Inventories get added with different parents

[Jelmer Vernooij]

  affects bzr
  status triaged
  importance high

When adding the same inventory twice, Bazaar crashes:

bzr: ERROR: bzrlib.errors.KnitCorrupt: Knit
<bzrlib.groupcompress._GCGraphIndex object at 0x36de5d0> corrupt:
inconsistent details in add_recor
ds: ('269 230 0 295', ((),)) ('269 230 0 295', ([],))

Traceback (most recent call last):
  File "/home/jelmer/bzr/bzr.dev/bzrlib/commands.py", line 831, in
exception_to_return_code
    return the_callable(*args, **kwargs)
  File "/home/jelmer/bzr/bzr.dev/bzrlib/commands.py", line 1026, in
run_bzr
    ret = run(*run_argv)
  File "/home/jelmer/bzr/bzr.dev/bzrlib/commands.py", line 643, in
run_argv_aliases
    return self.run(**all_cmd_args)
  File "/home/jelmer/.bazaar/dev-plugins/svn/commands.py", line 177,
in run
    to_revnum=to_revnum, prefix=prefix)
  File "/home/jelmer/.bazaar/dev-plugins/svn/convert.py", line 371, in
convert_repository
    RepositoryConverter(*args, **kwargs)
  File "/home/jelmer/.bazaar/dev-plugins/svn/convert.py", line 248, in
__init__
    revfinder, mapping, heads)
  File "/home/jelmer/.bazaar/dev-plugins/svn/convert.py", line 296, in
_fetch_to_shared_repo
    inter.fetch(needed=revfinder.get_missing())
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 1260, in
fetch
    use_replay=self._use_replay)
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 1191, in
_fetch_revisions
    parent_revmeta)
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 1136, in
_fetch_revision_switch
    report_inventory_contents(reporter, parent_revnum, start_empty)
  File "/home/jelmer/.bazaar/dev-plugins/svn/errors.py", line 129, in
convert
    return unbound(*args, **kwargs)
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 903, in
report_inventory_contents
    reporter.finish()
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 285, in close
    self._close()
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 446, in
_close
    self.editor._finish_commit()
  File "/home/jelmer/.bazaar/dev-plugins/svn/fetch.py", line 682, in
_finish_commit
    tuple([r for r in rev.parent_ids if r in
self.target.has_revisions(rev.parent_ids)]))
  File
"/home/jelmer/bzr/bzr.dev/bzrlib/repofmt/groupcompress_repo.py", line
704, in add_inventory_by_delta
    basis_revision_id, delta, new_revision_id, parents)
  File "/home/jelmer/bzr/bzr.dev/bzrlib/repository.py", line 1065, in
add_inventory_by_delta
    return (self.add_inventory(new_revision_id, basis_inv, parents),
  File "/home/jelmer/bzr/bzr.dev/bzrlib/repository.py", line 1013, in
add_inventory
    return self._add_inventory_checked(revision_id, inv, parents)
  File
"/home/jelmer/bzr/bzr.dev/bzrlib/repofmt/groupcompress_repo.py", line
675, in _add_inventory_checked
    inv_lines, check_content=False)
  File "/home/jelmer/bzr/bzr.dev/bzrlib/repository.py", line 1075, in
_inventory_add_lines
    check_content=check_content)[0]
  File "/home/jelmer/bzr/bzr.dev/bzrlib/groupcompress.py", line 1008,
in add_lines
    nostore_sha=nostore_sha))[0]
  File "/home/jelmer/bzr/bzr.dev/bzrlib/groupcompress.py", line 1500,
in _insert_record_stream
    flush()
  File "/home/jelmer/bzr/bzr.dev/bzrlib/groupcompress.py", line 1402,
in flush
    self._index.add_records(nodes, random_id=random_id)
  File "/home/jelmer/bzr/bzr.dev/bzrlib/groupcompress.py", line 1620,
in add_records
    ": %s %s" % ((value, node_refs), keys[key]))
KnitCorrupt: Knit <bzrlib.groupcompress._GCGraphIndex object at
0x36de5d0> corrupt: inconsistent details in add_records: ('269 230 0
295', ((),)) ('269 230 0 295', ([],))

bzr 1.17dev on python 2.5.4 (linux2)
arguments: ['/home/jelmer/bzr/bzr.dev/bzr', 'svn-import',
'--incremental', 'https://svn.sourceforge.net/svnroot/libexif', 'exif']
encoding: 'UTF-8', fsenc: 'UTF-8', lang: 'en_US.UTF-8'
plugins:
  builddeb /home/jelmer/.bazaar/dev-plugins/builddeb [unknown]
  bzrtools /home/jelmer/.bazaar/dev-plugins/bzrtools [1.16]
  cia /home/jelmer/.bazaar/dev-plugins/cia [1.0dev]
  colocated /home/jelmer/.bazaar/dev-plugins/colocated
[0.4.16dev]
  cvsps_import
/usr/lib/python2.5/site-packages/bzrlib/plugins/cvsps_import [unknown]
  dbus /home/jelmer/.bazaar/dev-plugins/dbus.py [unknown]
  email
/usr/lib/python2.5/site-packages/bzrlib/plugins/email [unknown]
  etckeeper
/usr/lib/python2.5/site-packages/bzrlib/plugins/etckeeper [unknown]
  fastimport
/usr/lib/python2.5/site-packages/bzrlib/plugins/fastimport [0.8dev]
  git /home/jelmer/.bazaar/dev-plugins/git [0.3.3dev]
  gtk /home/jelmer/.bazaar/dev-plugins/gtk [0.96.0.dev.1]
  hg /home/jelmer/.bazaar/dev-plugins/hg [unknown]
  jelmer /home/jelmer/.bazaar/dev-plugins/jelmer [unknown]
  launchpad
/home/jelmer/bzr/bzr.dev/bzrlib/plugins/launchpad [1.17dev]
  loggerhead /home/jelmer/.bazaar/dev-plugins/loggerhead [1.11]
  loom /home/jelmer/.bazaar/dev-plugins/loom [1.4dev]
  netrc_credential_store
/home/jelmer/bzr/bzr.dev/bzrlib/plugins/netrc_credential_store [1.17dev]
  pqm /home/jelmer/.bazaar/dev-plugins/pqm [1.4dev]
  qbzr /home/jelmer/.bazaar/dev-plugins/qbzr [0.12dev]
  rebase /home/jelmer/.bazaar/dev-plugins/rebase [0.5.1dev]
  search /home/jelmer/.bazaar/dev-plugins/search [1.7dev]
  stats
/usr/lib/python2.5/site-packages/bzrlib/plugins/stats [unknown]
  svn /home/jelmer/.bazaar/dev-plugins/svn [0.6.2dev]
  upload /home/jelmer/.bazaar/dev-plugins/upload [1.0dev]
  webdav /home/jelmer/.bazaar/dev-plugins/webdav [1.12]
*** Bazaar has encountered an internal error.
    Please report a bug at https://bugs.launchpad.net/bzr/+filebug
    including this traceback, and a description of what you
    were doing when the error occurred.

Some layers in bazaar seem to return parents as a list and others
return them as a tuple. Which of the two is correct here?

[John A Meinel (jameinel)]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

...

> KnitCorrupt: Knit <bzrlib.groupcompress._GCGraphIndex object at
> 0x36de5d0> corrupt: inconsistent details in add_records: ('269 230 0
> 295', ((),)) ('269 230 0 295', ([],))
>

Well, the second time you are adding it with an empty parent *list*
while it has an empty parent *tuple*.

It looks like we use "record.parents" to populate that field.

I don't know where that value is getting set.

We could do a safety trap against it with something like:
=== modified file 'bzrlib/groupcompress.py'
- --- bzrlib/groupcompress.py 2009-06-11 20:34:56 +0000
+++ bzrlib/groupcompress.py 2009-06-17 14:17:23 +0000
@@ -1610,7 +1610,7 @@
                                 "in parentless index.")
                     refs = ()
                     changed = True
- - keys[key] = (value, refs)
+ keys[key] = (value, tuple(refs))
         # check for dups
         if not random_id:
             present_nodes = self._get_entries(keys)

However, I'd really like to understand

1) Why do you need to add the inventory twice
2) Why the 'parents' entry is a list sometimes and a tuple other times.
(It is always a tuple when reading from bzr indexes, so I'm assuming it
is set somewhere else.)

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAko4+zkACgkQJdeBCYSNAAOCrwCgvXuB5t51G5eC3iKDhP+4Bezp
9a0AoI+Mosh1YBcsHdOxHfFzboBcrNZb
=RgIw
-----END PGP SIGNATURE-----

[Robert Collins (lifeless)]

tuples are the right representation. Perhaps its something from bzr-svn outputting an empty list?

[Jelmer Vernooij (jelmer)]

This was already fixed in bzr-svn 0.6.3.