bzr crashed with IndexError in git_to_bzr()

[Duncan McGreggor]

Binary package hint: bzr

I was using fast-import to isolate a subdirectory of a large branch. I used the following:

 $ bzr fast-export ~/lab/largeProject project.fe
 $ bzr fast-import-filter -i projectSubdir project.fe > subdir.fi
 $ mkdir filtered-project
 $ cd filtered-project
 $ bzr fast-import ../subdir.fi

This resulted in a "IndexError: list index out of range" exception.

ProblemType: Crash
DistroRelease: Ubuntu 10.10
Package: bzr 2.2.1-0ubuntu1
ProcVersionSignature: Ubuntu 2.6.35-22.34-generic 2.6.35.4
Uname: Linux 2.6.35-22-generic i686
Architecture: i386
BzrDebugFlags: set()
BzrVersion: 2.2.1
CommandLine: ['/usr/bin/bzr', 'fast-import', '../new.fi']
CrashDb: bzr
Date: Sun Oct 17 09:35:42 2010
ExecutablePath: /usr/bin/bzr
FileSystemEncoding: ANSI_X3.4-1968
InstallationMedia: Ubuntu-Netbook 10.10 "Maverick Meerkat" - Alpha i386 (20100803.1)
InterpreterPath: /usr/bin/python2.6
Locale: en_US.UTF-8
Platform: Linux-2.6.35-22-generic-i686-with-Ubuntu-10.10-maverick
ProcCmdline: /usr/bin/python /usr/bin/bzr fast-import ../new.fi
ProcEnviron:
 SHELL=/bin/bash
 PATH=(custom, user)
 LANG=en_US.UTF-8
 LANGUAGE=en_US.UTF-8
 LC_CTYPE=C
PythonVersion: 2.6.6
SourcePackage: bzr
Title: bzr crashed with IndexError in git_to_bzr()
UserEncoding: ANSI_X3.4-1968
UserGroups: adm admin cdrom dialout lpadmin plugdev sambashare

[Apport retracing service (apport)]

Thank you for taking the time to report this crash and helping to make Ubuntu better. This particular crash has already been reported and is a duplicate of bug #660540, so is being marked as such. Please look at the other bug report to see if there is any missing information that you can provide, or to see if there is a workaround for the bug. Additionally, any further discussion regarding the bug should occur in the other report. Please continue to report any other bugs you may find.

[Duncan McGreggor (oubiwann)]

Hrm, interesting. I added a pdb set_trace call to check this out. Looks like a string is being iterated over in a rather strange way. I can't imagine that 'refs/heads/master' is intended to be iterated over by-character. Rather, by a split on '/'.

Here's the interactive pdb session:

(Pdb) list
 26 the Bazaar branch names as values.
 27 """
 28 bazaar_names = {}
 29 for ref_name in sorted(ref_names):
 30 import pdb;pdb.set_trace()
 31 -> parts = ref_name.split('/')
 32 if parts[0] == 'refs':
 33 parts.pop(0)
 34 category = parts.pop(0)
 35 if category == 'heads':
 36 bazaar_name = self._git_to_bzr_name(parts[-1])
(Pdb) ref_name
'/'
(Pdb) ref_names
'refs/heads/master'
(Pdb) sorted(ref_names)
['/', '/', 'a', 'a', 'd', 'e', 'e', 'e', 'f', 'h', 'm', 'r', 'r', 's', 's', 's', 't']
(Pdb)

I added the following just before the 'for' loop:

        if not isinstance(ref_names, list):
            ref_names = [ref_names]

This resulted in another exception, however:

  File "/usr/lib/python2.6/dist-packages/bzrlib/revision.py", line 93, in _check_properties
    (value, name))
ValueError: invalid property value {'refs/heads/master': 'trunk'} for 'branch-nick'

After some hacking around, it seems that even though git_to_bzr() is returning a dict, when a string is expected. In particular, check out the following from a pdb session:

(Pdb) down
> /usr/lib/python2.6/dist-packages/bzrlib/plugins/fastimport/bzr_commit_handler.py(227)build_revision()
-> self.branch_ref)
(Pdb) list
222
223 def build_revision(self):
224 rev_props = self._legal_revision_properties(self.command.properties)
225 if 'branch-nick' not in rev_props:
226 rev_props['branch-nick'] = self.cache_mgr.branch_mapper.git_to_bzr(
227 -> self.branch_ref)
228 self._save_author_info(rev_props)
229 committer = self.command.committer
230 who = self._format_name_email(committer[0], committer[1])
231 message = self.command.message
232 if not _serializer_handles_escaping:

rev_props['branch-nick'] should be getting a string assigned to it, but instead, it's getting a dict.

I hacked around this (badly) just to get things working by returning the following from git_to_bzr:

return bazaar_names.values()[0]

Hope this helps...