In D5218#221743, @sera wrote:

Could repro again and the change fixes it

Great, thank you!

In D5218#221718, @sera wrote:

ERROR: Failed to compile shader 'shaders/glsl/compute_downscale.cs':
0:27(31): error: no function with name 'texture2D'
0:27(2): error: no matching function for call to `imageStore(image2D, ivec2, error)'; candidates are:

Should be fixed with the last update. To reproduce (I wasn't able locally) you need to choose OpenGL backend (restart after that), enable post processing, set scale to 200%.

Fixes compute_downscale.cs compilation

In D5218#221730, @sera wrote:

But might have been using opengl, maybe related to enabling preprocessing or some part thereof.

Scaling requires post processing be enabled indeed.

libxml2 v1.12.1 includes less header implicitly.

You mean v2.12.1?

In D5218#221718, @sera wrote:

After fiddling a while with this new feature and checking the console after I saw the above, but can't figure out how to reproduce it :)

Generally seems to work.

Could you post your userreport_hwdetect.txt?

In D5218#221693, @asterix wrote:

I suggest you post on forums like you did with testing Vulkan implementation.

Yeah, I will do after committing.

In D4425#221347, @phosit wrote:

So the queue limit was about memory-consumption? I thought it was about lag.

It was and still is about lag, but I'm also worrying that the queue capacity isn't limited anymore.

Have you measured performance for those changes?

In D5167#220539, @phosit wrote:

@vladislavbelov do you have a outstanding input?

In D5181#220476, @wraitii wrote:

I believe this also affected blinking territory, but I'm slightly less sure there - unlikely it would show up in game since it can only affect the center of a group of territory tiles, and unless I'm wrong we don't show borders when the borders are very small?

You mean like #1229?

In D5185#220429, @wraitii wrote:

You probably should fix the linter as well.

Yeah, thanks!

Could you make a test for that?

Not perfect but I can live with it.

This commit introduced serialization test failure (I've bisected it between 27104 and 27894):

Animation became noticeably out of sync with movement speed (reported by @abian):

In rP27908#61408, @phosit wrote:

How did you get the log? I didn't found anything on jenkins.

I believe it requires to login.

It'd be good to attach comparison images to such changes.

autobuild is broken (logs from Jenkins):

In D5163#219882, @phosit wrote:

I wrote it because of things like using StaticVector instead of std::optional. There are/where more but I don't remember.

LGTM.

I mean if we have a plan to restrict the CLogger scope by some class later then the change is unnecessary, if we don't have a plan then it's kind of ok (it's not perfect and in your terms a new dev might be "scared" because it looks like a workaround).

Fixed in rP27893.

A question: if there was a proper CApplication class (which controls lifetime of CLogger), would the LifetimeGuard be necessary?

Shouldn't it behave like the follow hotkey (F by default)?

In D5163#219492, @phosit wrote:

There are many things which "might be used in future". I don't like that.

I don't think that the word "might" is applicable here. The diff uses the code that was added.

Currently it's better to fix the behavior rather than making it always allocating. It was broken in rP23191. We construct CStrIntern with the same const char* many times per frame.

It should be mentioned in the changelog because there might be mods which use this function.

In D5127#219402, @Stan wrote:

From the ticket on Twitter, I'm not sure but I think the hangs are caused by the GC (Similar issue on RPI4):

Is there a more powerful combination with riscv64?

In D5154#219403, @phosit wrote:

Adding a class doesn't help much: we will still have to refactor code to be able to remove globals.

The main idea is to not make everything clean at once, but move towards by small steps. All polishing refactorings might be done afterwards. You might take a look my GL -> abstract API refactoring history. It took me more than a year (and it's still far from perfect). But it allowed me to add Vulkan support and make players life more comfortable in some cases (especially on macOS).

In D5115#217276, @abian wrote:

Thanks! Tested with "Segervillkor:", all okay.

Could you attach screenshots for long translations?

In D5127#218003, @xctan wrote:

Yes, for riscv64 support.

In D5154#219286, @phosit wrote:

Some libraries call such a class finally: https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines#Re-finally

Could you adjust your patch to have a context? You can find how to do that here: https://trac.wildfiregames.com/wiki/SubmittingPatches#Makingsomechanges