bzr convert to git things (bzr2git)

README.md

converting bzr 2 git things

Bug 1606973 was my original file, and then changes to bzr export.

The patch was improved by some others, and I'm storing my version of it here.

Features

The benefit of the patch is that you get:

• bug metadata converted: Bzr contains 'fixes' metadata (bzr commit --fixes) that does not get moved over to git. This updates the commit message exported to git to contain LP: #XXXXXX for each --fixes=lp:XXXXXX in bzr.

• bzr revno info: Often times bug or other references to code may say 'fixed in revno XXX'. That information gets lost in a conversion to git. The patch here updates commit messages to contain bzr-revno: XXX for each bzr revision.

Installing

I just installed from Ubuntu distro.

$ sudo apt-get install bzr-fastimport bzr

Patching

I just patched the package-installed version of fastexport.py. If you find that unacceptable, then update the path to the right directory and it will still apply.

$ ( cd /usr/lib/python2.7/dist-packages/bzrlib/plugins/ &&
  sudo patch -p0 --dry-run ) < fast-export.patch

Doing a convert

$ ./bzr2git bzr-dir git-dir
...
$ cd git-dir
$ git branch
master
$ git log
...

bzr2git

#!/bin/bash

VERBOSITY=0

error() { echo "$@" 1>&2; }
fail() { local r=$?;  [ $r -eq 0 ] && r=1; failrc "$r" "$@"; }
failrc() { local r=$1; shift; [ $# -eq 0 ] || error "$@"; exit $r; }

Usage() {
    cat <<EOF
Usage: ${0##*/} [ options ] bzr-repo new-git-dir

   convert bzr to git for 'bzr-repo' with output in 'new-git-dir'

   options:
      -v | --verbose    be more verbose
EOF
}

bad_Usage() { Usage 1>&2; [ $# -eq 0 ] || error "$@"; return 1; }
cleanup() {
    [ -z "${TEMP_D}" -o ! -d "${TEMP_D}" ] || rm -Rf "${TEMP_D}"
}

debug() {
    local level=${1}; shift;
    [ "${level}" -gt "${VERBOSITY}" ] && return
    error "${@}"
}

main() {
    local short_opts="hv"
    local long_opts="help,verbose"
    local getopt_out=""
    getopt_out=$(getopt --name "${0##*/}" \
        --options "${short_opts}" --long "${long_opts}" -- "$@") &&
        eval set -- "${getopt_out}" ||
        { bad_Usage; return; }

    local cur="" next="" out_d="" bzr_d=""

    while [ $# -ne 0 ]; do
        cur="$1"; next="$2";
        case "$cur" in
            -h|--help) Usage ; exit 0;;
            -v|--verbose) VERBOSITY=$((${VERBOSITY}+1));;
            --) shift; break;;
        esac
        shift;
    done

    [ $# -eq 2 ] || { bad_Usage "Expected 2 args, got $# ($*)"; return; }

    bzr_d="$1"
    out_d="$2"
    [ -d "$bzr_d" ] || fail "bzr-repo '$bzr_d' is not a directory"

    if [ ! -d "$out_d" ]; then
        debug 1 "mkdir -p \"$out_d\""
        mkdir -p "$out_d" || fail "failed mkdir -p $out_d"
    fi

    if [ ! -d "$out_d/.git" ]; then
        debug 1 "git init \"$out_d\""
        git init "$out_d" ||
            fail "failed to git init in $out_d"
    fi

    set -o pipefail
    local rcs=""
    bzr_export=( bzr fast-export --rewrite-tag-names --plain "$bzr_d" )
    git_import=( git fast-import )
    debug 1 "${bzr_export[*]}" "|" \
        "( cd \"$out_d\" && ${git_import[*]} )"
    bzr fast-export --plain "$bzr_d" | ( cd "$out_d" && "${git_import[@]}" )
    rcs=( "${PIPESTATUS[@]}" )
    [ "${rcs[0]}" != 0 ] &&
        failrc ${rcs[0]} "failed: ${bzr_export[*]}"
    [ "${rcs[1]}" != 0 ] &&
        failrc ${rcs[0]} "failed: ( cd $out_d && ${git_import[*]} )"

    debug 1 "( cd \"$out_d\" && git checkout master )"
    local out=""
    out=$( cd "$out_d" && git checkout master 2>&1 ) ||
        fail "failed git checkout master: $out"

    error "successfully created git directory in ${out_d}"
    return 0
}

main "$@"
# vi: ts=4 expandtab

fast-export.patch

--- fastimport/exporter.py.dist	2017-10-26 09:00:36.375185796 -0400
+++ fastimport/exporter.py	2017-12-18 12:28:39.386484215 -0500
@@ -57,6 +57,7 @@
     progress,
     trace,
     )
+from bzrlib.log import _compute_revno_str
 
 from bzrlib.plugins.fastimport import (
     helpers,
@@ -425,7 +426,70 @@
         # Filter the revision properties. Some metadata (like the
         # author information) is already exposed in other ways so
         # don't repeat it here.
+        addlines = []
+        message = revobj.message.encode("utf-8")
+        fixed_bugs = []
+        if b'\nLP: ' in message:
+            # Join any old LP: # in the bzr commit message to avoid duplicate
+            lines = []
+            for line in message.splitlines():
+                if not line.startswith(b'LP: '):
+                    lines.append(line)
+                    continue
+                replace_line = True
+                linebugs = []
+                for tok in line.split():
+                    if tok == b'LP:':
+                        continue
+                    tok = tok.replace(b"#", "").replace(b",", "")
+                    try:
+                        int(tok)
+                        linebugs.append(tok)
+                    except ValueError as e:
+                        sys.stderr.write(
+                            "WARNING: failed to parse '%s': %s\n" % (tok, line))
+                        replace_line = False
+                        break
+                if replace_line:
+                    fixed_bugs.extend(linebugs)
+                else:
+                    lines.append(line)
+            if fixed_bugs:
+                sys.stderr.write(
+                    "UPDATED: adjusted existing bzr LP: for %s\n" % fixed_bugs)
+            message = b'\n'.join(lines)
+            if fixed_bugs:
+                message = message.rstrip(b'\n') + b'\n'
+
         if self.plain_format:
+            if 'bugs' in revobj.properties:
+                # properties are a newline separated list of: bugurl action
+                #   http://bugs.launchpad.net/bug/XXXXX fixed
+                for bugstatement in revobj.properties["bugs"].split("\n"):
+                    bugurl, comment = bugstatement.decode().split()
+                    if not bugurl.startswith("https://launchpad.net/bugs/"):
+                        sys.stderr.write("unknown bug url: %s\n" % bugurl)
+                        continue
+                    if comment == "fixed":
+                        n = bugurl.rpartition("/")[2]
+                        fixed_bugs.append(bugurl.rpartition("/")[2])
+
+                # append to the message with LP: #XXX[, #XXX]
+                if fixed_bugs:
+                    fixed_bugs = sorted(set(fixed_bugs))
+                    addlines.append(
+                        "LP: %s" % ', '.join(["#" + n for n in fixed_bugs]))
+
+            if 'authors' in revobj.properties:
+                authors = revobj.properties['authors'].split('\n')
+                if len(authors) > 1:
+                    addlines.append("Original authors:")
+                    addlines.extend(["  - " + author for author in authors])
+
+            revno = _compute_revno_str(self.branch, revobj.revision_id)
+            if revno is not None:
+                addlines.append("bzr-revno: %s" % revno)
+
             properties = None
         else:
             properties = revobj.properties
@@ -435,9 +499,14 @@
                 except KeyError:
                     pass
 
+        if addlines:
+            if message[-1] != b"\n":
+                message += b"\n"
+            message = '\n'.join(
+                [message] + [a.encode("utf-8") for a in addlines])
         # Build and return the result
         return commands.CommitCommand(git_ref, str(mark), author_info,
-            committer_info, revobj.message.encode("utf-8"), from_, merges, iter(file_cmds),
+            committer_info, message, from_, merges, iter(file_cmds),
             more_authors=more_author_info, properties=properties)
 
     def _get_revision_trees(self, parent, revision_id):

snapcraft.yaml

## 2018-06-08 Currently not working, results in a segfault when run.
## not sure that patching is being done right.
## to check:
##   ( cd /snap/bzr2git/current/usr/lib/python2.*/dist-packages/
##     grep LP bzrlib/plugins/fastimport/exporter.py )
name: bzr2git
version: '0.1'
summary: Utility to convert bzr repos to git.
description: |
  Utility to convert bzr repos to git.

  This tool provides easy command to convert a bzr repo into git.
  It's additional features are:

    * bug metadata converted:
     This updates the commit message exported to git to contain
     'LP: #XXXXXX' for each `--fixes=lp:XXXXXX` in bzr.

    * bzr revno info: Often times bug or other references to code may
      say 'fixed in revno XXX'.  This updates git commit messages with
      'bzr-revno: XXX' for each bzr revision.

  Usage:
     bzr clone lp:your-repo
     bzr2git your-repo your-repo.git

grade: stable
confinement: strict

apps:
  bzr2git:
    command: bzr2git

parts:
  bzr2git:
    plugin: dump
    stage-packages:
      - bzr
      - bzr-fastimport
      - libc6
    override-build: |
      snapcraftctl build
      ( set -e
        plugdir=usr/lib/python2.7/dist-packages/bzrlib/plugins/
        cd ../install/$plugdir
        patch -p0 --dry-run
      ) < fast-export.patch
    organize:
      bzr2git: usr/bin/bzr2git

See also: my version of a bzr2git script (currently also a snap of the same name, but I'm willing to give it up if others want to maintain it). This supports iterative transitions, if you don't want a "flag day". But doesn't support bug metadata (or any bzr metadata, for that matter).