Not sure what the flip does, but does it have to be "duplicated" everywhere? Shouldn't it be handled by the backend somehow?

Should be good when adding this.

Index: LICENSE.txt
===================================================================
--- LICENSE.txt	(revision 27384)
+++ LICENSE.txt	(working copy)
@@ -86,6 +86,9 @@
   source/third_party/tinygettext/
     zlib

You need to update https://code.wildfiregames.com/source/0ad/browse/ps/trunk/libraries/LICENSE.txt$23 + Apache-2.0 / (WTFPL OR CC0-1.0) AND Apache-2.0
Also https://code.wildfiregames.com/source/0ad/browse/ps/trunk/LICENSE.txt$69
source/third_party/vma/* is MIT but since there sub license we should just link to the readme file.

Partial review. + 3 lint warnings

LGTM.

cc @phosit

Looks like EOL issues... Patch sounds good though. Did you see any difference?

Looks okay.

A simple grep for @paramseems to reveal they are almost always in the .h in the class declaration (18 *.cpp files vs 150 *.h files) And most of the time it's because the class declaration is defined in the *.cpp files. We should move them in the FSM.hfile.

Follows the conventions. LGTM.

Will fix copyright years when committing.

Gettinh it back. Thanks for the interest.

Some missed includes (Could be more)

Is it okay to have so many raw pointers?

Still quite a few lint warnings and ifs without space :)

So does it break UnitMotionFlying?

Worth noting that Ubuntu/Debian do not support that version of enet (They are on 1.3.13). https://repology.org/project/enet/badges

Few style comments. (This is not a complete review the idea is to try to reduce the patch length as much as possible)
Not sure about camelCase functions we usually prefer PascalCase. Can you fix the linter warnings? I see const issues and the like also prefer === to ==. if ( not if(. Any reason you didn't follow OOP for your JS classes?
Please read https://trac.wildfiregames.com/wiki/Coding_Conventions
Any reason you don't use enet_address_get_host_ip ? Would remove the need for all those new functions.

It should, yes. Also gitignored I suppose.

Feel free to update the patch :)

Tested on gles with low quality and high quality material. Didn't see anything weird.

Currently building on GLES. Just a thought, maybe it'd be nice, to have the renaming and cleaning in a separate diff. I mean the change in Brushes.cpp seems kinda useless.

Some notes (can't test right now)

A quick grep says it's the only occurence in binaries/data/mods/**

From the CI while running the debug tests.

In PostMessage All that matters is that components are processed in the same order on all plateforms to keep determinism.

Maybe, or a special soundgroup container?

Nah I'm still bad at it.

Question: What happens in the destructor and in the ReloadGroup method if initialization failed?

Made a color comparer if that helps.

Thanks for the explanation!
Do you have any profiling evidence? Maybe playing 2400 sounds on unit combat demo https://github.com/0ADMods/trailer_tools/blob/main/maps/scripts/combat_demo_huge.js

Regarding the change of behavior in LoadSoundGroup which should have probably been named TryLoadSoundGroup I'm wondering if it should be static and return a CSoundGroup object instance instead.

Should build now (Migrated CI to Clang8)

It seems Clang 7 doesn't support std::launder. Which is weird because according to https://en.cppreference.com/w/cpp/compiler_support/17 it should. I've created a docker with Clang8, I will be looking into integrating it to the CI this week.

Wondering if getonlychildshouldn't be getroot ? After all there should only be one root

After a lengthy discussion I'm convinced this is better than the json solution.

Add the json dependency and okay

Now got this

I tried on git got this (I'm a few commits back maybe something did not get applied correctly) will try to update tomorrow if i have some time

Did you look at the generated assembly, or is it just a guess?
This patch also allows components to subscribe other components to messages. Is this wanted?

Did you check the translations scripts to make sure those new strings were translated ?

Now you can close it manually.