Remap script path when registering class. #41025
Conversation
Was causing `class_name`-defined scripts to not being loaded in exported games due to the remap from `*.gd` to `*.gdc`/`*.gde`.
Faless
added
bug
topic:gdscript
regression
cherrypick:3.x
I think this is due to #40781 (CC @Rubonnek). I'm surprised about myself why I didn't think about this before because I've done similar path remapping workarounds/hacks: #33266 (basically the same issue but from the other end). 😃
According to #39797, yeah it indeed may be slow, and may impact #39841, but it may be different for If your solution works, I think it's ok then, especially since the task is to fix the regression first, then we should re-evaluate the way scripts are remapped, because it's very easy to break something with the current mechanism. |
| @@ -163,7 +164,7 @@ void ScriptServer::init_languages() { | |||
|
|
|||
| for (int i = 0; i < script_classes.size(); i++) { | |||
| Dictionary c = script_classes[i]; | |||
| if (!c.has("class") || !c.has("language") || !c.has("path") || !FileAccess::exists(c["path"]) || !c.has("base")) { | |||
| if (!c.has("class") || !c.has("language") || !c.has("path") || !FileAccess::exists(ResourceLoader::path_remap(c["path"])) || !c.has("base")) { | |||
There was a problem hiding this comment.
Would FileAccess::exists work in exported project anyways? May be it's better to omit the check in tools=no builds altogether?
EDIT: works, but I'm curious why? FileAccess works on pck files? 👀
EDIT2: yes:
Lines 53 to 56 in ec678c5
|
TBH I'm not familiar with this code and the release target is broken on master so I have not tested the export. But yes, if you need to check if the file exists you need to use the remapped path. |
|
Thanks! |
|
Cherry-picked for 3.2.3. |
akien-mga
removed
the
cherrypick:3.x
|
Reverted with #41501 as per #40584 (comment). |
Was causing
class_name-defined scripts to not being loaded in exported games due to the remap from*.gdto*.gdc/*.gde.Note: I could have used
ResourceLoader::existsdirectly, but it seems to come with a more complex logic (thus slower).But I'd like an opinion on this from someone more knowledgable on the topic (since e.g. ResourceLoader would allow us to specify the exptected type). @vnen @Xrayez ?
Fixes #40991