db289c1a1a
CodeGen v2 permits template parameters to be qualified. This means that if we call `make_field` with a template parameter it will be qualified. However, we don't qualify the types when generating meta functions such as `NameProvider` and `TypeHandler`. This means these qualified types don't match up with the expected type. Use `std::decay_t` when forwarding the type to `NameProvider` and `TypeHandler` so they're always the base type that they were generated with. Most of this is covered by `make_field`, but there are direct references to `TypeHandler<Ctx, T>` in a lot of `TypeHandler::type` fields. Fix the problematic types manually for now, there may need to be a better solution with meta functions for this in the future. Test Plan: - CI - Added a test for `std::unique_ptr<const uint64_t>` to exercise this. Failed before, passes after. - Added a test for `std::unique_ptr<const std::vector<uint64_t>>` to test a non-primitive type. Failed before, passes after. |
||
|---|---|---|
| .. | ||
| array_type.toml | ||
| boost_bimap_type.toml | ||
| caffe2_blob_type.toml | ||
| cxx11_list_type.toml | ||
| cxx11_string_type.toml | ||
| deque_list_type.toml | ||
| f14_fast_map.toml | ||
| f14_fast_set.toml | ||
| f14_node_map.toml | ||
| f14_node_set.toml | ||
| f14_value_map.toml | ||
| f14_value_set.toml | ||
| f14_vector_map.toml | ||
| f14_vector_set.toml | ||
| fb_string_type.toml | ||
| folly_iobuf_queue_type.toml | ||
| folly_iobuf_type.toml | ||
| folly_optional_type.toml | ||
| folly_small_heap_vector_map.toml | ||
| list_type.toml | ||
| map_seq_type.toml | ||
| multi_map_type.toml | ||
| multi_set_type.toml | ||
| optional_type.toml | ||
| pair_type.toml | ||
| priority_queue_container_adapter_type.toml | ||
| queue_container_adapter_type.toml | ||
| README.md | ||
| ref_wrapper_type.toml | ||
| repeated_field_type.toml | ||
| repeated_ptr_field_type.toml | ||
| seq_type.toml | ||
| set_type.toml | ||
| shrd_ptr_type.toml | ||
| small_vec_type.toml | ||
| sorted_vec_set_type.toml | ||
| stack_container_adapter_type.toml | ||
| std_map_type.toml | ||
| std_unordered_map_type.toml | ||
| std_unordered_multimap_type.toml | ||
| std_variant.toml | ||
| string_type.toml | ||
| thrift_isset_type.toml | ||
| try_type.toml | ||
| uniq_ptr_type.toml | ||
| unordered_multiset_type.toml | ||
| unordered_set_type.toml | ||
| weak_ptr_type.toml | ||
Container Type Definition Format
This document describes the format of the container definition files contained in this directory.
info
-
type_nameThe fully-qualified name of the container type. This is used to match against the names of types contained in an executable's debug information.
-
headerThe name of the C++ header file in which this container is defined.
-
stub_template_paramsThe indexes of template parameters which do not represent types stored within this container. These parameters will not be recursed into and measured by OI.
-
underlying_container_indexOnly used for container adapters. Points OI to the template parameter representing the underlying container to be measured.
-
required_featuresA set of feature names such as
tree-builder-v2which must be enabled for this container description to be included. Currently only supported with CodeGen v2 as that's their only use case and an implementation for CodeGen v1 would be untested.
codegen
-
declC++ code for the declaration of a
getSizeTypefunction for this container. -
funcC++ code for the definition of a
getSizeTypefunction for this container. -
processor.typeThe static type that will be filled in by this particular processor. Multiple processors are allowed and recommended.
-
processor.funcC++ code for the function body of a function with this signature:
void(result::Element& el, std::stack<Inst>& ins, ParsedData d)which is supplied with the data of the paired static type. -
traversal_funcC++ code for the function body of a function with this signature:
static types::st::Unit<DB> getSizeType(const T& container, ST returnArg)whereSTdefines the combined static type of each supplied processor. -
extraAny extra C++ code to be included directly. This code is not automatically wrapped in a namespace, so definitions from different container TOML files will collide if they share the same name.
## Changes introduced with TreeBuilder V2
declandfuncfields are ignored when using-ftree-builder-v2. TheTypeHandleris constructed fromtraversal_funcfield andprocessorentries.
Changes introduced with TypeGraph
typeNameandmatcherfields have been merged into the single fieldtype_name.nsnamespace definition is no longer required.underlyingContainerIndexrenamed tounderlying_container_index- The options for measuring / stubbing / removing template parameters have been reworked:
- By default all parameters are now measured. This includes the tail parameters in a variadic template.
- There is one option to specify template parameters which should not be measured:
stub_template_params. The type graph code will automatically determine what type of stubbing to apply to each parameter in this list.
Deprecated Options
-
ctypeA reference to the enum
ContainerTypeEnumin OI's source code. This is no longer required with Tree Builder v2. -
handlerC++ code for the definition of a
TypeHandlerclass for this container. Now generated fromtraversal_funcandprocessorentries. -
numTemplateParamsThe first
numTemplateParamstemplate parameters for this container represent the types for the data we want to process. If this is not set, use all a container's template parameters. All of a container's parameters will still be enumerated and output in CodeGen, regardless of this setting. -
allocatorIndexIndex of a template parameter representing an allocator. It will be not be used when CodeGenning this container.
-
replaceTemplateParamIndexIndexes of template parameters to be stubbed out, i.e. replaced with dummy structs of a given size. Used for custom hashers and comparers.