Support recursive types#2415
Open
eboasson wants to merge 30 commits into
Open
Conversation
Some generated files are committed in the repository, and for those the generated file contains hashes of the exact input version so we can ensure things stay in sync. In some cases it appended spaces, this commit fixes that.
Load MurmurHash3 32-bit blocks via memcpy instead of casting arbitrary byte input to uint32_t*. The public ddsrt_mh3 API accepts const void* data with no alignment precondition, so callers may pass byte buffers that are not 4-byte aligned.
Accept octet as the XML spelling for byte and report unknown non-basic types instead of dereferencing a failed lookup. Fix pre-existing sample layout calculations for int8, sequences, arrays, external members, and union data alignment. Also fix loading multiple XML type libraries and clean up parsed DOM trees, topic descriptors and raw sample buffers once they are no longer needed.
Add a pause option for discovery-sensitive experiments and an option to load and print the minimal TypeObject alongside the complete one.
Treat a final empty aggregate as FINAL when determining type extensibility from the CDR stream descriptor, allowing empty top-level structs without ADR/DLC/PLC instructions before RTS. Reuse the member-id hash table while walking descriptor MID operations and free duplicate MID records instead of leaking them.
Add a dedicated type-library trace category to the logging mask and configuration tables, plus a local TYPELIB_TRACE helper for the recursive type-library work.
Add recursive type-library support for strongly connected components, including received-SCC graph validation, transactional SCC installation, shared type-identifier identity helpers and hardened TypeLookup SCC reply handling.
Add configuration plumbing for disabling recursive types. When recursive dynamic type construction is rejected, abort partially built unidentified graphs cleanly and preserve the public dynamic-type error state.
Collect XML type declarations before materializing them, allowing struct and union predeclarations and typedef aliases to participate in recursive type graphs. Teach the dynsub type cache, type printing, sample scanning, sample printing, sample comparison and cleanup paths to follow SCC TypeIdentifiers and recursive references.
Clean up rejected TypeObjects, failed type-info and type-map generation, canonicalized dynamic-type graphs and SCC activation when recursive type graphs are rejected or only partially resolved. Hold TypeMap SCC members during import so components can refer to later slots without duplicate slot creation or use-after-free during failed cleanup.
Complete TypeObject import now copies type, collection and discriminator annotations back into xt_type storage, matching the TypeObject generator so regenerated complete TypeIdentifiers remain stable after import. Also make annotation ownership symmetric in copy/fini paths, deep-copy enum literal details for assignability temporaries, validate bitset member flags, and classify/generate plain collection TypeIdentifiers from their effective equivalence kind.
Reject invalid struct base chains, accept KeyHolder(T) for non-aggregate union case member types, and return BAD_PARAMETER instead of aborting when externally supplied TypeObjects cannot be encoded for hashing.
Gate fuzz target discovery only on LIB_FUZZING_ENGINE so local and OSS-Fuzz-compatible builds can expose fuzzer targets without building idlc.
This eliminates warnings from recent CMake versions.
Add generated assignability pairs and fix the fuzzer preconditions by keeping synthetic generated graph nodes in PARTIAL_RESOLVED state so validation still walks their nested hash references.
Build TypeInfo/TypeMap directly from validated generated type graphs, mutate them through existing corruption hooks, and exercise TypeInfo/TypeMap accessors and roundtrip fallback paths.
Extend the LPM type graph model with bitsets and annotation controls, and exercise imported type accessors, TypeIdentifier/TypeObject helpers, TypeInfo/TypeMap import paths and resolved typelib assignability.
Exercise collection TypeIdentifiers and collection TypeObjects directly, bias generated TypeInfo/TypeMap roots toward DDS top-level types with collection dependencies, and add focused invalid mutations for map keys, collection headers and stale collection TypeIdentifier references.
Registering equivalent recursive dynamic types in a different order could canonicalize a constructed type to an existing type-library entry while copying a full XTypes graph over an entry that already had one. For duplicate collection types this left an extra owned dependency reference behind, which showed up as a non-empty type library at shutdown. xmltype/dyntypelib exposed the ordering, but the bug is in the core dynamic type/type-library lifecycle. Only copy the constructed XTypes definition into an existing entry when that entry is a missing-definition placeholder, and reset the placeholder id to the canonical id computed during finalization. Add a direct dynamic type regression that registers the duplicate recursive sequence shape in the xmltype-exposed order.
If a char8 input is longer than 1 byte - assume user is trying to input char as an integer value
On behalf of test code on Windows.
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
2 participants
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.
The title is the high-level summary. In detail, there are of course various small other matters (a tweak to
.gitignore, some CMake stuff, etc.) that happened while working on this and that I couldn't be bothered to turn into separate commits.