remote.c
builtin-push.c
configure.ac
git_remote_helpers/.gitignore [new file with mode: 0644]
git_remote_helpers/Makefile [new file with mode: 0644]
git_remote_helpers/__init__.py [new file with mode: 0644]
git_remote_helpers/git/__init__.py [new file with mode: 0644]
git_remote_helpers/git/git.py [new file with mode: 0644]
git_remote_helpers/setup.py [new file with mode: 0644]
git_remote_helpers/util.py [new file with mode: 0644]
remote.c
remote.h
t/test-lib.sh
transport-helper.c
transport.c
transport.h
index a1e36d7e423e01de796ac5f866536280618199d1..76173dbc5deaf1fdbff316b42e0a42ce61f4f32d 100644 (file)
@@ -1461,6 +1461,10 @@ remote.<name>.tagopt::
        Setting this value to \--no-tags disables automatic tag following when
        fetching from remote <name>
 
+remote.<name>.vcs::
+       Setting this to a value <vcs> will cause git to interact with
+       the remote with the git-remote-<vcs> helper.
+
 remotes.<group>::
        The list of remotes which are fetched by "git remote update
        <group>".  See linkgit:git-remote[1].
index 8beb42dbb9952059650d51d2e7a5ca2efe9591fb..5cfdc0cfc553dc01f6f1c7cb185e79449ee0b0a8 100644 (file)
@@ -79,6 +79,17 @@ style string if it contains an LF.
 +
 Supported if the helper has the "push" capability.
 
+'import' <name>::
+       Produces a fast-import stream which imports the current value
+       of the named ref. It may additionally import other refs as
+       needed to construct the history efficiently. The script writes
+       to a helper-specific private namespace. The value of the named
+       ref should be written to a location in this namespace derived
+       by applying the refspecs from the "refspec" capability to the
+       name of the ref.
++
+Supported if the helper has the "import" capability.
+
 If a fatal error occurs, the program writes the error message to
 stderr and exits. The caller should expect that a suitable error
 message has been printed if the child closes the connection without
@@ -99,6 +110,19 @@ CAPABILITIES
 'push'::
        This helper supports the 'push' command.
 
+'import'::
+       This helper supports the 'import' command.
+
+'refspec' 'spec'::
+       When using the import command, expect the source ref to have
+       been written to the destination ref. The earliest applicable
+       refspec takes precedence. For example
+       "refs/heads/*:refs/svn/origin/branches/*" means that, after an
+       "import refs/heads/name", the script has written to
+       refs/svn/origin/branches/name. If this capability is used at
+       all, it must cover all refs reported by the list command; if
+       it is not used, it is effectively "*:*"
+
 REF LIST ATTRIBUTES
 -------------------
 
@@ -107,6 +131,10 @@ REF LIST ATTRIBUTES
        commands.  A helper might chose to acquire the ref list by
        opening a different type of connection to the destination.
 
+'unchanged'::
+       This ref is unchanged since the last import or fetch, although
+       the helper cannot necessarily determine what value that produced.
+
 OPTIONS
 -------
 'option verbosity' <N>::
index 4a1e5bcc4def58a5d8f05a368fea69cde8ddb74e..2ad7e366ef925f99b8a903f3f3a936ee72052127 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -168,6 +168,8 @@ all::
 #
 # Define NO_PERL if you do not want Perl scripts or libraries at all.
 #
+# Define NO_PYTHON if you do not want Python scripts or libraries at all.
+#
 # Define NO_TCLTK if you do not want Tcl/Tk GUI.
 #
 # The TCL_PATH variable governs the location of the Tcl interpreter
@@ -341,6 +343,7 @@ LIB_H =
 LIB_OBJS =
 PROGRAMS =
 SCRIPT_PERL =
+SCRIPT_PYTHON =
 SCRIPT_SH =
 TEST_PROGRAMS =
 
@@ -379,6 +382,7 @@ SCRIPT_PERL += git-svn.perl
 
 SCRIPTS = $(patsubst %.sh,%,$(SCRIPT_SH)) \
          $(patsubst %.perl,%,$(SCRIPT_PERL)) \
+         $(patsubst %.py,%,$(SCRIPT_PYTHON)) \
          git-instaweb
 
 # Empty...
@@ -434,8 +438,12 @@ endif
 ifndef PERL_PATH
        PERL_PATH = /usr/bin/perl
 endif
+ifndef PYTHON_PATH
+       PYTHON_PATH = /usr/bin/python
+endif
 
 export PERL_PATH
+export PYTHON_PATH
 
 LIB_FILE=libgit.a
 XDIFF_LIB=xdiff/lib.a
@@ -1346,6 +1354,10 @@ ifeq ($(PERL_PATH),)
 NO_PERL=NoThanks
 endif
 
+ifeq ($(PYTHON_PATH),)
+NO_PYTHON=NoThanks
+endif
+
 QUIET_SUBDIR0  = +$(MAKE) -C # space to separate -C and subdir
 QUIET_SUBDIR1  =
 
@@ -1393,6 +1405,7 @@ prefix_SQ = $(subst ','\'',$(prefix))
 
 SHELL_PATH_SQ = $(subst ','\'',$(SHELL_PATH))
 PERL_PATH_SQ = $(subst ','\'',$(PERL_PATH))
+PYTHON_PATH_SQ = $(subst ','\'',$(PYTHON_PATH))
 TCLTK_PATH_SQ = $(subst ','\'',$(TCLTK_PATH))
 
 LIBS = $(GITLIBS) $(EXTLIBS)
@@ -1439,6 +1452,9 @@ ifndef NO_TCLTK
 endif
 ifndef NO_PERL
        $(QUIET_SUBDIR0)perl $(QUIET_SUBDIR1) PERL_PATH='$(PERL_PATH_SQ)' prefix='$(prefix_SQ)' all
+endif
+ifndef NO_PYTHON
+       $(QUIET_SUBDIR0)git_remote_helpers $(QUIET_SUBDIR1) PYTHON_PATH='$(PYTHON_PATH_SQ)' prefix='$(prefix_SQ)' all
 endif
        $(QUIET_SUBDIR0)templates $(QUIET_SUBDIR1)
 
@@ -1566,11 +1582,41 @@ $(patsubst %.perl,%,$(SCRIPT_PERL)) git-instaweb: % : unimplemented.sh
        mv $@+ $@
 endif # NO_PERL
 
+
 ifdef JSMIN
 gitweb/gitweb.min.js: gitweb/gitweb.js
        $(QUIET_GEN)$(JSMIN) <$< >$@
 endif # JSMIN
 
+ifndef NO_PYTHON
+$(patsubst %.py,%,$(SCRIPT_PYTHON)): GIT-CFLAGS
+$(patsubst %.py,%,$(SCRIPT_PYTHON)): % : %.py
+       $(QUIET_GEN)$(RM) $@ $@+ && \
+       INSTLIBDIR=`MAKEFLAGS= $(MAKE) -C git_remote_helpers -s \
+               --no-print-directory prefix='$(prefix_SQ)' DESTDIR='$(DESTDIR_SQ)' \
+               instlibdir` && \
+       sed -e '1{' \
+           -e '        s|#!.*python|#!$(PYTHON_PATH_SQ)|' \
+           -e '}' \
+           -e 's|^import sys.*|&; \\\
+                  import os; \\\
+                  sys.path[0] = os.environ.has_key("GITPYTHONLIB") and \\\
+                                os.environ["GITPYTHONLIB"] or \\\
+                                "@@INSTLIBDIR@@"|' \
+           -e 's|@@INSTLIBDIR@@|'"$$INSTLIBDIR"'|g' \
+           $@.py >$@+ && \
+       chmod +x $@+ && \
+       mv $@+ $@
+else # NO_PYTHON
+$(patsubst %.py,%,$(SCRIPT_PYTHON)): % : unimplemented.sh
+       $(QUIET_GEN)$(RM) $@ $@+ && \
+       sed -e '1s|#!.*/sh|#!$(SHELL_PATH_SQ)|' \
+           -e 's|@@REASON@@|NO_PYTHON=$(NO_PYTHON)|g' \
+           unimplemented.sh >$@+ && \
+       chmod +x $@+ && \
+       mv $@+ $@
+endif # NO_PYTHON
+
 configure: configure.ac
        $(QUIET_GEN)$(RM) $@ $<+ && \
        sed -e 's/@@GIT_VERSION@@/$(GIT_VERSION)/g' \
@@ -1796,6 +1842,9 @@ install: all
 ifndef NO_PERL
        $(MAKE) -C perl prefix='$(prefix_SQ)' DESTDIR='$(DESTDIR_SQ)' install
 endif
+ifndef NO_PYTHON
+       $(MAKE) -C git_remote_helpers prefix='$(prefix_SQ)' DESTDIR='$(DESTDIR_SQ)' install
+endif
 ifndef NO_TCLTK
        $(MAKE) -C gitk-git install
        $(MAKE) -C git-gui gitexecdir='$(gitexec_instdir_SQ)' install
@@ -1912,6 +1961,9 @@ clean:
 ifndef NO_PERL
        $(RM) gitweb/gitweb.cgi
        $(MAKE) -C perl clean
+endif
+ifndef NO_PYTHON
+       $(MAKE) -C git_remote_helpers clean
 endif
        $(MAKE) -C templates/ clean
        $(MAKE) -C t/ clean
index caf3025031d83dc860b22c7798dd5f40b6b89ed2..5df8b0f72c8bb5dfa7fa7df1edfec90405031f47 100644 (file)
@@ -362,9 +362,10 @@ int cmd_clone(int argc, const char **argv, const char *prefix)
        const char *repo_name, *repo, *work_tree, *git_dir;
        char *path, *dir;
        int dest_exists;
-       const struct ref *refs, *remote_head, *mapped_refs;
+       const struct ref *refs, *remote_head;
        const struct ref *remote_head_points_at;
        const struct ref *our_head_points_at;
+       struct ref *mapped_refs;
        struct strbuf key = STRBUF_INIT, value = STRBUF_INIT;
        struct strbuf branch_top = STRBUF_INIT, reflog_msg = STRBUF_INIT;
        struct transport *transport = NULL;
index 5b7db616dcf5cc3bb178b2c4dbb989322c6b2374..8654fa7a2dbe2c1ca76a493a4322447de5219b99 100644 (file)
@@ -322,7 +322,10 @@ static int store_updated_refs(const char *raw_url, const char *remote_name,
        if (!fp)
                return error("cannot open %s: %s\n", filename, strerror(errno));
 
-       url = transport_anonymize_url(raw_url);
+       if (raw_url)
+               url = transport_anonymize_url(raw_url);
+       else
+               url = xstrdup("foreign");
        for (rm = ref_map; rm; rm = rm->next) {
                struct ref *ref = NULL;
 
@@ -819,7 +822,7 @@ static int fetch_one(struct remote *remote, int argc, const char **argv)
        if (!remote)
                die("Where do you want to fetch from today?");
 
-       transport = transport_get(remote, remote->url[0]);
+       transport = transport_get(remote, NULL);
        if (verbosity >= 2)
                transport->verbose = verbosity <= 3 ? verbosity : 3;
        if (verbosity < 0)
index b5bad0c184fc1ebc49759f211781ecd4031fd027..70f5622d9d49aae4080b38ee4487cc6e403a9d2c 100644 (file)
@@ -89,7 +89,7 @@ int cmd_ls_remote(int argc, const char **argv, const char *prefix)
        remote = remote_get(dest);
        if (!remote->url_nr)
                die("remote %s has no configured URL", dest);
-       transport = transport_get(remote, remote->url[0]);
+       transport = transport_get(remote, NULL);
        if (uploadpack != NULL)
                transport_set_option(transport, TRANS_OPT_UPLOADPACK, uploadpack);
 
index 356d7c1fd3a6fae9558e93eb4f240242a60da368..dcfb53f1884d2648993214b4378afbd1f83760b0 100644 (file)
@@ -87,6 +87,37 @@ static void setup_default_push_refspecs(void)
        }
 }
 
+static int push_with_options(struct transport *transport, int flags)
+{
+       int err;
+       int nonfastforward;
+       if (receivepack)
+               transport_set_option(transport,
+                                    TRANS_OPT_RECEIVEPACK, receivepack);
+       if (thin)
+               transport_set_option(transport, TRANS_OPT_THIN, "yes");
+
+       if (flags & TRANSPORT_PUSH_VERBOSE)
+               fprintf(stderr, "Pushing to %s\n", transport->url);
+       err = transport_push(transport, refspec_nr, refspec, flags,
+                            &nonfastforward);
+       if (err != 0)
+               error("failed to push some refs to '%s'", transport->url);
+
+       err |= transport_disconnect(transport);
+
+       if (!err)
+               return 0;
+
+       if (nonfastforward && advice_push_nonfastforward) {
+               printf("To prevent you from losing history, non-fast-forward updates were rejected\n"
+                      "Merge the remote changes before pushing again.  See the 'non-fast-forward'\n"
+                      "section of 'git push --help' for details.\n");
+       }
+
+       return 1;
+}
+
 static int do_push(const char *repo, int flags)
 {
        int i, errs;
@@ -135,33 +166,19 @@ static int do_push(const char *repo, int flags)
                url = remote->url;
                url_nr = remote->url_nr;
        }
-       for (i = 0; i < url_nr; i++) {
-               struct transport *transport =
-                       transport_get(remote, url[i]);
-               int err;
-               int nonfastforward;
-               if (receivepack)
-                       transport_set_option(transport,
-                                            TRANS_OPT_RECEIVEPACK, receivepack);
-               if (thin)
-                       transport_set_option(transport, TRANS_OPT_THIN, "yes");
-
-               if (flags & TRANSPORT_PUSH_VERBOSE)
-                       fprintf(stderr, "Pushing to %s\n", url[i]);
-               err = transport_push(transport, refspec_nr, refspec, flags,
-                                    &nonfastforward);
-               err |= transport_disconnect(transport);
-
-               if (!err)
-                       continue;
-
-               error("failed to push some refs to '%s'", url[i]);
-               if (nonfastforward && advice_push_nonfastforward) {
-                       printf("To prevent you from losing history, non-fast-forward updates were rejected\n"
-                              "Merge the remote changes before pushing again.  See the 'non-fast-forward'\n"
-                              "section of 'git push --help' for details.\n");
+       if (url_nr) {
+               for (i = 0; i < url_nr; i++) {
+                       struct transport *transport =
+                               transport_get(remote, url[i]);
+                       if (push_with_options(transport, flags))
+                               errs++;
                }
-               errs++;
+       } else {
+               struct transport *transport =
+                       transport_get(remote, NULL);
+
+               if (push_with_options(transport, flags))
+                       errs++;
        }
        return !!errs;
 }
index 4625b8672bf5b0dcc977737c2c5086c5c1aa54a6..78345ebb60240e6642f319467defe3e07d4aa263 100644 (file)
@@ -276,6 +276,9 @@ GIT_ARG_SET_PATH(shell)
 # Define PERL_PATH to provide path to Perl.
 GIT_ARG_SET_PATH(perl)
 #
+# Define PYTHON_PATH to provide path to Python.
+GIT_ARG_SET_PATH(python)
+#
 # Define ZLIB_PATH to provide path to zlib.
 GIT_ARG_SET_PATH(zlib)
 #
diff --git a/git_remote_helpers/.gitignore b/git_remote_helpers/.gitignore
new file mode 100644 (file)
index 0000000..2247d5f
--- /dev/null
@@ -0,0 +1,2 @@
+/build
+/dist
diff --git a/git_remote_helpers/Makefile b/git_remote_helpers/Makefile
new file mode 100644 (file)
index 0000000..c62dfd0
--- /dev/null
@@ -0,0 +1,35 @@
+#
+# Makefile for the git_remote_helpers python support modules
+#
+pysetupfile:=setup.py
+
+# Shell quote (do not use $(call) to accommodate ancient setups);
+DESTDIR_SQ = $(subst ','\'',$(DESTDIR))
+
+ifndef PYTHON_PATH
+       PYTHON_PATH = /usr/bin/python
+endif
+ifndef prefix
+       prefix = $(HOME)
+endif
+ifndef V
+       QUIET = @
+       QUIETSETUP = --quiet
+endif
+
+PYLIBDIR=$(shell $(PYTHON_PATH) -c \
+        "import sys; \
+        print 'lib/python%i.%i/site-packages' % sys.version_info[:2]")
+
+all: $(pysetupfile)
+       $(QUIET)$(PYTHON_PATH) $(pysetupfile) $(QUIETSETUP) build
+
+install: $(pysetupfile)
+       $(PYTHON_PATH) $(pysetupfile) install --prefix $(DESTDIR_SQ)$(prefix)
+
+instlibdir: $(pysetupfile)
+       @echo "$(DESTDIR_SQ)$(prefix)/$(PYLIBDIR)"
+
+clean:
+       $(QUIET)$(PYTHON_PATH) $(pysetupfile) $(QUIETSETUP) clean -a
+       $(RM) *.pyo *.pyc
diff --git a/git_remote_helpers/__init__.py b/git_remote_helpers/__init__.py
new file mode 100644 (file)
index 0000000..00f69cb
--- /dev/null
@@ -0,0 +1,16 @@
+#!/usr/bin/env python
+
+"""Support library package for git remote helpers.
+
+Git remote helpers are helper commands that interfaces with a non-git
+repository to provide automatic import of non-git history into a Git
+repository.
+
+This package provides the support library needed by these helpers..
+The following modules are included:
+
+- git.git - Interaction with Git repositories
+
+- util - General utility functionality use by the other modules in
+         this package, and also used directly by the helpers.
+"""
diff --git a/git_remote_helpers/git/__init__.py b/git_remote_helpers/git/__init__.py
new file mode 100644 (file)
index 0000000..e69de29
diff --git a/git_remote_helpers/git/git.py b/git_remote_helpers/git/git.py
new file mode 100644 (file)
index 0000000..a383e6c
--- /dev/null
@@ -0,0 +1,678 @@
+#!/usr/bin/env python
+
+"""Functionality for interacting with Git repositories.
+
+This module provides classes for interfacing with a Git repository.
+"""
+
+import os
+import re
+import time
+from binascii import hexlify
+from cStringIO import StringIO
+import unittest
+
+from git_remote_helpers.util import debug, error, die, start_command, run_command
+
+
+def get_git_dir ():
+    """Return the path to the GIT_DIR for this repo."""
+    args = ("git", "rev-parse", "--git-dir")
+    exit_code, output, errors = run_command(args)
+    if exit_code:
+        die("Failed to retrieve git dir")
+    assert not errors
+    return output.strip()
+
+
+def parse_git_config ():
+    """Return a dict containing the parsed version of 'git config -l'."""
+    exit_code, output, errors = run_command(("git", "config", "-z", "-l"))
+    if exit_code:
+        die("Failed to retrieve git configuration")
+    assert not errors
+    return dict([e.split('\n', 1) for e in output.split("\0") if e])
+
+
+def git_config_bool (value):
+    """Convert the given git config string value to True or False.
+
+    Raise ValueError if the given string was not recognized as a
+    boolean value.
+
+    """
+    norm_value = str(value).strip().lower()
+    if norm_value in ("true", "1", "yes", "on", ""):
+        return True
+    if norm_value in ("false", "0", "no", "off", "none"):
+        return False
+    raise ValueError("Failed to parse '%s' into a boolean value" % (value))
+
+
+def valid_git_ref (ref_name):
+    """Return True iff the given ref name is a valid git ref name."""
+    # The following is a reimplementation of the git check-ref-format
+    # command.  The rules were derived from the git check-ref-format(1)
+    # manual page.  This code should be replaced by a call to
+    # check_ref_format() in the git library, when such is available.
+    if ref_name.endswith('/') or \
+       ref_name.startswith('.') or \
+       ref_name.count('/.') or \
+       ref_name.count('..') or \
+       ref_name.endswith('.lock'):
+        return False
+    for c in ref_name:
+        if ord(c) < 0x20 or ord(c) == 0x7f or c in " ~^:?*[":
+            return False
+    return True
+
+
+class GitObjectFetcher(object):
+
+    """Provide parsed access to 'git cat-file --batch'.
+
+    This provides a read-only interface to the Git object database.
+
+    """
+
+    def __init__ (self):
+        """Initiate a 'git cat-file --batch' session."""
+        self.queue = []  # List of object names to be submitted
+        self.in_transit = None  # Object name currently in transit
+
+        # 'git cat-file --batch' produces binary output which is likely
+        # to be corrupted by the default "rU"-mode pipe opened by
+        # start_command.  (Mode == "rU" does universal new-line
+        # conversion, which mangles carriage returns.) Therefore, we
+        # open an explicitly binary-safe pipe for transferring the
+        # output from 'git cat-file --batch'.
+        pipe_r_fd, pipe_w_fd = os.pipe()
+        pipe_r = os.fdopen(pipe_r_fd, "rb")
+        pipe_w = os.fdopen(pipe_w_fd, "wb")
+        self.proc = start_command(("git", "cat-file", "--batch"),
+                                  stdout = pipe_w)
+        self.f = pipe_r
+
+    def __del__ (self):
+        """Verify completed communication with 'git cat-file --batch'."""
+        assert not self.queue
+        assert self.in_transit is None
+        self.proc.stdin.close()
+        assert self.proc.wait() == 0  # Zero exit code
+        assert self.f.read() == ""  # No remaining output
+
+    def _submit_next_object (self):
+        """Submit queue items to the 'git cat-file --batch' process.
+
+        If there are items in the queue, and there is currently no item
+        currently in 'transit', then pop the first item off the queue,
+        and submit it.
+
+        """
+        if self.queue and self.in_transit is None:
+            self.in_transit = self.queue.pop(0)
+            print >> self.proc.stdin, self.in_transit[0]
+
+    def push (self, obj, callback):
+        """Push the given object name onto the queue.
+
+        The given callback function will at some point in the future
+        be called exactly once with the following arguments:
+        - self - this GitObjectFetcher instance
+        - obj  - the object name provided to push()
+        - sha1 - the SHA1 of the object, if 'None' obj is missing
+        - t    - the type of the object (tag/commit/tree/blob)
+        - size - the size of the object in bytes
+        - data - the object contents
+
+        """
+        self.queue.append((obj, callback))
+        self._submit_next_object()  # (Re)start queue processing
+
+    def process_next_entry (self):
+        """Read the next entry off the queue and invoke callback."""
+        obj, cb = self.in_transit
+        self.in_transit = None
+        header = self.f.readline()
+        if header == "%s missing\n" % (obj):
+            cb(self, obj, None, None, None, None)
+            return
+        sha1, t, size = header.split(" ")
+        assert len(sha1) == 40
+        assert t in ("tag", "commit", "tree", "blob")
+        assert size.endswith("\n")
+        size = int(size.strip())
+        data = self.f.read(size)
+        assert self.f.read(1) == "\n"
+        cb(self, obj, sha1, t, size, data)
+        self._submit_next_object()
+
+    def process (self):
+        """Process the current queue until empty."""
+        while self.in_transit is not None:
+            self.process_next_entry()
+
+    # High-level convenience methods:
+
+    def get_sha1 (self, objspec):
+        """Return the SHA1 of the object specified by 'objspec'.
+
+        Return None if 'objspec' does not specify an existing object.
+
+        """
+        class _ObjHandler(object):
+            """Helper class for getting the returned SHA1."""
+            def __init__ (self, parser):
+                self.parser = parser
+                self.sha1 = None
+
+            def __call__ (self, parser, obj, sha1, t, size, data):
+                # FIXME: Many unused arguments. Could this be cheaper?
+                assert parser == self.parser
+                self.sha1 = sha1
+
+        handler = _ObjHandler(self)
+        self.push(objspec, handler)
+        self.process()
+        return handler.sha1
+
+    def open_obj (self, objspec):
+        """Return a file object wrapping the contents of a named object.
+
+        The caller is responsible for calling .close() on the returned
+        file object.
+
+        Raise KeyError if 'objspec' does not exist in the repo.
+
+        """
+        class _ObjHandler(object):
+            """Helper class for parsing the returned git object."""
+            def __init__ (self, parser):
+                """Set up helper."""
+                self.parser = parser
+                self.contents = StringIO()
+                self.err = None
+
+            def __call__ (self, parser, obj, sha1, t, size, data):
+                """Git object callback (see GitObjectFetcher documentation)."""
+                assert parser == self.parser
+                if not sha1:  # Missing object
+                    self.err = "Missing object '%s'" % obj
+                else:
+                    assert size == len(data)
+                    self.contents.write(data)
+
+        handler = _ObjHandler(self)
+        self.push(objspec, handler)
+        self.process()
+        if handler.err:
+            raise KeyError(handler.err)
+        handler.contents.seek(0)
+        return handler.contents
+
+    def walk_tree (self, tree_objspec, callback, prefix = ""):
+        """Recursively walk the given Git tree object.
+
+        Recursively walk all subtrees of the given tree object, and
+        invoke the given callback passing three arguments:
+        (path, mode, data) with the path, permission bits, and contents
+        of all the blobs found in the entire tree structure.
+
+        """
+        class _ObjHandler(object):
+            """Helper class for walking a git tree structure."""
+            def __init__ (self, parser, cb, path, mode = None):
+                """Set up helper."""
+                self.parser = parser
+                self.cb = cb
+                self.path = path
+                self.mode = mode
+                self.err = None
+
+            def parse_tree (self, treedata):
+                """Parse tree object data, yield tree entries.
+
+                Each tree entry is a 3-tuple (mode, sha1, path)
+
+                self.path is prepended to all paths yielded
+                from this method.
+
+                """
+                while treedata:
+                    mode = int(treedata[:6], 10)
+                    # Turn 100xxx into xxx
+                    if mode > 100000:
+                        mode -= 100000
+                    assert treedata[6] == " "
+                    i = treedata.find("\0", 7)
+                    assert i > 0
+                    path = treedata[7:i]
+                    sha1 = hexlify(treedata[i + 1: i + 21])
+                    yield (mode, sha1, self.path + path)
+                    treedata = treedata[i + 21:]
+
+            def __call__ (self, parser, obj, sha1, t, size, data):
+                """Git object callback (see GitObjectFetcher documentation)."""
+                assert parser == self.parser
+                if not sha1:  # Missing object
+                    self.err = "Missing object '%s'" % (obj)
+                    return
+                assert size == len(data)
+                if t == "tree":
+                    if self.path:
+                        self.path += "/"
+                    # Recurse into all blobs and subtrees
+                    for m, s, p in self.parse_tree(data):
+                        parser.push(s,
+                                    self.__class__(self.parser, self.cb, p, m))
+                elif t == "blob":
+                    self.cb(self.path, self.mode, data)
+                else:
+                    raise ValueError("Unknown object type '%s'" % (t))
+
+        self.push(tree_objspec, _ObjHandler(self, callback, prefix))
+        self.process()
+
+
+class GitRefMap(object):
+
+    """Map Git ref names to the Git object names they currently point to.
+
+    Behaves like a dictionary of Git ref names -> Git object names.
+
+    """
+
+    def __init__ (self, obj_fetcher):
+        """Create a new Git ref -> object map."""
+        self.obj_fetcher = obj_fetcher
+        self._cache = {}  # dict: refname -> objname
+
+    def _load (self, ref):
+        """Retrieve the object currently bound to the given ref.
+
+        The name of the object pointed to by the given ref is stored
+        into this mapping, and also returned.
+
+        """
+        if ref not in self._cache:
+            self._cache[ref] = self.obj_fetcher.get_sha1(ref)
+        return self._cache[ref]
+
+    def __contains__ (self, refname):
+        """Return True if the given refname is present in this cache."""
+        return bool(self._load(refname))
+
+    def __getitem__ (self, refname):
+        """Return the git object name pointed to by the given refname."""
+        commit = self._load(refname)
+        if commit is None:
+            raise KeyError("Unknown ref '%s'" % (refname))
+        return commit
+
+    def get (self, refname, default = None):
+        """Return the git object name pointed to by the given refname."""
+        commit = self._load(refname)
+        if commit is None:
+            return default
+        return commit
+
+
+class GitFICommit(object):
+
+    """Encapsulate the data in a Git fast-import commit command."""
+
+    SHA1RE = re.compile(r'^[0-9a-f]{40}$')
+
+    @classmethod
+    def parse_mode (cls, mode):
+        """Verify the given git file mode, and return it as a string."""
+        assert mode in (644, 755, 100644, 100755, 120000)
+        return "%i" % (mode)
+
+    @classmethod
+    def parse_objname (cls, objname):
+        """Return the given object name (or mark number) as a string."""
+        if isinstance(objname, int):  # Object name is a mark number
+            assert objname > 0
+            return ":%i" % (objname)
+
+        # No existence check is done, only checks for valid format
+        assert cls.SHA1RE.match(objname)  # Object name is valid SHA1
+        return objname
+
+    @classmethod
+    def quote_path (cls, path):
+        """Return a quoted version of the given path."""
+        path = path.replace("\\", "\\\\")
+        path = path.replace("\n", "\\n")
+        path = path.replace('"', '\\"')
+        return '"%s"' % (path)
+
+    @classmethod
+    def parse_path (cls, path):
+        """Verify that the given path is valid, and quote it, if needed."""
+        assert not isinstance(path, int)  # Cannot be a mark number
+
+        # These checks verify the rules on the fast-import man page
+        assert not path.count("//")
+        assert not path.endswith("/")
+        assert not path.startswith("/")
+        assert not path.count("/./")
+        assert not path.count("/../")
+        assert not path.endswith("/.")
+        assert not path.endswith("/..")
+        assert not path.startswith("./")
+        assert not path.startswith("../")
+
+        if path.count('"') + path.count('\n') + path.count('\\'):
+            return cls.quote_path(path)
+        return path
+
+    def __init__ (self, name, email, timestamp, timezone, message):
+        """Create a new Git fast-import commit, with the given metadata."""
+        self.name = name
+        self.email = email
+        self.timestamp = timestamp
+        self.timezone = timezone
+        self.message = message
+        self.pathops = []  # List of path operations in this commit
+
+    def modify (self, mode, blobname, path):
+        """Add a file modification to this Git fast-import commit."""
+        self.pathops.append(("M",
+                             self.parse_mode(mode),
+                             self.parse_objname(blobname),
+                             self.parse_path(path)))
+
+    def delete (self, path):
+        """Add a file deletion to this Git fast-import commit."""
+        self.pathops.append(("D", self.parse_path(path)))
+
+    def copy (self, path, newpath):
+        """Add a file copy to this Git fast-import commit."""
+        self.pathops.append(("C",
+                             self.parse_path(path),
+                             self.parse_path(newpath)))
+
+    def rename (self, path, newpath):
+        """Add a file rename to this Git fast-import commit."""
+        self.pathops.append(("R",
+                             self.parse_path(path),
+                             self.parse_path(newpath)))
+
+    def note (self, blobname, commit):
+        """Add a note object to this Git fast-import commit."""
+        self.pathops.append(("N",
+                             self.parse_objname(blobname),
+                             self.parse_objname(commit)))
+
+    def deleteall (self):
+        """Delete all files in this Git fast-import commit."""
+        self.pathops.append("deleteall")
+
+
+class TestGitFICommit(unittest.TestCase):
+
+    """GitFICommit selftests."""
+
+    def test_basic (self):
+        """GitFICommit basic selftests."""
+
+        def expect_fail (method, data):
+            """Verify that the method(data) raises an AssertionError."""
+            try:
+                method(data)
+            except AssertionError:
+                return
+            raise AssertionError("Failed test for invalid data '%s(%s)'" %
+                                 (method.__name__, repr(data)))
+
+    def test_parse_mode (self):
+        """GitFICommit.parse_mode() selftests."""
+        self.assertEqual(GitFICommit.parse_mode(644), "644")
+        self.assertEqual(GitFICommit.parse_mode(755), "755")
+        self.assertEqual(GitFICommit.parse_mode(100644), "100644")
+        self.assertEqual(GitFICommit.parse_mode(100755), "100755")
+        self.assertEqual(GitFICommit.parse_mode(120000), "120000")
+        self.assertRaises(AssertionError, GitFICommit.parse_mode, 0)
+        self.assertRaises(AssertionError, GitFICommit.parse_mode, 123)
+        self.assertRaises(AssertionError, GitFICommit.parse_mode, 600)
+        self.assertRaises(AssertionError, GitFICommit.parse_mode, "644")
+        self.assertRaises(AssertionError, GitFICommit.parse_mode, "abc")
+
+    def test_parse_objname (self):
+        """GitFICommit.parse_objname() selftests."""
+        self.assertEqual(GitFICommit.parse_objname(1), ":1")
+        self.assertRaises(AssertionError, GitFICommit.parse_objname, 0)
+        self.assertRaises(AssertionError, GitFICommit.parse_objname, -1)
+        self.assertEqual(GitFICommit.parse_objname("0123456789" * 4),
+                         "0123456789" * 4)
+        self.assertEqual(GitFICommit.parse_objname("2468abcdef" * 4),
+                         "2468abcdef" * 4)
+        self.assertRaises(AssertionError, GitFICommit.parse_objname,
+                          "abcdefghij" * 4)
+
+    def test_parse_path (self):
+        """GitFICommit.parse_path() selftests."""
+        self.assertEqual(GitFICommit.parse_path("foo/bar"), "foo/bar")
+        self.assertEqual(GitFICommit.parse_path("path/with\n and \" in it"),
+                         '"path/with\\n and \\" in it"')
+        self.assertRaises(AssertionError, GitFICommit.parse_path, 1)
+        self.assertRaises(AssertionError, GitFICommit.parse_path, 0)
+        self.assertRaises(AssertionError, GitFICommit.parse_path, -1)
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo//bar")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo/bar/")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "/foo/bar")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo/./bar")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo/../bar")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo/bar/.")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "foo/bar/..")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "./foo/bar")
+        self.assertRaises(AssertionError, GitFICommit.parse_path, "../foo/bar")
+
+
+class GitFastImport(object):
+
+    """Encapsulate communication with git fast-import."""
+
+    def __init__ (self, f, obj_fetcher, last_mark = 0):
+        """Set up self to communicate with a fast-import process through f."""
+        self.f = f  # File object where fast-import stream is written
+        self.obj_fetcher = obj_fetcher  # GitObjectFetcher instance
+        self.next_mark = last_mark + 1  # Next mark number
+        self.refs = set()  # Keep track of the refnames we've seen
+
+    def comment (self, s):
+        """Write the given comment in the fast-import stream."""
+        assert "\n" not in s, "Malformed comment: '%s'" % (s)
+        self.f.write("# %s\n" % (s))
+
+    def commit (self, ref, commitdata):
+        """Make a commit on the given ref, with the given GitFICommit.
+
+        Return the mark number identifying this commit.
+
+        """
+        self.f.write("""\
+commit %(ref)s
+mark :%(mark)i
+committer %(name)s <%(email)s> %(timestamp)i %(timezone)s
+data %(msgLength)i
+%(msg)s
+""" % {
+    'ref': ref,
+    'mark': self.next_mark,
+    'name': commitdata.name,
+    'email': commitdata.email,
+    'timestamp': commitdata.timestamp,
+    'timezone': commitdata.timezone,
+    'msgLength': len(commitdata.message),
+    'msg': commitdata.message,
+})
+
+        if ref not in self.refs:
+            self.refs.add(ref)
+            parent = ref + "^0"
+            if self.obj_fetcher.get_sha1(parent):
+                self.f.write("from %s\n" % (parent))
+
+        for op in commitdata.pathops:
+            self.f.write(" ".join(op))
+            self.f.write("\n")
+        self.f.write("\n")
+        retval = self.next_mark
+        self.next_mark += 1
+        return retval
+
+    def blob (self, data):
+        """Import the given blob.
+
+        Return the mark number identifying this blob.
+
+        """
+        self.f.write("blob\nmark :%i\ndata %i\n%s\n" %
+                     (self.next_mark, len(data), data))
+        retval = self.next_mark
+        self.next_mark += 1
+        return retval
+
+    def reset (self, ref, objname):
+        """Reset the given ref to point at the given Git object."""
+        self.f.write("reset %s\nfrom %s\n\n" %
+                     (ref, GitFICommit.parse_objname(objname)))
+        if ref not in self.refs:
+            self.refs.add(ref)
+
+
+class GitNotes(object):
+
+    """Encapsulate access to Git notes.
+
+    Simulates a dictionary of object name (SHA1) -> Git note mappings.
+
+    """
+
+    def __init__ (self, notes_ref, obj_fetcher):
+        """Create a new Git notes interface, bound to the given notes ref."""
+        self.notes_ref = notes_ref
+        self.obj_fetcher = obj_fetcher  # Used to get objects from repo
+        self.imports = []  # list: (objname, note data blob name) tuples
+
+    def __del__ (self):
+        """Verify that self.commit_notes() was called before destruction."""
+        if self.imports:
+            error("Missing call to self.commit_notes().")
+            error("%i notes are not committed!", len(self.imports))
+
+    def _load (self, objname):
+        """Return the note data associated with the given git object.
+
+        The note data is returned in string form. If no note is found
+        for the given object, None is returned.
+
+        """
+        try:
+            f = self.obj_fetcher.open_obj("%s:%s" % (self.notes_ref, objname))
+            ret = f.read()
+            f.close()
+        except KeyError:
+            ret = None
+        return ret
+
+    def __getitem__ (self, objname):
+        """Return the note contents associated with the given object.
+
+        Raise KeyError if given object has no associated note.
+
+        """
+        blobdata = self._load(objname)
+        if blobdata is None:
+            raise KeyError("Object '%s' has no note" % (objname))
+        return blobdata
+
+    def get (self, objname, default = None):
+        """Return the note contents associated with the given object.
+
+        Return given default if given object has no associated note.
+
+        """
+        blobdata = self._load(objname)
+        if blobdata is None:
+            return default
+        return blobdata
+
+    def import_note (self, objname, data, gfi):
+        """Tell git fast-import to store data as a note for objname.
+
+        This method uses the given GitFastImport object to create a
+        blob containing the given note data.  Also an entry mapping the
+        given object name to the created blob is stored until
+        commit_notes() is called.
+
+        Note that this method only works if it is later followed by a
+        call to self.commit_notes() (which produces the note commit
+        that refers to the blob produced here).
+
+        """
+        if not data.endswith("\n"):
+            data += "\n"
+        gfi.comment("Importing note for object %s" % (objname))
+        mark = gfi.blob(data)
+        self.imports.append((objname, mark))
+
+    def commit_notes (self, gfi, author, message):
+        """Produce a git fast-import note commit for the imported notes.
+
+        This method uses the given GitFastImport object to create a
+        commit on the notes ref, introducing the notes previously
+        submitted to import_note().
+
+        """
+        if not self.imports:
+            return
+        commitdata = GitFICommit(author[0], author[1],
+                                 time.time(), "0000", message)
+        for objname, blobname in self.imports:
+            assert isinstance(objname, int) and objname > 0
+            assert isinstance(blobname, int) and blobname > 0
+            commitdata.note(blobname, objname)
+        gfi.commit(self.notes_ref, commitdata)
+        self.imports = []
+
+
+class GitCachedNotes(GitNotes):
+
+    """Encapsulate access to Git notes (cached version).
+
+    Only use this class if no caching is done at a higher level.
+
+    Simulates a dictionary of object name (SHA1) -> Git note mappings.
+
+    """
+
+    def __init__ (self, notes_ref, obj_fetcher):
+        """Set up a caching wrapper around GitNotes."""
+        GitNotes.__init__(self, notes_ref, obj_fetcher)
+        self._cache = {}  # Cache: object name -> note data
+
+    def __del__ (self):
+        """Verify that GitNotes' destructor is called."""
+        GitNotes.__del__(self)
+
+    def _load (self, objname):
+        """Extend GitNotes._load() with a local objname -> note cache."""
+        if objname not in self._cache:
+            self._cache[objname] = GitNotes._load(self, objname)
+        return self._cache[objname]
+
+    def import_note (self, objname, data, gfi):
+        """Extend GitNotes.import_note() with a local objname -> note cache."""
+        if not data.endswith("\n"):
+            data += "\n"
+        assert objname not in self._cache
+        self._cache[objname] = data
+        GitNotes.import_note(self, objname, data, gfi)
+
+
+if __name__ == '__main__':
+    unittest.main()
diff --git a/git_remote_helpers/setup.py b/git_remote_helpers/setup.py
new file mode 100644 (file)
index 0000000..4d434b6
--- /dev/null
@@ -0,0 +1,17 @@
+#!/usr/bin/env python
+
+"""Distutils build/install script for the git_remote_helpers package."""
+
+from distutils.core import setup
+
+setup(
+    name = 'git_remote_helpers',
+    version = '0.1.0',
+    description = 'Git remote helper program for non-git repositories',
+    license = 'GPLv2',
+    author = 'The Git Community',
+    author_email = 'git@vger.kernel.org',
+    url = 'http://www.git-scm.com/',
+    package_dir = {'git_remote_helpers': ''},
+    packages = ['git_remote_helpers', 'git_remote_helpers.git'],
+)
diff --git a/git_remote_helpers/util.py b/git_remote_helpers/util.py
new file mode 100644 (file)
index 0000000..dce83e6
--- /dev/null
@@ -0,0 +1,194 @@
+#!/usr/bin/env python
+
+"""Misc. useful functionality used by the rest of this package.
+
+This module provides common functionality used by the other modules in
+this package.
+
+"""
+
+import sys
+import os
+import subprocess
+
+
+# Whether or not to show debug messages
+DEBUG = False
+
+def notify(msg, *args):
+    """Print a message to stderr."""
+    print >> sys.stderr, msg % args
+
+def debug (msg, *args):
+    """Print a debug message to stderr when DEBUG is enabled."""
+    if DEBUG:
+        print >> sys.stderr, msg % args
+
+def error (msg, *args):
+    """Print an error message to stderr."""
+    print >> sys.stderr, "ERROR:", msg % args
+
+def warn(msg, *args):
+    """Print a warning message to stderr."""
+    print >> sys.stderr, "warning:", msg % args
+
+def die (msg, *args):
+    """Print as error message to stderr and exit the program."""
+    error(msg, *args)
+    sys.exit(1)
+
+
+class ProgressIndicator(object):
+
+    """Simple progress indicator.
+
+    Displayed as a spinning character by default, but can be customized
+    by passing custom messages that overrides the spinning character.
+
+    """
+
+    States = ("|", "/", "-", "\\")
+
+    def __init__ (self, prefix = "", f = sys.stdout):
+        """Create a new ProgressIndicator, bound to the given file object."""
+        self.n = 0  # Simple progress counter
+        self.f = f  # Progress is written to this file object
+        self.prev_len = 0  # Length of previous msg (to be overwritten)
+        self.prefix = prefix  # Prefix prepended to each progress message
+        self.prefix_lens = [] # Stack of prefix string lengths
+
+    def pushprefix (self, prefix):
+        """Append the given prefix onto the prefix stack."""
+        self.prefix_lens.append(len(self.prefix))
+        self.prefix += prefix
+
+    def popprefix (self):
+        """Remove the last prefix from the prefix stack."""
+        prev_len = self.prefix_lens.pop()
+        self.prefix = self.prefix[:prev_len]
+
+    def __call__ (self, msg = None, lf = False):
+        """Indicate progress, possibly with a custom message."""
+        if msg is None:
+            msg = self.States[self.n % len(self.States)]
+        msg = self.prefix + msg
+        print >> self.f, "\r%-*s" % (self.prev_len, msg),
+        self.prev_len = len(msg.expandtabs())
+        if lf:
+            print >> self.f
+            self.prev_len = 0
+        self.n += 1
+
+    def finish (self, msg = "done", noprefix = False):
+        """Finalize progress indication with the given message."""
+        if noprefix:
+            self.prefix = ""
+        self(msg, True)
+
+
+def start_command (args, cwd = None, shell = False, add_env = None,
+                   stdin = subprocess.PIPE, stdout = subprocess.PIPE,
+                   stderr = subprocess.PIPE):
+    """Start the given command, and return a subprocess object.
+
+    This provides a simpler interface to the subprocess module.
+
+    """
+    env = None
+    if add_env is not None:
+        env = os.environ.copy()
+        env.update(add_env)
+    return subprocess.Popen(args, bufsize = 1, stdin = stdin, stdout = stdout,
+                            stderr = stderr, cwd = cwd, shell = shell,
+                            env = env, universal_newlines = True)
+
+
+def run_command (args, cwd = None, shell = False, add_env = None,
+                 flag_error = True):
+    """Run the given command to completion, and return its results.
+
+    This provides a simpler interface to the subprocess module.
+
+    The results are formatted as a 3-tuple: (exit_code, output, errors)
+
+    If flag_error is enabled, Error messages will be produced if the
+    subprocess terminated with a non-zero exit code and/or stderr
+    output.
+
+    The other arguments are passed on to start_command().
+
+    """
+    process = start_command(args, cwd, shell, add_env)
+    (output, errors) = process.communicate()
+    exit_code = process.returncode
+    if flag_error and errors:
+        error("'%s' returned errors:\n---\n%s---", " ".join(args), errors)
+    if flag_error and exit_code:
+        error("'%s' returned exit code %i", " ".join(args), exit_code)
+    return (exit_code, output, errors)
+
+
+def file_reader_method (missing_ok = False):
+    """Decorator for simplifying reading of files.
+
+    If missing_ok is True, a failure to open a file for reading will
+    not raise the usual IOError, but instead the wrapped method will be
+    called with f == None.  The method must in this case properly
+    handle f == None.
+
+    """
+    def _wrap (method):
+        """Teach given method to handle both filenames and file objects.
+
+        The given method must take a file object as its second argument
+        (the first argument being 'self', of course).  This decorator
+        will take a filename given as the second argument and promote
+        it to a file object.
+
+        """
+        def _wrapped_method (self, filename, *args, **kwargs):
+            if isinstance(filename, file):
+                f = filename
+            else:
+                try:
+                    f = open(filename, 'r')
+                except IOError:
+                    if missing_ok:
+                        f = None
+                    else:
+                        raise
+            try:
+                return method(self, f, *args, **kwargs)
+            finally:
+                if not isinstance(filename, file) and f:
+                    f.close()
+        return _wrapped_method
+    return _wrap
+
+
+def file_writer_method (method):
+    """Decorator for simplifying writing of files.
+
+    Enables the given method to handle both filenames and file objects.
+
+    The given method must take a file object as its second argument
+    (the first argument being 'self', of course).  This decorator will
+    take a filename given as the second argument and promote it to a
+    file object.
+
+    """
+    def _new_method (self, filename, *args, **kwargs):
+        if isinstance(filename, file):
+            f = filename
+        else:
+            # Make sure the containing directory exists
+            parent_dir = os.path.dirname(filename)
+            if not os.path.isdir(parent_dir):
+                os.makedirs(parent_dir)
+            f = open(filename, 'w')
+        try:
+            return method(self, f, *args, **kwargs)
+        finally:
+            if not isinstance(filename, file):
+                f.close()
+    return _new_method
index b979a9642b81b456ab92cdf57e8eda944f01f92e..e3afecdb1069a244493e906b5e16a96dc47535d3 100644 (file)
--- a/remote.c
+++ b/remote.c
@@ -53,6 +53,11 @@ static struct rewrites rewrites_push;
 #define BUF_SIZE (2048)
 static char buffer[BUF_SIZE];
 
+static int valid_remote(const struct remote *remote)
+{
+       return (!!remote->url) || (!!remote->foreign_vcs);
+}
+
 static const char *alias_url(const char *url, struct rewrites *r)
 {
        int i, j;
@@ -441,6 +446,8 @@ static int handle_config(const char *key, const char *value, void *cb)
        } else if (!strcmp(subkey, ".proxy")) {
                return git_config_string((const char **)&remote->http_proxy,
                                         key, value);
+       } else if (!strcmp(subkey, ".vcs")) {
+               return git_config_string(&remote->foreign_vcs, key, value);
        }
        return 0;
 }
@@ -668,6 +675,16 @@ static struct refspec *parse_push_refspec(int nr_refspec, const char **refspec)
        return parse_refspec_internal(nr_refspec, refspec, 0, 0);
 }
 
+void free_refspec(int nr_refspec, struct refspec *refspec)
+{
+       int i;
+       for (i = 0; i < nr_refspec; i++) {
+               free(refspec[i].src);
+               free(refspec[i].dst);
+       }
+       free(refspec);
+}
+
 static int valid_remote_nick(const char *name)
 {
        if (!name[0] || is_dot_or_dotdot(name))
@@ -690,14 +707,14 @@ struct remote *remote_get(const char *name)
 
        ret = make_remote(name, 0);
        if (valid_remote_nick(name)) {
-               if (!ret->url)
+               if (!valid_remote(ret))
                        read_remotes_file(ret);
-               if (!ret->url)
+               if (!valid_remote(ret))
                        read_branches_file(ret);
        }
-       if (name_given && !ret->url)
+       if (name_given && !valid_remote(ret))
                add_url_alias(ret, name);
-       if (!ret->url)
+       if (!valid_remote(ret))
                return NULL;
        ret->fetch = parse_fetch_refspec(ret->fetch_refspec_nr, ret->fetch_refspec);
        ret->push = parse_push_refspec(ret->push_refspec_nr, ret->push_refspec);
@@ -810,6 +827,23 @@ static int match_name_with_pattern(const char *key, const char *name,
        return ret;
 }
 
+char *apply_refspecs(struct refspec *refspecs, int nr_refspec,
+                    const char *name)
+{
+       int i;
+       char *ret = NULL;
+       for (i = 0; i < nr_refspec; i++) {
+               struct refspec *refspec = refspecs + i;
+               if (refspec->pattern) {
+                       if (match_name_with_pattern(refspec->src, name,
+                                                   refspec->dst, &ret))
+                               return ret;
+               } else if (!strcmp(refspec->src, name))
+                       return strdup(refspec->dst);
+       }
+       return NULL;
+}
+
 int remote_find_tracking(struct remote *remote, struct refspec *refspec)
 {
        int find_src = refspec->src == NULL;
index d0aba81ead1847e43a971362659abf1c1737c12f..8b7ecf9197f2a9210299c9700cc88a92a796f714 100644 (file)
--- a/remote.h
+++ b/remote.h
@@ -11,6 +11,8 @@ struct remote {
        const char *name;
        int origin;
 
+       const char *foreign_vcs;
+
        const char **url;
        int url_nr;
        int url_alloc;
@@ -89,6 +91,11 @@ void ref_remove_duplicates(struct ref *ref_map);
 int valid_fetch_refspec(const char *refspec);
 struct refspec *parse_fetch_refspec(int nr_refspec, const char **refspec);
 
+void free_refspec(int nr_refspec, struct refspec *refspec);
+
+char *apply_refspecs(struct refspec *refspecs, int nr_refspec,
+                    const char *name);
+
 int match_refs(struct ref *src, struct ref **dst,
               int nr_refspec, const char **refspec, int all);
 
index ec3336aba5a65a468bc6ce71f33a9cca76dbfe0f..4a40520595e9d932640dd603b28c4472c945e63b 100644 (file)
@@ -638,6 +638,15 @@ test -d ../templates/blt || {
        error "You haven't built things yet, have you?"
 }
 
+if test -z "$GIT_TEST_INSTALLED"
+then
+       GITPYTHONLIB="$(pwd)/../git_remote_helpers/build/lib"
+       export GITPYTHONLIB
+       test -d ../git_remote_helpers/build || {
+               error "You haven't built git_remote_helpers yet, have you?"
+       }
+fi
+
 if ! test -x ../test-chmtime; then
        echo >&2 'You need to build test-chmtime:'
        echo >&2 'Run "make test-chmtime" in the source (toplevel) directory'
@@ -729,6 +738,7 @@ case $(uname -s) in
 esac
 
 test -z "$NO_PERL" && test_set_prereq PERL
+test -z "$NO_PYTHON" && test_set_prereq PYTHON
 
 # test whether the filesystem supports symbolic links
 ln -s x y 2>/dev/null && test -h y 2>/dev/null && test_set_prereq SYMLINKS
index 5078c7100f16e9d7ac08aad05e3797ad5b6626bc..11f3d7ec52893dcdee127d8493f1897c6c091b31 100644 (file)
@@ -6,6 +6,7 @@
 #include "diff.h"
 #include "revision.h"
 #include "quote.h"
+#include "remote.h"
 
 struct helper_data
 {
@@ -13,8 +14,12 @@ struct helper_data
        struct child_process *helper;
        FILE *out;
        unsigned fetch : 1,
+               import : 1,
                option : 1,
                push : 1;
+       /* These go from remote name (as in "list") to private name */
+       struct refspec *refspecs;
+       int refspec_nr;
 };
 
 static struct child_process *get_helper(struct transport *transport)
@@ -22,6 +27,9 @@ static struct child_process *get_helper(struct transport *transport)
        struct helper_data *data = transport->data;
        struct strbuf buf = STRBUF_INIT;
        struct child_process *helper;
+       const char **refspecs = NULL;
+       int refspec_nr = 0;
+       int refspec_alloc = 0;
 
        if (data->helper)
                return data->helper;
@@ -55,7 +63,25 @@ static struct child_process *get_helper(struct transport *transport)
                        data->option = 1;
                if (!strcmp(buf.buf, "push"))
                        data->push = 1;
+               if (!strcmp(buf.buf, "import"))
+                       data->import = 1;
+               if (!data->refspecs && !prefixcmp(buf.buf, "refspec ")) {
+                       ALLOC_GROW(refspecs,
+                                  refspec_nr + 1,
+                                  refspec_alloc);
+                       refspecs[refspec_nr++] = strdup(buf.buf + strlen("refspec "));
+               }
+       }
+       if (refspecs) {
+               int i;
+               data->refspec_nr = refspec_nr;
+               data->refspecs = parse_fetch_refspec(refspec_nr, refspecs);
+               for (i = 0; i < refspec_nr; i++) {
+                       free((char *)refspecs[i]);
+               }
+               free(refspecs);
        }
+       strbuf_release(&buf);
        return data->helper;
 }
 
@@ -72,7 +98,6 @@ static int disconnect_helper(struct transport *transport)
                free(data->helper);
                data->helper = NULL;
        }
-       free(data);
        return 0;
 }
 
@@ -154,8 +179,18 @@ static void standard_options(struct transport *t)
        set_helper_option(t, "verbosity", buf);
 }
 
+static int release_helper(struct transport *transport)
+{
+       struct helper_data *data = transport->data;
+       free_refspec(data->refspec_nr, data->refspecs);
+       data->refspecs = NULL;
+       disconnect_helper(transport);
+       free(transport->data);
+       return 0;
+}
+
 static int fetch_with_fetch(struct transport *transport,
-                           int nr_heads, const struct ref **to_fetch)
+                           int nr_heads, struct ref **to_fetch)
 {
        struct helper_data *data = transport->data;
        int i;
@@ -197,8 +232,64 @@ static int fetch_with_fetch(struct transport *transport,
        return 0;
 }
 
+static int get_importer(struct transport *transport, struct child_process *fastimport)
+{
+       struct child_process *helper = get_helper(transport);
+       memset(fastimport, 0, sizeof(*fastimport));
+       fastimport->in = helper->out;
+       fastimport->argv = xcalloc(5, sizeof(*fastimport->argv));
+       fastimport->argv[0] = "fast-import";
+       fastimport->argv[1] = "--quiet";
+
+       fastimport->git_cmd = 1;
+       return start_command(fastimport);
+}
+
+static int fetch_with_import(struct transport *transport,
+                            int nr_heads, struct ref **to_fetch)
+{
+       struct child_process fastimport;
+       struct child_process *helper = get_helper(transport);
+       struct helper_data *data = transport->data;
+       int i;
+       struct ref *posn;
+       struct strbuf buf = STRBUF_INIT;
+
+       if (get_importer(transport, &fastimport))
+               die("Couldn't run fast-import");
+
+       for (i = 0; i < nr_heads; i++) {
+               posn = to_fetch[i];
+               if (posn->status & REF_STATUS_UPTODATE)
+                       continue;
+
+               strbuf_addf(&buf, "import %s\n", posn->name);
+               write_in_full(helper->in, buf.buf, buf.len);
+               strbuf_reset(&buf);
+       }
+       disconnect_helper(transport);
+       finish_command(&fastimport);
+       free(fastimport.argv);
+       fastimport.argv = NULL;
+
+       for (i = 0; i < nr_heads; i++) {
+               char *private;
+               posn = to_fetch[i];
+               if (posn->status & REF_STATUS_UPTODATE)
+                       continue;
+               if (data->refspecs)
+                       private = apply_refspecs(data->refspecs, data->refspec_nr, posn->name);
+               else
+                       private = strdup(posn->name);
+               read_ref(private, posn->old_sha1);
+               free(private);
+       }
+       strbuf_release(&buf);
+       return 0;
+}
+
 static int fetch(struct transport *transport,
-                int nr_heads, const struct ref **to_fetch)
+                int nr_heads, struct ref **to_fetch)
 {
        struct helper_data *data = transport->data;
        int i, count;
@@ -214,6 +305,9 @@ static int fetch(struct transport *transport,
        if (data->fetch)
                return fetch_with_fetch(transport, nr_heads, to_fetch);
 
+       if (data->import)
+               return fetch_with_import(transport, nr_heads, to_fetch);
+
        return -1;
 }
 
@@ -343,6 +437,22 @@ static int push_refs(struct transport *transport,
        return 0;
 }
 
+static int has_attribute(const char *attrs, const char *attr) {
+       int len;
+       if (!attrs)
+               return 0;
+
+       len = strlen(attr);
+       for (;;) {
+               const char *space = strchrnul(attrs, ' ');
+               if (len == space - attrs && !strncmp(attrs, attr, len))
+                       return 1;
+               if (!*space)
+                       return 0;
+               attrs = space + 1;
+       }
+}
+
 static struct ref *get_refs_list(struct transport *transport, int for_push)
 {
        struct helper_data *data = transport->data;
@@ -379,6 +489,12 @@ static struct ref *get_refs_list(struct transport *transport, int for_push)
                        (*tail)->symref = xstrdup(buf.buf + 1);
                else if (buf.buf[0] != '?')
                        get_sha1_hex(buf.buf, (*tail)->old_sha1);
+               if (eon) {
+                       if (has_attribute(eon + 1, "unchanged")) {
+                               (*tail)->status |= REF_STATUS_UPTODATE;
+                               read_ref((*tail)->name, (*tail)->old_sha1);
+                       }
+               }
                tail = &((*tail)->next);
        }
        strbuf_release(&buf);
@@ -399,6 +515,6 @@ int transport_helper_init(struct transport *transport, const char *name)
        transport->get_refs_list = get_refs_list;
        transport->fetch = fetch;
        transport->push_refs = push_refs;
-       transport->disconnect = disconnect_helper;
+       transport->disconnect = release_helper;
        return 0;
 }
index 7362ec09b2cbc6752489286a8280c16d3519f163..3eea836a33a56aaa99eec78bb4850b02888e377b 100644 (file)
@@ -204,7 +204,7 @@ static struct ref *get_refs_via_rsync(struct transport *transport, int for_push)
 }
 
 static int fetch_objs_via_rsync(struct transport *transport,
-                               int nr_objs, const struct ref **to_fetch)
+                               int nr_objs, struct ref **to_fetch)
 {
        struct strbuf buf = STRBUF_INIT;
        struct child_process rsync;
@@ -379,7 +379,7 @@ static struct ref *get_refs_from_bundle(struct transport *transport, int for_pus
 }
 
 static int fetch_refs_from_bundle(struct transport *transport,
-                              int nr_heads, const struct ref **to_fetch)
+                              int nr_heads, struct ref **to_fetch)
 {
        struct bundle_transport_data *data = transport->data;
        return unbundle(&data->header, data->fd);
@@ -457,7 +457,7 @@ static struct ref *get_refs_via_connect(struct transport *transport, int for_pus
 }
 
 static int fetch_refs_via_pack(struct transport *transport,
-                              int nr_heads, const struct ref **to_fetch)
+                              int nr_heads, struct ref **to_fetch)
 {
        struct git_transport_data *data = transport->data;
        char **heads = xmalloc(nr_heads * sizeof(*heads));
@@ -788,8 +788,26 @@ struct transport *transport_get(struct remote *remote, const char *url)
                die("No remote provided to transport_get()");
 
        ret->remote = remote;
+
+       if (!url && remote && remote->url)
+               url = remote->url[0];
        ret->url = url;
 
+       /* maybe it is a foreign URL? */
+       if (url) {
+               const char *p = url;
+
+               while (isalnum(*p))
+                       p++;
+               if (!prefixcmp(p, "::"))
+                       remote->foreign_vcs = xstrndup(url, p - url);
+       }
+
+       if (remote && remote->foreign_vcs) {
+               transport_helper_init(ret, remote->foreign_vcs);
+               return ret;
+       }
+
        if (!prefixcmp(url, "rsync:")) {
                ret->get_refs_list = get_refs_via_rsync;
                ret->fetch = fetch_objs_via_rsync;
@@ -896,16 +914,17 @@ const struct ref *transport_get_remote_refs(struct transport *transport)
        return transport->remote_refs;
 }
 
-int transport_fetch_refs(struct transport *transport, const struct ref *refs)
+int transport_fetch_refs(struct transport *transport, struct ref *refs)
 {
        int rc;
        int nr_heads = 0, nr_alloc = 0, nr_refs = 0;
-       const struct ref **heads = NULL;
-       const struct ref *rm;
+       struct ref **heads = NULL;
+       struct ref *rm;
 
        for (rm = refs; rm; rm = rm->next) {
                nr_refs++;
                if (rm->peer_ref &&
+                   !is_null_sha1(rm->old_sha1) &&
                    !hashcmp(rm->peer_ref->old_sha1, rm->old_sha1))
                        continue;
                ALLOC_GROW(heads, nr_heads + 1, nr_alloc);
index e4e6177e2632b4a764703c11ef0a8033eafea037..9e74406fa203a59920cb24884d7893b45d876a70 100644 (file)
@@ -18,11 +18,48 @@ struct transport {
        int (*set_option)(struct transport *connection, const char *name,
                          const char *value);
 
+       /**
+        * Returns a list of the remote side's refs. In order to allow
+        * the transport to try to share connections, for_push is a
+        * hint as to whether the ultimate operation is a push or a fetch.
+        *
+        * If the transport is able to determine the remote hash for
+        * the ref without a huge amount of effort, it should store it
+        * in the ref's old_sha1 field; otherwise it should be all 0.
+        **/
        struct ref *(*get_refs_list)(struct transport *transport, int for_push);
-       int (*fetch)(struct transport *transport, int refs_nr, const struct ref **refs);
+
+       /**
+        * Fetch the objects for the given refs. Note that this gets
+        * an array, and should ignore the list structure.
+        *
+        * If the transport did not get hashes for refs in
+        * get_refs_list(), it should set the old_sha1 fields in the
+        * provided refs now.
+        **/
+       int (*fetch)(struct transport *transport, int refs_nr, struct ref **refs);
+
+       /**
+        * Push the objects and refs. Send the necessary objects, and
+        * then, for any refs where peer_ref is set and
+        * peer_ref->new_sha1 is different from old_sha1, tell the
+        * remote side to update each ref in the list from old_sha1 to
+        * peer_ref->new_sha1.
+        *
+        * Where possible, set the status for each ref appropriately.
+        *
+        * The transport must modify new_sha1 in the ref to the new
+        * value if the remote accepted the change. Note that this
+        * could be a different value from peer_ref->new_sha1 if the
+        * process involved generating new commits.
+        **/
        int (*push_refs)(struct transport *transport, struct ref *refs, int flags);
        int (*push)(struct transport *connection, int refspec_nr, const char **refspec, int flags);
 
+       /** get_refs_list(), fetch(), and push_refs() can keep
+        * resources (such as a connection) reserved for futher
+        * use. disconnect() releases these resources.
+        **/
        int (*disconnect)(struct transport *connection);
        char *pack_lockfile;
        signed verbose : 3;
@@ -74,7 +111,7 @@ int transport_push(struct transport *connection,
 
 const struct ref *transport_get_remote_refs(struct transport *transport);
 
-int transport_fetch_refs(struct transport *transport, const struct ref *refs);
+int transport_fetch_refs(struct transport *transport, struct ref *refs);
 void transport_unlock_pack(struct transport *transport);
 int transport_disconnect(struct transport *transport);
 char *transport_anonymize_url(const char *url);