* changes TypeResolver to accept ResolveTargetKind in all methods (#354) this is needed to be able to encode jagged arrays in instructions * removes redundant parameter from IApiDriverDefinitionsFactory.Field() overload (#354) * WIP: fix array SmokeTest for SRM (#354) * WIP: set field RVA value in SRM (#354) * adds FieldInitializationData struct to abstract field initializer values and field initialized with data embedded in the binary (#354) Issues: 1. RawDataType types are not emitted as nested types 2. Missing .pack/.size directives in these types 3. Missing .data directive with the actual byte values 4. field ArrayInitializationTests.intArrayField defined in the private implementation details type ?? * fixes nested types declaration in SRM (#354) * removes unused method (#354) this method is a left over from a previous refactoring which I forgot to remove * improves association of fields/methods with its owning type (#354) The list of methods/fields of a type depends on the type the list of methods/fields of the next method. If the type defines methods/fields, the token of the first method/field is used to define the list of methods/field. All methods/fields with a token >= the first method/field token and <= first method/field token of the next type are associated with the type. Unfortunately neither the "Common Language Infrastructure" docs nor the Sytem.Reflection.Metadata documentation is clear wrt that (at least IMHO). My understanding on how methods/fields gets associated with a type is derived from dumping metadata tables for varios scenarios and observing the actual values (a simple, but very effective way to dump the metadata tables is to use the tool from git@github.com:kant2002/MetadataDumper.git) * abstracts class size/packing size properties and take them into account in SRM (#354) * minor refactor * moves ParameterSpec type to TypeSystem namespace (#354) * fixes nested type definition in SRM (#354) In System.Reflection.Metadata, when processing type definitions a type reference is added and this type reference is used whenever that type is referenced. Type references for nested types should have the outer type as its resolution scope (as opposed to the main module handle) * disables test that depends on setting up .data section in the PE which I am having a hard time (#354) I think that setting that section is the last missing bit to be able to enable that test again * removes unused method in Mono.Cecil API driver (#354) * enables abstract class integration test and fixes accessibility being hardcoded as public on instance ctors (#354) * fixes base ctor registration/resolution for base types and enables InheritanceSameCompilationUnitTest (#354) * enables InheritanceTest (base extern type) (#354) * enables *InnerClassTest* and fixes a bunch of problems with order ot type/member/local variables definition (#354) * enables/fixes a bunch of simple interface definition (#354) * adds support for declaring properties in interface definition (#354) * trying to fix CI; it seems that after .NET 10 release CI image is using that version instead 9.x * adds support for (non generic) interface implementation (#354) for now only testing simple interface defined in BCL * enable simple value type test and type with static ctor (#354) * fixes parameter name missing in method definitions (#354) also includes some cleanups * enables simple readonly struct test (#354) * enables partial class integration test (#354) just noticed that both Mono.Cecil/System.Reflection.Metadata have problems in partial type handling (filled #357 to follow up) * enables expression integration tests that are passing on SRM (#358) * fixes string concatenation through operator + (#358) * enables TestNewPrimitive and TestValueTypeAddress integration tests (#358) also specify target of type resolution in some invocations of ITypeResolver.ResolveAny() to fix these failing tests * explicitly pass ResolveTargetKind in some invocations of ITypeResolver.ResolveAny() (#358) goal is to ensure the correct syntax is used in SRM when resolving types * fixes test expectation after changes in string concatanation logic introduced in 7b5d49e (#358) * fixes ternary operator integration test (#358) * improves parameter handling in SRM (#358) * fixes integration test expectation for tests that were correct in SRM (#358) * fix typo in test * big refactoring; adss TypeResolutionContext to ITypeResolver.ResolveX() (#358) * enable all fields integration tests (#358) * enables codeblock integration tests on SRM (#358) * enables ExpressionsTestCase.TestDoubleLocalVariableInitializationComplex() (#358) * fixes modulus operator test on SRM (#358) 1. Invalid variable name due to usage of ToDisplayString() in some scenarios 2. Invalid code generated due to missinterpretation of call target (operand was being interpreted as a string) * removes left over code used for debugging/investigation * renames parameter to reflect atual type * refactor: make SystemReflectionMetadataTypeResolver.RsolveForTargetKind() private there was a single reference to this method which has been replaced with a call to ResolveAny() instead * refactor: change order of tests to match the definition order (#358) this makes it easier to find next test to enable / fix * fix operand types not being correctly set when emitting instructions in multiple places * ensures fields are emitted in the same order as its parent definition types (#358) * fixes variable name for auto property backing field being invalid * fix field constant/initialization handling after adding field references (#358) after introducing a field reference for each visited field, ITypeResolver started resolving the field reference variable and the code ended trying to add constants/initialization data to that reference (as opposed to the field definition) * fixes token used as first field/method for types with no fields/methods (#358) in this scenarios, the index of the entry following the last entry in the respective table (field/method definition) should be used. * fixes assembly comparer not reporting missing fields (#358) * enables / reorganizes expression tests (#358) * fixes code generation when loading/storing local variables in some scenarios (#358) * enables post/pre increment tests in SRM (#358) * fixes range expression tests on SRM (#358) * improves exception message to make debugging/investigation easier * code cleanup * reenables expression tests already working on SRM (#358) * reenables integration method tests already working on SRM (#358) * reenables integration operators tests already working on SRM (#358) * reenables integration value type tests already working on SRM (#358) * reenables integration types tests already working on SRM (#358) * reenables integration miscelaneous tests already working on SRM (#358) * fixes when IlContext is set when defining events (#358) * removes unecessary generated code (#358) removed code had 2 issues: 1. It was specific for Mono.Cecil in a method that handles both SRM/Mono.Cecil 2. Some other part of the code (inside Mono.Cecil driver) was already adding exactly the same code. * removed some more Mono.Cecil specific code related to event handler processing (#358) hitted a wall since the generated event code depends on Interlocked.CompareExchange<T>() (i.e. a generic overload) and support for generics in the SRM driver is inexisting in Cecilifier * style: improves formating of generated code on SRM (#350) * feat: takes target API into account when generating .csproj * feat: remember last used api driver * test: disable tests for events on SRM test was enabled by mistake and is not working * refactor: minor typos and doc changes * refactor: use existing extension method when looking for method variables (#360) this ensures consistency since the same extension method is used when registering variables * refactor: improve forwarded method declararion handling (#360) Unfortunatelly the code mapping from IMethodSymbol -> MethodAttributes enum is incomplete and cecilifier code just assumes that the attributes are incorrect and resets them (at least with the correct value). At some point we should either improve the mapping and remove this reset code or remove the mapping (i.e. rely on the reseting code) and make it clear in the generated code that this reseting is happening * refactor: use standard approach to warn users * fix: handle event symbols when visiting nodes representing event raising (#360) * fix: ensures Invoke() method is invoked in implicit event rising (#359) * fix: applying attributes to enum/enum members (#358) 1. Removes some APIs only used by tests 2. Changes how attributes are processed in SRM. Specifically changes how these are handled for fields aligning it to how types are handled. * fix: attributes with explicit targets on SRM (#358) Enables test TestAttributesOnExplicitTargets * fix usage of attributes with arguments/named arguments on SRM (#358) * reduces verbosity by caching resolved types (as opposed to re-resolving multiple times) * bumps version to 2.25.0

Uh oh!

Go over generated code and fix/improve formatting #350

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions