compiler: add support for the go keyword on interface methods #2202
Conversation
aykevl
force-pushed
the
go-on-interface
branch
from
4abe316 to
e2f6ad2
Compare
|
Updated to fix merge conflicts and (hopefully) fix the CI failures. |
This commit simplifies the IR a little bit: instead of calling pseudo-functions runtime.interfaceImplements and runtime.interfaceMethod, real declared functions are being called that are then defined in the interface lowering pass. This should simplify the interaction between various transformation passes. It also reduces the number of lines of code, which is generally a good thing.
This is a feature that was long missing, but because of the previous refactor, it is now trivial to implement.
aykevl
force-pushed
the
go-on-interface
branch
from
e2f6ad2 to
d3043c7
Compare
| uintptrType := mod.Context().IntType(llvm.NewTargetData(mod.DataLayout()).PointerSize() * 8) | ||
|
|
||
| defer llvm.VerifyModule(mod, llvm.PrintMessageAction) | ||
| // Look up the (reflect.Value).Implements() method. |
There was a problem hiding this comment.
I think I am missing something: why can't we look this function up by name?
There was a problem hiding this comment.
Because it is named something like interface:{Align:func:{}{basic:int},AssignableTo:func:{named:reflect.Type}{basic:bool},Bits:func:{}{basic:int},Comparable:func:{}{basic:bool},ConvertibleTo:func:{named:reflect.Type}{basic:bool},Elem:func:{}{named:reflect.Type},Field:func:{basic:int}{named:reflect.StructField},FieldAlign:func:{}{basic:int},Implements:func:{named:reflect.Type}{basic:bool},Key:func:{}{named:reflect.Type},Kind:func:{}{named:reflect.Kind},Len:func:{}{basic:int},Name:func:{}{basic:string},NumField:func:{}{basic:int},NumMethod:func:{}{basic:int},Size:func:{}{basic:uintptr},String:func:{}{basic:string}}.Implements$invoke and therefore depends on the precise set of methods in src/reflect/type.go.
So yeah, we could check that name directly, but every change to the reflect.Type interface would also need a change here. Doable, but kind of a burden.
There was a problem hiding this comment.
Oh, I think I misunderstood what this was doing. Alright.
|
I think most of this looks fine, just a bit confused (see the question i posted above) |
|
Might have been better to rebase @aykevl can you please take a look? Thanks! |
Most of this PR is a refactor of interface lowering, to make future changes possible. The result of it is that it is now trivial to add support for the go keyword on interface methods, as requested in #1961.