~bzr-pqm/bzr/bzr.dev : changes

~bzr-pqm/bzr/bzr.dev » Changes from revision 4343.3.21

Rev	Summary	Authors	Date
4343.3.21	Implement get_missing_parents in terms of _KeyRefs. Implement get_missing_parents in terms of _KeyRefs. This gives _GCGraphIndex a _key_dependencies variable. Things still fail, because get_missing_parent_inventories() assumes that fileids_altered_by_revision_ids will return file_ids for revisions outside of the actual revisions requested. CHKInventoryRepository actually explicitly eliminates those keys. We may just need a custom get_missing_parent_inventories, but one way or another we have issues.	John Arbash Meinel	15 years ago
4343.3.20	Copy the track_external_parent_refs tests over to GCVF. Copy the track_external_parent_refs tests over to GCVF. Of course, the new format doesn't actually need to implement _key_dependencies in order to do this. There really should be a direct test on KnitVersionedFiles if we are going to be using _index._key_dependencies as an outside VF API. (not quite public, but accessed directly from the Repo.)	John Arbash Meinel	15 years ago
4343.3.19	NEWS update for fixing bug #373455 NEWS update for fixing bug #373455	John Arbash Meinel	15 years ago
4343.3.18	merge bzr.dev 4354 to get the latest NEWS merge bzr.dev 4354 to get the latest NEWS	John Arbash Meinel	15 years ago
4343.3.17	Finally found a way to write a test case to exercise the cod... Finally found a way to write a test case to exercise the code path. The issue is that branch.bzrdir.sprout('target') does not actually use PendingAncestryResult, you have to supply a revision-id. So it required: 1) --dev6 format 2) branching from a stacked branch 3) stacked branch was across HPSS 4) Branching with a specific revision id I guess I'm not too suprised it wasn't caught earlier. Though it also hints that we need some clearer interfaces. I guess the problem is that when branching a stacked branch locally, we don't clear fallback_vfs, whereas when doing it from a remote location, we don't have fallback_vfs, and instead have a chained set of fetches.	John Arbash Meinel	15 years ago
4343.3.16	Change test_default_stacking to use the existing 'make_smart... Change test_default_stacking to use the existing 'make_smart_server' code. With added benefit that this properly cleans up the smart server	John Arbash Meinel	15 years ago
4343.3.15	remove test_chk_bytes, it wasn't being used anyway remove test_chk_bytes, it wasn't being used anyway	John Arbash Meinel	15 years ago
4343.3.14	Remove bogus 'test_foo' Remove bogus 'test_foo'	John Arbash Meinel	15 years ago
4343.3.13	Update the fetch tests to make sure we are properly testing ... Update the fetch tests to make sure we are properly testing a smart base branch	John Arbash Meinel	15 years ago
4343.3.12	instead of trusting the Search, use a get_parent_map call. instead of trusting the Search, use a get_parent_map call. We'll try to fix search up next.	John Arbash Meinel	15 years ago
4343.3.11	Change PendingAncestryResult to strip ghosts from .get_keys(... Change PendingAncestryResult to strip ghosts from .get_keys()	John Arbash Meinel	15 years ago
4343.3.10	Add a per_repository_reference test with real stacked repos. Add a per_repository_reference test with real stacked repos. This adds a test that sets up a stacked repository with an extra revision versus the stacked-on location, and then asserts you can branch from this. I would think this would be tested somewhere, but it wasn't a per-repository-reference test, so when I broke the code, it didn't fail an obvious test. This now changes the code so that if there are no keys to be fetched for the 'missing_keys', then it doesn't try to get a stream from that vf. There wasn't any real reason to do so, and it avoids trying to get a stream for chk_bytes when the vf doesn't exist.	John Arbash Meinel	15 years ago
4343.3.9	Remove an accidental 'bork'. Remove an accidental 'bork'. Though it does show I was correct about the code not being used.	John Arbash Meinel	15 years ago
4343.3.8	Some cleanup passes. Some cleanup passes. move test_resume_chk_bytes from per_repository_referece to test_pack_repository, as it is a bit more implementation specific. Make sure we don't allow get_stream_for_missing_keys() if we have a pending stream.	John Arbash Meinel	15 years ago
4343.3.7	Update the suspend/resume/commit/abort_write_group tests for... Update the suspend/resume/commit/abort_write_group tests for CHK1. It turns out that the parameterized tests weren't actually parameterized, always only using the default format. Added a test that 'commit_write_group()' also doesn't leave any detritus around in upload, and puts things into packs and indices correctly. Updated the 'suspend' test to test that all indices are present, and not just that the .pack file is. Updated the ResumedPack.finish() and ResumedPack.abort() to handle .cix entries.	John Arbash Meinel	15 years ago
4343.3.6	Take out the _delta_across_repos flag. Take out the _delta_across_repos flag. Instead, we now check that the either commit_write_group fails, or get_record_stream(..., True) succeeds. Also, change the make_source_with_delta_record so that it generates the closest thing to a delta for chk formats. We have to use 'insert_record_stream', because 'add_lines' doesn't delta for chk formats.	John Arbash Meinel	15 years ago
4343.3.5	small comment small comment	John Arbash Meinel	15 years ago
4343.3.4	Custom implementation for GroupCHKStreamSource.get_stream_fo... Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys. This now ensures that all referenced chk pages will be transmitted when we are requested to send missing inventories. Also, adding a file to use for testing 'chk_bytes' specific information when testing stacking.	John Arbash Meinel	15 years ago
4343.3.3	Be sure to check for missing compression parents in chk_byte... Be sure to check for missing compression parents in chk_bytes	John Arbash Meinel	15 years ago
4343.3.2	All stacking tests seem to be passing for dev6 repos All stacking tests seem to be passing for dev6 repos 1) Add _GCGraphIndex.scan_unvalidated_index() It currently is a no-op because we a) Don't cache the possible external parent references b) Don't have external compression parents, so we can't have missing compression parents that we need to scan either. 2) Introduce a .resumed_pack_factory and ResumedGCPack We only need this to handle the fact that '_get_external_refs' is always empty. It would probably be nicer if we could do this with a different flag.. 3) RepositoryFormat._deltas_across_repos I needed a way to disable the 'cannot commit because we have a delta' tests for GC repositories, because they currently have no way to represent a delta between pack files. So I added this flag for now. 4) ResumedPack now properly supports chk_index, and we add a test to ensure this fact.	John Arbash Meinel	15 years ago

Older »