syscall: improve NewCallback documentation and panic message #26138
Comments
|
Ah, "output parameter" appears in the reflect documentation, which is probably where this terminology came from. That may be a bug in the reflect package documentation. :) |
|
@aclements I totally agree. syscall.NewCallback documentation is bad. syscall.NewCallback was used to allow for calling WindowProc ( https://msdn.microsoft.com/en-us/library/windows/desktop/ms633573(v=vs.85).aspx ) and service handler ( https://docs.microsoft.com/en-us/windows/desktop/api/winsvc/nc-winsvc-lphandler_function_ex ). Alex |
|
I could help with this. The documentation has to be updated in |
Please, do.
Both. I suggest you create CL for syscall first. And then we could copy it to x/sys. Thank you. Alex |
|
Change https://golang.org/cl/126035 mentions this issue: |
What version of Go are you using (
go version)?Tip (commit 65d55a1)
Does this issue reproduce with the latest release?
Yes.
syscall.NewCallback(andsyscall.NewCallbackCDecl) don't describe what's expected of the type of their argument beyond "it must be a function". But they do have requirements. For example, if you pass a function with no argument and no results, they panic with "compileCallback: function must have one output parameter". I don't actually know what this means a priori, and theNewCallbackdocumentation doesn't tell me. The term "output parameter" doesn't appear in the Go spec, and it could (and does in this case) mean a result value, or it could mean a C-style output parameter which is actually an input argument that points to where to put the result.We should document
NewCallback*'s requirements and clarify the panic message for when they're violated./cc @alexbrainman
The text was updated successfully, but these errors were encountered: