UOP Support fixes - #178
Merged
Merged
Conversation
- Map the uop tile visibility word onto both multi.mul int32s in both directions (0x0001 -> flags, 0x0100 -> extra) instead of collapsing them into one boolean; 8207 of 186695 shipped tiles carry 0x0100 - Skip build/multicollection/housing.bin when loading the uop - its first two DWORDs parse as a valid (multiId 7, 1 tile) header - RebuildTiles now only guarantees a tile on the anchor: it no longer deletes invisible (0x1) tiles or reorders the list, which on shipped data dropped 122 tiles and reshuffled 119 multis - Raise MaximumMultiIndex to 0x2710; the shipped uop holds ids up to 9000, which the old 0x2200 bound silently dropped - Write multi.idx only up to the highest populated multi, with extra 0 instead of -1, matching the client files and the packer - Keep per-tile component ids in a multi-components.txt sidecar so a mul -> uop repack does not strip a boat's tiller man, hatch and planks or a customisable house's doors - Write MultiCollection.uop as a version 4 container (first block at 0x28) with the 12 byte per-entry header, and set the entry hash to the Adler32 of that header as the shipped files do - Compress multi entries at zlib level 7 to land within 0.2% of the size the client's own packer produced - Require housing.bin and force Zlib for MultiCollection; reject Mythic, throw on compression failure and delete a partial output instead of leaving a half written uop behind
- Use the compressed length as the on-disk byte count for animation, map and gump uop entries; decompressedLength only equals it while the entry is stored, so compressed entries were read short - Decode animation frames stored as flag 3 (zlib wrapped Mythic) instead of handing Mythic bytes to the frame parser as pixels - Fix Entry6D.Extra: getter and setter disagreed on the packing order, so every UOP gump reported its width and height swapped - Add FileIndex.CacheDimensions - Seek and the indexer hand out a boxed copy, so writing dimensions to entry.Extra1 was silently discarded - Probe compressed gump entries for real content, so EA's 0x0 placeholder gumps (29, 33, 34, 37, 47, 49, 98 ...) stop listing as valid and failing to draw on a compressed client - Raise the gump id ceiling to 0x12000: 7.0.98.1 and later ship ids 69971..69985 above the old 0xFFFF bound. Gumps.Save now truncates gumpidx.mul to the last real row rather than padding with zeroes - Read and write the full 0x28 byte sound name block; the previous 32 bytes prefixed 8 bytes of per-build junk onto every sound - Fix the staidx tail loop in TileMatrix, which compared bytes against a block count and left short-file tail blocks zeroed instead of -1 - Reject compressed map uop entries with a clear message - the block reader slices straight into the file and can only handle stored ones - Packer: pick the container shape the newest client uses (sound moves to version 4), write 1000 entry blocks for version 5, emit the version 4 entry header for every type using that layout, and stamp one timestamp per file so a repack of the same input is identical - Packer: drop idx rows that are empty or point past the end of the mul rather than packing truncated data, and log what was dropped - Packer: pad artidx.mul to the 0x13FDC High Seas threshold so an unpacked art set is not downgraded to pre-Stygian-Abyss limits - Packer: seek map chunks through BaseStream (a large custom facet overflows int) and leave a genuinely larger custom map untrimmed - Export and import the trailing High Seas int32 of a multi tile in the CSV and XML formats, replacing the always-empty Cliloc column - Reuse UopUtils.HashFileName in the packer instead of a second copy of hashlittle2 - UopPacker UI: default compression per file type, warn before an unusual choice, and confirm before packing multis with no component sidecar; batch mode now uses multi.mul/multi.idx so single-file and batch modes interoperate
- Replace the register-named transcription of HashFileName with a direct port of Bob Jenkins' lookup3 hashlittle2, the function the client runs at 0x0042C9B2. Behaviour is unchanged - verified bit-for-bit against the previous implementation over 82k inputs covering every block-boundary length class - Add HashWord2, the word-oriented lookup3 sibling used for 32 bit word input, returning the client's low output word
- Mirror the gump content probe and dimension write-back from Ultima.Gumps, so both sides agree on which ids are valid; otherwise the compare tabs report differences that do not exist - Raise the second client's gump ceiling to 0x12000 and drive the compared id range from both loaded clients instead of a hardcoded 0x10000, so ids above 0xFFFF are reachable - Fix SecondEntry6D.Extra, which had a second pair of backing fields that only Extra wrote to, making a write through one view invisible to the other - Make SecondFileIndex disposable and re-open the mul stream in one place; SecondArt no longer closes the index's shared handle after every tile, which forced a re-open for the next one - Dispose the outgoing index and bitmap cache when a second client is loaded, after the tabs drop the cached instances they parked in PictureBox.BackgroundImage - Guard the SecondArt and SecondGump entry points against being called before a second client is loaded - Report an unreadable or compressed map uop once in a message box and unload it, instead of rethrowing from OnPaint and OnMouseMove - Re-apply the "differences only" filter after loading a second client
- A list box addresses rows through a 16 bit index in LB_ITEMFROMPOINT and in its scroll bar, so once the gump id ceiling moved to 0x12000 every row above 65535 aliased onto a row near the start of the list. With free slots shown the list is 73728 rows, so scrolling or paging into its top jumped back to around row 8192 - Row identity now lives in an ascending list of ids rather than in the control's items, keeping a row position and a gump id distinct when free slots are hidden - HasGumpId and Search binary search that list instead of walking every row - HasGumpId ran on each selection change for ids >= 50000 - The four insert paths collapse into one that keeps the list ordered - Paint the selection from the tracked row: the item handed to DrawItem is a fresh one built in RetrieveVirtualItem, so its State never carries the selection and every row drew as selected - Keep the list pane out of the form's resizing (FixedPanel.Panel1, as the other tabs do) and clamp it to 450 pixels; a row is a 105 pixel thumbnail plus two lines of text, the rest was empty background - Guard the extract image entries against an empty selection
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
multi.mulint32s instead of one boolean (8207 of 186695 shipped tiles used to lose a bit)multi-components.txtsidecar, so repacking no longer strips a boat's tiller man, hatch and planks or a house's doorshousing.binis no longer parsed as multi 7, and the multi id ceiling now covers the ids up to 9000 that ship in MultiCollection.uophousing.bin/ component sidecar are prompted forgumpidx.mulandmulti.idxare truncated at the last real entry instead of padded