ensure base was bzr-git generated when applying deltas

On Sat, 2011-03-12 at 15:59 +0000, Eli Zaretskii wrote:
> Public bug reported:
>
> I tried to clone a remote git repository into a branch inside a sharewd
> bzr repo. This crashed with the following error message:
>
> bzr: ERROR: An inconsistent delta was supplied involving '<unknown>',
> 'intl' reason: Parent is not present in resulting inventory.
What versions of Bazaar, bzr-git and Dulwich are you using?

Can you run "bzr check" / "bzr reconcile" in the repository ? This error
is caused by bzr-git processing the existing revisions in that
repository and generating Git SHA1s for them. One of these revisions is
causing bzr-git to break.

  status incomplete

Cheers,

Jelmer

Sorry, forgot to mention the versions. They are the latest ones I could find: bzr-git 0.5.4 and dulwich 0.7.0.
The version of Bazaar is 2.3.0, installed with a stand-alone installer on Windows XP SP2.

"bzr check" reports nothing special, see below. Is there a point in running reconcile, given this?

Checking working tree at 'D:/gnu/bzr/idutils/mytrunk'.
Checking branch at 'file:///D:/gnu/bzr/idutils/mytrunk/'.
Checking working tree at 'D:/gnu/bzr/idutils/trunk'.
Checking branch at 'file:///D:/gnu/bzr/idutils/trunk/'.
Checking working tree at 'D:/gnu/bzr/idutils/trunk.remote'.
Checking branch at 'file:///D:/gnu/bzr/idutils/trunk.remote/'.
Checking working tree at 'D:/gnu/bzr/idutils/w32'.
Checking branch at 'file:///D:/gnu/bzr/idutils/w32/'.
Checking repository at 'file:///D:/gnu/bzr/idutils/'.
checked repository file:///D:/gnu/bzr/idutils/ format RepositoryFormat2a()
   438 revisions
  1361 file-ids
checked branch file:///D:/gnu/bzr/idutils/w32/ format Branch format 7
checked branch file:///D:/gnu/bzr/idutils/mytrunk/ format Branch format 7
checked branch file:///D:/gnu/bzr/idutils/trunk/ format Branch format 7
checked branch file:///D:/gnu/bzr/idutils/trunk.remote/ format Branch format 7

On Sat, 2011-03-12 at 16:40 +0000, Eli Zaretskii wrote:
> Sorry, forgot to mention the versions. They are the latest ones I could find: bzr-git 0.5.4 and dulwich 0.7.0.
> The version of Bazaar is 2.3.0, installed with a stand-alone installer on Windows XP SP2.
>
> "bzr check" reports nothing special, see below. Is there a point in
> running reconcile, given this?
Can you run bzr git-import again with "BZR_PDB=1" set? that should land
you in a debugger. If you can paste the backtrace from that exception
("bt" in the debugger) here that would be useful.

Cheers,

Jelmer

Here's what "bt" in the debugger shows:

bzr: ERROR: An inconsistent delta was supplied involving '<unknown>', 'intl'
reason: Parent is not present in resulting inventory.
**** entering debugger
> d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
(Pdb) bt
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(926)exception_to_return_code(
)
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(1126)run_bzr()
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(691)run_argv_aliases()
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(713)run()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(135)run_simple()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(165)_do_with_cleanups()
  d:\usr\bazaar\plugins\git\commands.py(92)run()
  d:\usr\bazaar\plugins\git\fetch.py(534)fetch()
  d:\usr\bazaar\plugins\git\fetch.py(596)fetch_objects()
  d:\usr\bazaar\plugins\git\fetch.py(465)import_git_objects()
  d:\usr\bazaar\plugins\git\fetch.py(380)import_git_commit()
  d:\gnu\bzr\idutils\bzrlib\repofmt\groupcompress_repo.pyo(986)add_inventory_by_
delta()
> d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
(Pdb)

Sorry, that was with "git-import". Here's the traceback from "bzr branch":

bzr: ERROR: An inconsistent delta was supplied involving '<unknown>', 'intl'
reason: Parent is not present in resulting inventory.
**** entering debugger
> d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
(Pdb) bt
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(926)exception_to_return_code()
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(1126)run_bzr()
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(691)run_argv_aliases()
  d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(713)run()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(135)run_simple()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(165)_do_with_cleanups()
  d:\gnu\bzr\idutils\bzrlib\builtins.pyo(1227)run()
  d:\gnu\bzr\idutils\bzrlib\controldir.pyo(374)sprout()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(131)run()
  d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(165)_do_with_cleanups()
  d:\gnu\bzr\idutils\bzrlib\controldir.pyo(425)_sprout()
  d:\gnu\bzr\idutils\bzrlib\repository.pyo(1781)fetch()
  d:\usr\bazaar\plugins\git\fetch.py(534)fetch()
  d:\usr\bazaar\plugins\git\fetch.py(596)fetch_objects()
  d:\usr\bazaar\plugins\git\fetch.py(465)import_git_objects()
  d:\usr\bazaar\plugins\git\fetch.py(380)import_git_commit()
  d:\gnu\bzr\idutils\bzrlib\repofmt\groupcompress_repo.pyo(986)add_inventory_by_delta()
> d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
(Pdb)

On Sat, 2011-03-12 at 19:31 +0000, Eli Zaretskii wrote:
> Sorry, that was with "git-import". Here's the traceback from "bzr
> branch":
>
> bzr: ERROR: An inconsistent delta was supplied involving '<unknown>', 'intl'
> reason: Parent is not present in resulting inventory.
> **** entering debugger
> > d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
> (Pdb) bt
> d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(926)exception_to_return_code()
> d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(1126)run_bzr()
> d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(691)run_argv_aliases()
> d:\usr\bazaar\lib\library.zip\bzrlib\commands.py(713)run()
> d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(135)run_simple()
> d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(165)_do_with_cleanups()
> d:\gnu\bzr\idutils\bzrlib\builtins.pyo(1227)run()
> d:\gnu\bzr\idutils\bzrlib\controldir.pyo(374)sprout()
> d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(131)run()
> d:\gnu\bzr\idutils\bzrlib\cleanup.pyo(165)_do_with_cleanups()
> d:\gnu\bzr\idutils\bzrlib\controldir.pyo(425)_sprout()
> d:\gnu\bzr\idutils\bzrlib\repository.pyo(1781)fetch()
> d:\usr\bazaar\plugins\git\fetch.py(534)fetch()
> d:\usr\bazaar\plugins\git\fetch.py(596)fetch_objects()
> d:\usr\bazaar\plugins\git\fetch.py(465)import_git_objects()
> d:\usr\bazaar\plugins\git\fetch.py(380)import_git_commit()
> d:\gnu\bzr\idutils\bzrlib\repofmt\groupcompress_repo.pyo(986)add_inventory_by_delta()
> > d:\gnu\bzr\idutils\bzrlib\inventory.pyo(1776)create_by_apply_delta()
> (Pdb)
Hmm, looks like it's creating a delta against an existing revision in
that repository. Did you use earlier versions of bzr-git with this
repository perchance?

Cheers,

Jelmer

I don't think I ever used bzr-git in that repository. However, one of the branches in the repository was created by "bzr fast-import" using a stream exported from the same git repository. (The export was done on another machine, and the resulting .fi file was copied to this machine and used to create a branch in this repository.)

Could this be the source of the problem?

On Sat, 2011-03-12 at 20:21 +0000, Eli Zaretskii wrote:
> I don't think I ever used bzr-git in that repository. However, one of
> the branches in the repository was created by "bzr fast-import" using a
> stream exported from the same git repository. (The export was done on
> another machine, and the resulting .fi file was copied to this machine
> and used to create a branch in this repository.)
>
> Could this be the source of the problem?
Maybe. There is clearly a buggy revision here. Can you print the
problematic revision with "print new_revision_id" and the current
revision with "print self.revision_id" in the pdb prompt?

Cheers,

Jelmer

Here:

(Pdb) print new_revision_id
git-v1:802eb1d3b0f18993e882369cf02452ded696897c
(Pdb) print self.revision_id
<email address hidden>
(Pdb)

On Sat, 2011-03-12 at 22:49 +0000, Eli Zaretskii wrote:
> Here:
>
> (Pdb) print new_revision_id
> git-v1:802eb1d3b0f18993e882369cf02452ded696897c
> (Pdb) print self.revision_id
> <email address hidden>
> (Pdb)
Euhm, WTF?

Does "bzr git-objects 802eb1d3b0f18993e882369cf02452ded696897c" work on
that repository ?

Cheers,

Jelmer

Seems so:

D:\gnu\bzr\idutils>bzr git-objects 802eb1d3b0f18993e882369cf02452ded696897c
tree ca8cc72963b395e49271f23d436ecc561868c3ff
parent 9432121c2ddc629db098b8d194275e0f8a1ce270
author Jim Meyering <email address hidden> 826693640 +0000
committer Jim Meyering <email address hidden> 826693640 +0000

.

On Sun, 2011-03-13 at 04:02 +0000, Eli Zaretskii wrote:
> Seems so:
>
> D:\gnu\bzr\idutils>bzr git-objects 802eb1d3b0f18993e882369cf02452ded696897c
> tree ca8cc72963b395e49271f23d436ecc561868c3ff
> parent 9432121c2ddc629db098b8d194275e0f8a1ce270
> author Jim Meyering <email address hidden> 826693640 +0000
> committer Jim Meyering <email address hidden> 826693640 +0000

 summary "ensure base was bzr-git generated when applying deltas"
 status triaged
 importance medium

This looks like it is indeed a clash between the revisions generated by
bzr-git and bzr-fastimport.

Cheers,

Jelmer