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

~bzr-pqm/bzr/bzr.dev » Changes from revision 3735.32.20

Rev	Summary	Authors	Date
3735.32.20	groupcompress now copies the blocks exactly as they were giv... groupcompress now copies the blocks exactly as they were given. One major concern here is that 'topo_sort' is not particularly stable. For example, given a history of a=>b=>c=>d and e=>f=>g=>h, it easily groups the contents as, h,a,b,c,d,e,f,g. Which is interleaving unrelated histories. This will actually cause transmission of the e-h group 2x, and cause effective 'bloat'. We can still tell 'get_record_stream' to remove some of this. Also, autopack still needs to be told to not re-use blocks.	John Arbash Meinel	15 years ago
3735.32.19	Get rid of the 'delta' flag to GroupCompressor. It didn't do... Get rid of the 'delta' flag to GroupCompressor. It didn't do anything anyway.	John Arbash Meinel	15 years ago
3735.32.18	We now support generating a network stream. We now support generating a network stream. It can then be read back in on the other end. Though we don't support inserting the byte-stream directly yet.	John Arbash Meinel	15 years ago
3735.32.17	We now round-trip the wire_bytes. We now round-trip the wire_bytes.	John Arbash Meinel	15 years ago
3735.32.16	We now have a general header for the GC block. We now have a general header for the GC block. It is pretty much just the trade between the GC Label structure and the formatting used by knit-delta-closure. Probably the main difference is compressing the bytes, but I found that to be pretty important for the GC Label, and it isn't hard to do.	John Arbash Meinel	15 years ago
3735.32.15	Change the GroupCompressBlock code to allow not recording 'e... Change the GroupCompressBlock code to allow not recording 'end'. As long as we know what gc block we need to read, we don't care about how long a record is. As it is already recorded in the data stream. We already have read the whole compressed block into memory, so we aren't saving a round trip. Though if we want trivial stripping of unused content, we may want to keep the extra data in the index. It still is good to have the functionality available to the extract() function.	John Arbash Meinel	15 years ago
3735.32.14	Move the tests over to testing the LazyGroupContentManager o... Move the tests over to testing the LazyGroupContentManager object.	John Arbash Meinel	15 years ago
3735.32.13	Allow the sha1 to be None for a factory. Allow the sha1 to be None for a factory.	John Arbash Meinel	15 years ago
3735.32.12	Add groupcompress-block[-ref] as valid stream types. Add groupcompress-block[-ref] as valid stream types. However, we now encounter that the versionedfile tests assume that Factory objects know the content's sha1 sum if we request them with delta_closure=False. I assume this is because we had been reading the knit content headers to get the sha1sum, but I think Andrew was planning on changing that anyway. The VF tests expose the lack of sha1, but otherwise things seem to work if we pass back the lazy block, rather than extracting it ahead of time.	John Arbash Meinel	15 years ago
3735.32.11	Add tests for the ability to do partial decompression withou... Add tests for the ability to do partial decompression without knowing the final length.	John Arbash Meinel	15 years ago
3735.32.10	test that we support reading from the gc blocks that didn't ... test that we support reading from the gc blocks that didn't have their lengths.	John Arbash Meinel	15 years ago
3735.32.9	Use a 32kB extension, since that is the max window size for ... Use a 32kB extension, since that is the max window size for zlib.	John Arbash Meinel	15 years ago
3735.32.8	Some tests for the LazyGroupCompressFactory Some tests for the LazyGroupCompressFactory	John Arbash Meinel	15 years ago
3735.32.7	Implement partial decompression support. Implement partial decompression support.	John Arbash Meinel	15 years ago
3735.32.6	A bit of reworking changes things so content is expanded at ... A bit of reworking changes things so content is expanded at extract() time.	John Arbash Meinel	15 years ago
3735.32.5	Change the parsing code to start out just holding the compre... Change the parsing code to start out just holding the compressed bytes.	John Arbash Meinel	15 years ago
3735.32.4	Change the byte representation of a groupcompress block. Change the byte representation of a groupcompress block. We now include the length of the compressed and uncompressed content. The important bit is the length of uncompressed content, as it allows to potentially pre-allocate the decomp buffer. But more important than that, is that it allows us to know how much of the content we will be processing during 'extract', just from the index entries and that header block.	John Arbash Meinel	15 years ago
3735.32.3	Start doing some direct GCVF tests. Start doing some direct GCVF tests.	John Arbash Meinel	15 years ago
3735.32.2	The 'delta' flag has no effect on the content (all GC is del... The 'delta' flag has no effect on the content (all GC is delta'd), However, it can mess up make_pack_factory because the GCVF doesn't think there are references, but the GCGraphIndex does.	John Arbash Meinel	15 years ago
3735.32.1	Fix the VF WalkingContent checks. Fix the VF WalkingContent checks.	John Arbash Meinel	15 years ago

Older »