2a format should be the default but some tests fail

I have a plugin at https://code.launchpad.net/mbp/+junk/default2a that globally sets bzr's default format to 2a. Use care if you install that.

Using this plugin and running selftest shows several failures. The first I'm looking at is

test: bzrlib.tests.blackbox.test_branch.TestBranch.test_branch_hardlink
failure: bzrlib.tests.blackbox.test_branch.TestBranch.test_branch_hardlink [
RemoteException: Traceback (most recent call last):
  File "/home/mbp/bzr/398668-2a-default/bzrlib/tests/blackbox/test_branch.py", line 99, in test_branch_hardlink
    self.assertEqual(source_stat, target_stat)
AssertionError: not equal:
a = posix.stat_result(st_mode=33188, st_ino=745538, st_dev=64519L, st_nlink=1, st_uid=1000, st_gid=1000, st_size=25, st_atime=1249265516, st_mtime=1249265516, st_ctime=1249265516)
b = posix.stat_result(st_mode=33188, st_ino=745583, st_dev=64519L, st_nlink=1, st_uid=1000, st_gid=1000, st_size=25, st_atime=1249265516, st_mtime=1249265516, st_ctime=1249265516)

apparently meaning that branch --hardlink does not hardlink the working copy files if it fails.

The issue in the previous comment is bug 408193.

Bug 375013 is another failure, the only one in the hpss blackbox tests. I haven't looked at it yet.

Bug 408199 is another shallow failure, now fixed.

Just skimming over the current failures - bug 408824 and also there are also some failures related to stacking:

test: bzrlib.tests.blackbox.test_status.BranchStatus.test_tree_status_specific_files
error: bzrlib.tests.blackbox.test_status.BranchStatus.test_tree_status_specific_files [
RemoteException: Traceback (most recent call last):
  File "/home/mbp/bzr/398668-2a-default/bzrlib/tests/blackbox/test_status.py", line 259, in test_tree_status_specific_files
    short=True, revision=revs)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/status.py", line 162, in show_tree_status
    if nonexistents:
UnboundLocalError: local variable 'nonexistents' referenced before assignment

]
test: bzrlib.tests.blackbox.test_status.CheckoutStatus.test_tree_status_specific_files
error: bzrlib.tests.blackbox.test_status.CheckoutStatus.test_tree_status_specific_files [
RemoteException: Traceback (most recent call last):
  File "/home/mbp/bzr/398668-2a-default/bzrlib/tests/blackbox/test_status.py", line 259, in test_tree_status_specific_files
    short=True, revision=revs)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/status.py", line 162, in show_tree_status
    if nonexistents:
UnboundLocalError: local variable 'nonexistents' referenced before assignment

]

and in testing stacking across hpss

test: bzrlib.tests.per_bzrdir.test_bzrdir.TestBzrDir.test_format_initialize_on_transport_ex_default_stack_on(BzrDirMetaFormat1)
error: bzrlib.tests.per_bzrdir.test_bzrdir.TestBzrDir.test_format_initialize_on_transport_ex_default_stack_on(BzrDirMetaFormat1) [
RemoteException: Traceback (most recent call last):
  File "/home/mbp/bzr/398668-2a-default/bzrlib/tests/per_bzrdir/test_bzrdir.py", line 1272, in test_format_initialize_on_transport_ex_default_s
tack_on
    t, need_meta=True, repo_format_name=repo_name, stacked_on=None)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/tests/per_bzrdir/test_bzrdir.py", line 1323, in assertInitializeEx
    self.bzrdir_format.initialize_on_transport_ex(t, **kwargs)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/bzrdir.py", line 1948, in initialize_on_transport_ex
    make_working_trees, shared_repo)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/bzrdir.py", line 3624, in acquire_repository
    possible_transports=[self._bzrdir.transport])
  File "/home/mbp/bzr/398668-2a-default/bzrlib/bzrdir.py", line 3571, in _add_fallback
    repository.add_fallback_repository(stacked_repo)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/repository.py", line 965, in add_fallback_repository
    self._check_fallback_repository(repository)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/repository.py", line 981, in _check_fallback_repository
    return InterRepository._assert_same_model(self, repository)
  File "/home/mbp/bzr/398668-2a-default/bzrlib/repository.py", line 3308, in _assert_same_model
    "different rich-root support")
IncompatibleRepositories: CHKInventoryRepository('file:///tmp/testbzr-4TD5P3.tmp/bzrlib.tests.per_bzrdir.test_bzrdir.TestBzrDir.test_format_ini
tialize_on_transport_ex_default_stack_on%28BzrDirMetaFormat1%29/work/stack-on/.bzr/repository/')
is not compatible with
KnitPackRepository('file:///tmp/testbzr-4TD5P3.tmp/bzrlib.tests.p...

I've made a branch lp:~lifeless/bzr/bug-398668 which sets the format to 2a. Attached to this bug is a subunit log of the failing tests.

2009/8/12 Robert Collins <email address hidden>:
> I've made a branch lp:~lifeless/bzr/bug-398668 which sets the format to
> 2a. Attached to this bug is a subunit log of the failing tests.

Be aware I might still have some fixes in review or awaiting merge.

--
Martin <http://launchpad.net/~mbp/>

On Wed, 2009-08-12 at 11:11 +0000, Martin Pool wrote:
> 2009/8/12 Robert Collins <email address hidden>:
> > I've made a branch lp:~lifeless/bzr/bug-398668 which sets the format to
> > 2a. Attached to this bug is a subunit log of the failing tests.
>
> Be aware I might still have some fixes in review or awaiting merge.

Thats true; OTOH I looked at all the branches, and the hardlink one
hasn't got consensus; the status one I landed, and I didn't see any
other completed branches.

-Rob

nb: my subunit run was missing a default change for branch and trees; I'll do another full run once the known failures are fixed.

2009/8/13 Robert Collins <email address hidden>:
> On Wed, 2009-08-12 at 11:11 +0000, Martin Pool wrote:
>> 2009/8/12 Robert Collins <email address hidden>:
>> > I've made a branch lp:~lifeless/bzr/bug-398668 which sets the format to
>> > 2a. Attached to this bug is a subunit log of the failing tests.
>>
>> Be aware I might still have some fixes in review or awaiting merge.
>
> Thats true; OTOH I looked at all the branches, and the hardlink one
> hasn't got consensus; the status one I landed, and I didn't see any
> other completed branches.

OK, thanks for checking.

My patch to give a warning was merged in
https://code.edge.launchpad.net/~mbp/bzr/408193-hardlink

The rest of that is still a bug but not a 2.0 blocker.

--
Martin <http://launchpad.net/~mbp/>

My branch here has a number of fixes:
bzr+ssh://bazaar.launchpad.net/~lifeless/bzr/bug-398668

Many are pending review in lp; others aren't separate enough to the act of changing are just done in the branch.

bzr+ssh://bazaar.launchpad.net/~lifeless/bzr/bug-398668 is looking good:
FAILED (failures=12, errors=16, known_failure_count=23)

OK, so there now seem to be more failures in stacking; they might be fixed by the default-rich-root fix, or maybe not.

I'm seeing several failures still like this:

FAIL: per_branch.test_stacking.TestStacking.test_clone_stacking_policy_handling(BzrBranchFormat5)
    UnstackableBranchFormat not raised

FAIL: per_branch.test_stacking.TestStacking.test_clone_stacking_policy_handling(BzrBranchFormat6)
    UnstackableBranchFormat not raised

ERROR: per_branch.test_stacking.TestStacking.test_sprout_stacking_policy_handling(BzrBranchFormat5)
    The branch 'file:///tmp/testbzr-CUXNDJ.tmp/bzrlib.tests.per_branch.test_stacking.TestStacking.test_sprout_stacking_policy_handling%28BzrBranchFormat5%29/work/target/'(Branch format 5) is not a stackable format. You will need to upgrade the branch to permit branch stacking.

ERROR: per_branch.test_stacking.TestStacking.test_sprout_stacking_policy_handling(BzrBranchFormat6)
    The branch 'file:///tmp/testbzr-CUXNDJ.tmp/bzrlib.tests.per_branch.test_stacking.TestStacking.test_sprout_stacking_policy_handling%28BzrBranchFormat6%29/work/target/'(Branch format 6) is not a stackable format. You will need to upgrade the branch to permit branch stacking.

ERROR: per_branch.test_stacking.TestStacking.test_sprout_to_smart_server_stacking_policy_handling(BzrBranchFormat5)
    The branch 'bzr://127.0.0.1:54204/extra/target/'(Remote BZR Branch) is not a stackable format. You will need to upgrade the branch to permit branch stacking.

ERROR: per_branch.test_stacking.TestStacking.test_sprout_to_smart_server_stacking_policy_handling(BzrBranchFormat6)
    The branch 'bzr://127.0.0.1:52552/extra/target/'(Remote BZR Branch) is not a stackable format. You will need to upgrade the branch to permit branch stacking.

On Fri, 2009-08-14 at 12:21 +0000, Martin Pool wrote:
> OK, so there now seem to be more failures in stacking; they might be
> fixed by the default-rich-root fix, or maybe not.

I think we should delete default-rich-root as part of the move to 2.0.

-Rob

2009/8/15 Robert Collins <email address hidden>:
> On Fri, 2009-08-14 at 12:21 +0000, Martin Pool wrote:
>> OK, so there now seem to be more failures in stacking; they might be
>> fixed by the default-rich-root fix, or maybe not.
>
> I think we should delete default-rich-root as part of the move to 2.0.

At least some of these were https://bugs.edge.launchpad.net/bzr/+bug/413541.

I'd be happy to delete that name; we'll also need to remove some code
that internally relies on it; it's not quite totally trivial. Maybe
we could hide it.

--
Martin <http://launchpad.net/~mbp/>

2009/8/15 Robert Collins <email address hidden>:
> On Fri, 2009-08-14 at 12:21 +0000, Martin Pool wrote:
>> OK, so there now seem to be more failures in stacking; they might be
>> fixed by the default-rich-root fix, or maybe not.
>
> I think we should delete default-rich-root as part of the move to 2.0.

My default2a plugin now does that and I'll post more test results with
that in place.

--
Martin <http://launchpad.net/~mbp/>

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

Martin Pool wrote:
> 2009/8/15 Robert Collins <email address hidden>:
>> On Fri, 2009-08-14 at 12:21 +0000, Martin Pool wrote:
>>> OK, so there now seem to be more failures in stacking; they might be
>>> fixed by the default-rich-root fix, or maybe not.
>> I think we should delete default-rich-root as part of the move to 2.0.
>
> At least some of these were
> https://bugs.edge.launchpad.net/bzr/+bug/413541.
>
> I'd be happy to delete that name; we'll also need to remove some code
> that internally relies on it; it's not quite totally trivial. Maybe
> we could hide it.
>

I would also mention that it might break scripts that people were using
to convert from svn, for example.

I would rather make it hidden than removed. Or possibly deprecate it.

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

iEYEARECAAYFAkqGvl8ACgkQJdeBCYSNAAMXvACePXHctuUoRAT4xOC+cliPR74U
UQ8AoJUWSwJ9muLSp+WpudUF5WSEnjSk
=XViw
-----END PGP SIGNATURE-----

hidden would be fine with me.

The reason I want it to go from the ui is that its == 'default' now, and
will add confusion.

-Rob

I'm driving this at the moment ;)
My bzr+ssh://bazaar.launchpad.net/~lifeless/bzr/bug-398668 is down to 12 errors based on the reduce pattern. Nearly time for a full run and cleanup.

No tests failing (of the test that were failing) in my branch now.

Will run full suite tomorrow and land the many reviewed incremental fixes.

4 new failures:
per_repository.test_repository.TestRepository.test_clone_unstackable_branch_preserves_stackable_repo_format(RepositoryFormatKnit1)
per_repository.test_repository.TestRepository.test_clone_unstackable_branch_preserves_stackable_repo_format(RepositoryFormat7)
bzrlib.tests.test_branch.TestDefaultFormat.test_default_format
bzrlib.tests.test_remote.TestRepositoryInsertStream.test_stream_with_inventory_deltas

fixed