Instead of bailing when children have duplicate filesystem names,
syncback now resolves collisions by appending incrementing suffixes
(e.g. Foo, Foo1, Foo2). This handles both init-renamed children and
any other name collisions. Meta stem derivation is now path-based
to correctly handle collision suffixes and dotted names.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Drop the strip_suffix(extension) approach for computing adjacent meta
file names. Instead, use the instance name directly (slugified if it
has invalid filesystem chars, prefixed with '_' if it's "init"). This
is the same logic as the original code plus init-prefix handling, and
correctly preserves dots in instance names like "Name.new".
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Two bugs:
1. Meta stem fallback used raw instance name (unslugged), so names with
forbidden chars like '/' would create bogus directory components in
the meta path. Fix: fallback now slugifies + init-prefixes, matching
name_for_inst.
2. AdjacentMetadata name check used split('.').next() to extract the
filesystem stem, breaking dotted names like "Name.new" (stem became
"Name", mismatched the instance name, wrote an unnecessary name
property). Fix: check the conditions that cause name_for_inst to
diverge (invalid chars or init-prefix) directly instead of comparing
path stems.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous fix used split('.').next() to get the meta stem from the
snapshot path, which only takes the first dot-segment. This broke names
containing dots (e.g. "Name.new" → "Name.new.luau" would produce
"Name.meta.json" instead of "Name.new.meta.json").
Strip the full middleware extension (e.g. ".server.luau", ".txt") from
the snapshot path filename instead. This correctly handles all cases:
Name.new.luau → Name.new → Name.new.meta.json
_Init.server.luau → _Init → _Init.meta.json
Name.new.txt → Name.new → Name.new.meta.json
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When a script/txt/csv child is renamed by name_for_inst (e.g. "Init" →
"_Init.luau"), the adjacent meta file must follow the same name. All
three callers were using the Roblox instance name to construct the meta
path, producing "Init.meta.json" instead of "_Init.meta.json" — which
collides with the parent directory's "init.meta.json" on
case-insensitive file systems.
Fix by deriving the meta stem from the first dot-segment of the
snapshot path file name, which already holds the resolved name.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* Mostly mechanical port bits
* Almost there
* It builds again!
* Turn on all the code again
* Tests compiling but not passing
* Stub work for value resolution
* Implement resolution minus enums and derived properties
* Implement property descriptor resolution
* Update referent snapshots
* Update unions test project
Using a place file instead of a model yields better
error messages in Roblox Studio.
* Add easy shortcut to testing with local rbx-dom
* Update rbx-dom
* Add enum resolution
* Update init.meta.json to use UnresolvedValue
* Expand value resolution support, add test
* Filter SharedString values from web API
* Add 'property' builder method to InstanceSnapshot
* Change InstanceSnapshot/InstanceBuilder boundary
* Fix remove_file crash
* rustfmt
* Update to latest rbx_dom_lua
* Update dependencies, including rbx_dom_weak
* Update to latest rbx-dom
* Update dependencies
* Update rbx-dom, fixing more bugs
* Remove experimental warning on binary place builds
* Remove unused imports
* First take at flattening middleware for simpler code and better perf
* Undo debug prints
* Fix using wrong path in snapshot_from_vfs
* Disable some broken tests
* Re-enable (mistakenly?) disabled CSV test
* Fix some tests
* Update project file tests
* Fix benchmark
* vroom
* Port dir middleware
* Filter rules
* Directory metadata
* Project support
* Enable Lua support
* StringValue support
* CSV
* rbxm, rbxmx, and rbxlx
* JSON models
* Clean up some warnings
* Strip out PathMap
* Unwatch paths when they're reported as removed
* Fix 'rojo upload' behavior
* Upgrade to Insta 0.13.1
* Update dependencies
* Release 0.6.0-alpha.2
* Fix bad merge
* Replace MemoryBackend with InMemoryFs
* Sledgehammer tests into passing for now
* Txt middleware
* Update easy snapshot tests
* Lua tests
* Project middleware tests
* Try to fix test failures by sorting
* Port first set of serve session tests
* Add InMemoryFs::raise_event
* Finish porting serve session tests
* Remove UI code for introspecting VFS for now
* VFS docs
* Drop plugin context on the floor
* Remove redirect from old context to new context
* Pass InstanceContext via & instead of &mut reference
* Re-use context value in ChangeProcessor from metadata
This PR refactors all of the methods on `Vfs` from accepting `&mut self` to
accepting `&self` and keeping data wrapped in a mutex. This builds on previous
changes to make reference count file contents and cleans up the last places
where we're returning borrowed data out of the VFS interface.
Once this change lands, there are two possible directions we can go that I see:
* Conservative: Refactor all remaining `&mut Vfs` handles to `&Vfs`
* Interesting: Embrace ref counting by changing `Vfs` methods to accept `self:
Arc<Self>`, which makes the `VfsEntry` API no longer need an explicit `Vfs`
argument for its operations.
* Change VfsFetcher to be immutable with internal locking
* Refactor Vfs::would_be_resident
* Refactor Vfs::read_if_not_exists
* Refactor Vfs::raise_file_removed
* Refactor Vfs::raise_file_changed
* Add Vfs::get_internal as bits of Vfs::get
* Switch Vfs to use internal locking
* Migrate all Vfs methods from &mut self to &self
* Make VfsEntry access Vfs immutably
* Remove outer VFS locking (#260)
* Refactor all snapshot middleware to accept &Vfs instead of &mut Vfs
* Remove outer VFS Mutex across the board
