self. completion suggesting incorrect code

[am45-001]

Tested versions

Tested in v4.2.stable.official [46dc277], v4.2.1.stable.official [b09f793].

System information

Godot v4.2.stable - Windows 10.0.19045 - Vulkan (Forward+) - dedicated Radeon RX 580 Series (Advanced Micro Devices, Inc.; 31.0.21905.1001) - AMD Ryzen 5 3600 6-Core Processor (12 Threads)

Issue description

The self keyword are assumed to be classes by code completion (unlike before in v4.1 and previous versions), and only display constants, static functions, and new().

Edit: After further investigation by @Vilcrow, I jumped the gun on Singletons suffering the same issue as the self keyword. Apologies to @MitrB for stealing your thunder on this one.

Steps to reproduce

• In any project open in version 4.2, create or open a script.
• Type "self." and observe the code completion suggestions.
• You should see something like this:
  

Minimal reproduction project (MRP)

BugProject.zip

[yosoyfreeman]

Confirmed in Linux too.

[dalexeev]

• See also self. autocomplete broken #85568.

[grumpOldman]

I'm pretty sure this is a regression as I could use this before and waiting for 4.3 seems odd.

[AThousandShips]

It being marked for 4.3 doesn't mean you have to wait for 4.3, if a compatible bug fix can be made it can be cherry picked for 4.2.x

We only assign a released milestone to bugs if they are unique to that milestone, like a regression specifically in 4.2 that isn't in the development branch

[am45-001]

I'm pretty sure this is a regression as I could use this before and waiting for 4.3 seems odd.

as a workaround you can use an external code editor like VSCode, and it will get around the missing instance variable and method issues. It will still kind of treat self and singletons like classes, but you'll actually be able to code with it.

[am45-001]

I'm pretty sure this is a regression as I could use this before and waiting for 4.3 seems odd.

as a workaround you can use an external code editor like VSCode, and it will get around the missing instance variable and method issues. It will still kind of treat self and singletons like classes, but you'll actually be able to code with it.

Well, this is awkward, I just tried this again and it doesn't work [edit] for self specifically, singletons work fine.[/edit]

Edit: the real workaround is casting self as the class it is, which will force the completer to treat it as an instance.

Edit 2: The final workaround I've come up with is to assign self to a variable while casting self as the class it is. It treat the variable as an instance otherwise.

[HolonProduction]

Did a bisect. Was introduced in 56e2fad. (#73196)
The changes to gdscript_editor.cpp don't make sense to me and don't seem to have gotten any discussion during the PR. Could you explain how they relate to the issue @Vilcrow

[Vilcrow]

@HolonProduction changes to gdscript_editor.cpp have been made to correct #72353 which led to the incorrect completion. I've tried to figure it out a little bit now, but so far there are no results. I will try to fix this bug as soon as I have time.

[HolonProduction]

Yeah I read the isssue and you also described what caused the issue in you PR. What I don't understand is how these changes:

if (p_context.type != GDScriptParser::COMPLETION_SUPER_METHOD) {
    r_type.type = p_context.current_class->get_datatype();
} else {
    r_type.type = p_context.current_class->base_type;
}

are related to the cause of the issue that you described in the PR (you said it had something todo with a wrong position being used in CodeEdit).

This issue (as in #85833) results from p_context.current_class->get_datatype() being a meta type (the class, not an instance of the class; that is why you only see new. So this should just be switched back to how the type was constructed before.

I also don't see what you are trying to do with p_context.type != GDScriptParser::COMPLETION_SUPER_METHOD. Could you give an example case were this would be false.

So yeah IMO I would just try to revert all changes to gdscript_editor.cpp. But you propably know better what this code is supposed to do.

[MitrB]

@am45-001 No need to apologise, I wasn't going to be very involved either way

[am45-001]

@Vilcrow would it be appropriate for me to upload the custom version of 4.2.1 with your fix here? Or would I need to publish it somewhere else while we wait for the next version?