Gotta have to check those includes, lots are missing.

I deliberately omitted thous wich are already included in Execution.h. I'll add them in the next diff

I don't think the condition is useful without context. But i'm ok if i should leave it that way.

In D4761#202710, @vladislavbelov wrote:

Have you checked MSVC output in case of an error?

In D4743#201822, @phosit wrote:

Acording to https://trac.wildfiregames.com/wiki/Coding_Conventions#Errorreporting we should not use LOGWARNING:
"only use things defined in source/lib/"

It is already UTF-8‽ So it should be possible to replace wchar_t by char in most places.

Realy nice improvement... and less controversal than the allocator.

Acording to https://trac.wildfiregames.com/wiki/Coding_Conventions#Errorreporting we should not use LOGWARNING:
"only use things defined in source/lib/"

In D4743#201740, @Stan wrote:

Might try filesystem if that works. I don't remember if nothing worked, or only a subset.

In D4743#201732, @Stan wrote:

They trigger Assertion popups. You're free to ignore suppress them. Except in Atlas where it goes south and crashes the entire game.

Handle error also on windows (hopefully i can't test ^^)

Why do WARN_ON_ERROR and WARN_RETURN abort the program? shouldn't they warn as the name says?

In D4743#201709, @Langbart wrote:

Before applying the patch I would get the following output

Terminal output

❯ pyrogenesis
TIMER| InitVfs: 2.513 ms
FILES| Main log written to '/Users/paria/Library/Application Support/0ad/logs/mainlog.html'
FILES| Interesting log written to '/Users/paria/Library/Application Support/0ad/logs/interestinglog.html'
TIMER| CONFIG_Init: 6.691 ms
Function call failed: return value was -110301 (Error during IO)
Location: file_system.cpp:123 (GetDirectoryEntries)

Call stack:

(error while dumping stack: Function not supported)
errno = 2 (Error during IO)
OS error = ?

I also got an "Error during IO" but why? it is not EIO?

Include logger

fix test

In D4705#201569, @Stan wrote:

Header date, the test has been lost,

In D4735#201571, @marder wrote:

In D4735#201544, @phosit wrote:

In most civs this looks weird: wall; tower; tower; wall; tower
IMO we shouldn't sort by phase.

Could you expand on that / how would you sort it ?
Rather by function e.g. eco buildings p1, eco buildings p2, military buildings p1 .... Ect?

• braces
• function names
• CStr

Did i miss something?

In most civs this looks weird: wall; tower; tower; wall; tower
IMO we shouldn't sort by phase.

I forgot to add the test file

• Test
• helper function
• return std::string

In D4730#201162, @phosit wrote:

I'll try to embed the SharedState in to Future. or i join the diff with one that adds continuation.

Never mind my comments about lib/rand.h
just implement CUserReporter::LoadUserID() in terms of ps_generate_guid()

I would

• rewrite lib/rand.h to use "better" randomnes (and std distribution).
• use this new implementation were sys_generate_random_bytes is used now.

In D4705#201372, @Stan wrote:

Somewhere in lib we have some stuff about random used by the soundmanager there IIRC.

It doesn't realy solve the problem.

In D4705#201366, @Stan wrote:

Can you make this a helper function or something to remove the duplication ?

Apply comments from vladislavbelov

Use falback if entroby is zero.

In D4705#201348, @vladislavbelov wrote:

Don't we have regression here?

std::random_device may be implemented in terms of an implementation-defined pseudo-random number engine if a non-deterministic source (e.g. a hardware device) is not available to the implementation. In this case each std::random_device object may generate the same number sequence.

So it might produce the same sequence for each std::random_device. So more players might have the same ID on the same platform version.

In D4730#201118, @jprahman wrote:

Usually the overhead from std::function<> comes from one of two sources:

1. Heap allocation for the functor/function pointer/callable std::function is wrapping (https://devblogs.microsoft.com/oldnewthing/20200514-00/?p=103749), including CPU cache misses when following the pointer
2. Indirection when invoking the type erased std::function through a virtual method call

I'd expect 1. to dominate in most cases given a cache miss all the way out to main memory takes ~100+ clock cycles. If we really want to avoid std::function overhead, we can write our own std::function lookalike with more in-situ storage to avoid dereferencing a pointer in the hot path (https://jprahman.substack.com/p/software-performance-cache-locality). Alternatively, lambda captures used for callbacks can have fewer captures to fit in the typical 16 bytes of in-situ storage most std::function implementations offer.

• One coment from Stan
• Lint

optional -> unique_ptr
ENSURE more in the constructor
reset to nullptr in the destructor

remove repackaging in to tuple

In D4728#200776, @Stan wrote:

Does it make that code slower?

Should g_Profiler2 be reset in the destructor? to nullptr or to the pointer g_Profiler2 was bevore the constructor was called?.

I don't see a problem with this going into a26

forgot context
rename NameCmp -> NameLess

if constexpr -> #if

I messed it up with another diff

coments from Vladislav
CStr could not be constructed from std::string_view. Now CStr inherits all constructors from std::string

some thaughts on algorithm, feel free to ignore them :)

Are we talking about introducing those icons. I like the them but they are a litle to big.
Sometimes the icon is over the icon of the CC and sometimes below it.

rP26939 solves the "main" issue but this diff go's further

suggested by wraitii: value readability over optimization

remove two unused members of Impl

In my understanding is a singleton-class a class made for only one object: static Instance method; editing global/static "single" variables in constructor/destructor... by single i mean if it stores this inside a vector(to track all instances) its not a singleton-class but when it stores this in a single global pointer it is a singleton-class since it "shadows" the old pointer.
A singleton(-object) i would refer to as an object of a singleton-class. Otherwise every global variables would be a singleton since there is only one of it.

• assigne g_Profiler2 also in the tests
• edit some coments in Profiler2.h

In D4682#199514, @wraitii wrote:

Per https://trac.wildfiregames.com/wiki/Coding_Conventions, we have Prefer global variables over singletons, because then they're not trying to hide their ugliness. It's definitely something I kind of like, because singletons by design make it annoying to have two of something, and that can occasionally be useful.

I like basically all of the changes here, but I think you should just make it a regular global.

not a singleton anymore.

Yes looks good, thanks for solfing the tiket.

Remove note about msvc

adjust comments in main

The last fix was kind of overkill

hopefuly fix for msvc-debug

good

new CRenderer;
delete &g_Renderer;

should not generate memory leaks.

In D4682#199088, @Stan wrote:

Doesn't that crash if you actually try to use it? IIRC you need to new it somewhere and &~dtor it elsewhere

I use Sway. The window opens always on the same monitor and not were my mouse is. Minor problem and likely not our responsability.

which version of xcode does jenkins use? string_view neads at least xcode10

explicit std::array in tests

g_Profiler2 is now a macro

remove g_Profiler2 use Singleton

reverse protoVertex
and msvs should build