[4.3] Hot reload of @tool scripts stops working after first reload

[dmlary]

Tested versions

• Reproduces in v4.3.stable.official [77dcf97]
• Fixed in v4.4.dev.custom_build [621cadc] (confirmed fixed in this commit; prior commit still has issue)

System information

Godot v4.3.stable - macOS 14.6.1 - Vulkan (Forward+) - integrated Apple M3 Max - Apple M3 Max (16 Threads)

Issue description

@tool scripts do not hot reload after the first hot reload when using an external editor. During the first hot reload the following error is generated:

modules/gdscript/gdscript.cpp:745 - Condition "!p_keep_state && has_instances" is true. Returning: ERR_ALREADY_IN_USE

Note this is not the same issue as #97238 which is 4.4 specific.

Steps to reproduce

• Create a new project

• Add a Node3D

• Attach a script to that Node3D

• Make script a @tool script, and add a print to the _ready() function; save it

• Switch back to godot editor

• Save & reopen scene in editor (seems to be necessary on macos right now)

• Confirm print from _ready()

• Edit script to change print message; save script

• Switch back to godot editor

• Godot will display the ERR_ALREADY_IN_USE error

• Close and reopen scene and updated _ready() message will be printed

• Edit script, changing the print message again: save script

• Switch back to godot editor

• Godot will not print any new message

• close & reopens scene and the message from the previous edit will be shown

Minimal reproduction project (MRP)

hot-reload-issue.zip

[AThousandShips]

Thank you for your report, consolidating in:

• ERROR: Condition "!p_keep_state && has_instances" is true. Returning: ERR_ALREADY_IN_USE #97138

This has been fixed, and will be available in 4.4.dev3

[dmlary]

@AThousandShips This bug is present in 4.3. I put a comment in #97168 asking for a cherry-pick to 4.3.1, and they asked for a separate 4.3 issue.

[AThousandShips]

My bad, reopening, didn't notice

[dmlary]

Also confirmed that commit from #97168 (621cadc) fixes the issue. Prior to that commit, the problem still reproduces.

[AThousandShips]

Prior to that commit, the problem still reproduces.

Did you confirm it was actually occurring in between? Or did you cherry-pick the commit to 4.3?

[dmlary]

@AThousandShips bah, you're right. I only confirmed it was fixed on 4.4 branch. I'll see if I can cherry pick it.