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

~bzr-pqm/bzr/bzr.dev » Changes from revision 5847.4.21

Rev	Summary	Authors	Date
5847.4.21	We were raising exceptions improperly. A couple more tests c... We were raising exceptions improperly. A couple more tests cases, so far so good.	John Arbash Meinel	13 years ago
5847.4.20	Handle a bunch of rename cases. Handle a bunch of rename cases.	John Arbash Meinel	13 years ago
5847.4.19	Get the ugly 'exists, but at a different location' working. Get the ugly 'exists, but at a different location' working.	John Arbash Meinel	13 years ago
5847.4.18	And now we add a bunch more tests, and finally have a failin... And now we add a bunch more tests, and finally have a failing case.	John Arbash Meinel	13 years ago
5847.4.17	The 25ms was the time spent checking that we didn't alter th... The 25ms was the time spent checking that we didn't alter the containing WT. Switch to using temp files everywhere, since we don't actually need to write anything to disk. (Except DirState requires a disk file.) Now down to ~2ms per test, which is a great feeling.	John Arbash Meinel	13 years ago
5847.4.16	Switching from BranchBuilder to Inventory directly dropped u... Switching from BranchBuilder to Inventory directly dropped us from 73ms to 26ms. Probably as good as we'll get.	John Arbash Meinel	13 years ago
5847.4.15	two simple tests in the new framework. two simple tests in the new framework. Not quite as fast as I would like, but it at least works.	John Arbash Meinel	13 years ago
5847.4.14	Create a direct helper for testing UpdateBasisByDelta cases. Create a direct helper for testing UpdateBasisByDelta cases. Hopefully this will be simpler to manage, and set up the edge cases. It also allows us to directly inspect the dirstate object.	John Arbash Meinel	13 years ago
5847.4.13	Mostly more test cases, some small progress in getting Mostly more test cases, some small progress in getting update_basis_by_delta cases working.	John Arbash Meinel	13 years ago
5847.4.12	(broken) Expand the test coverage for cases we care about. (broken) Expand the test coverage for cases we care about. There are now a fair number of more edge cases if tree0 can diverge from tree1 when calling update_basis_by_delta. The biggest thing is that we now have to handle renames correctly. Just snapshotting progress, since it will take a bit more to actually get the tests passing.	John Arbash Meinel	13 years ago
5847.4.11	Update some documentation, also fix a serious bug in the old... Update some documentation, also fix a serious bug in the old code. Specifically, any real_add is going to fail with _get_entry(1,) because the _get_entry code notices that the entry is 'a' on the basis tree. So we fall back to _get_entry(0). Another option comes to mind, though, which is to always call _get_entry(0,) without a file_id, then check the file_id, and only call _get_entry(1, ) if the [0] file_id doesn't match. That might save some bisection lookups in common cases.	John Arbash Meinel	13 years ago
5847.4.10	A few more bug fixes. A few more bug fixes. Handle when a location changes file_id. We were looking up the path in the active tree, we need to look it up in the basis_tree. Add a direct test for the 'switch' case, which is when setting a parent_tree to something new, and the basis_tree is no longer in the available repository. Add a couple more direct successful delta checks. (most of the test_inv tests were just about failing with bad deltas.)	John Arbash Meinel	13 years ago
5847.4.9	Have to set update=True if you don't want to do the work Have to set update=True if you don't want to do the work	John Arbash Meinel	13 years ago
5847.4.8	Fix some 'switch' tests. Fix some 'switch' tests. Basically, with switch a WorkingTree can be pointing to a revision that will no longer be present in the repository the WT is associated with. So if we can't lookup the basis_tree from the repo, just fall back to the regular 'set_parent_trees' code.	John Arbash Meinel	13 years ago
5847.4.7	Merge bzr.dev 5887, to get the update_basis_by_delta changes Merge bzr.dev 5887, to get the update_basis_by_delta changes	John Arbash Meinel	13 years ago
5847.4.6	Do another round of timing tests in a Launchpad tree. Do another round of timing tests in a Launchpad tree. Not everything improved nearly as much, but a few things improved considerably.	John Arbash Meinel	13 years ago
5847.4.5	Line up the column a little bit nicer. Line up the column a little bit nicer.	John Arbash Meinel	13 years ago
5847.4.4	Update the performance list in whats-new. Update the performance list in whats-new. We generally have 3 bits that we improved 1) Make loading the whole inventory not be pathologically slow. 2) Make more cases use optimized iter_changes() results (ones that don't require loading the whole inventory) 3) Update the WT basis cache using a delta, rather than requiring it to be set from scratch. (1) drops us from 3+minutes down to ~15s for each tree we have to load. (2) Mostly because we don't have to spend those 15s, we get down to a couple seconds here and there (3) This covers a few more cases that (2) didn't directly apply. Stuff like 'pull' has to do iter_changes (which we made faster in (2)) and then set_parent_trees (which was made faster in 3)	John Arbash Meinel	13 years ago
5847.4.3	Merge bzr.dev to get the other delta-based changes. Merge bzr.dev to get the other delta-based changes.	John Arbash Meinel	13 years ago
5847.4.2	uncommit is now down to 2.7s, document it fixing bug #780544 uncommit is now down to 2.7s, document it fixing bug #780544	John Arbash Meinel	13 years ago

Older »