AROS build fixes and code cleanup.

[Kalamatee]

Apply updates to the AROS build so that it works now that dhewm3 uses SDL2, and newer OpenAL requirements.
Use consistent formatting in the code, so it matches existing code.
Adjust startup code to allocate 1MB stack space.
Check in the directory the game has been launched from, for mod components before the game directory so it doesn't use dooms config instead of the mods.
Don't use auto-opened bsdsocket.library, but open it ourselves (and close on exit).
Remove old dll code - use the dynmodule library provided by AROS.

[Kalamatee]

What's needed to get these fixes merged @DanielGibson?

[DanielGibson]

time and motivation (for review) on my side

[DanielGibson]

patching Dear ImGui is problematic, because I usually just copy over the source files of new upstream versions..
Could you try to get this upstream so next time I update ImGui it doesn't break AROS?

[Kalamatee]

I'll look into that, thanks.

…

On Wed, 22 Oct 2025, 02:51 Daniel Gibson, ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In neo/libs/imgui/imgui.cpp <#690 (comment)>: > @@ -15037,6 +15037,12 @@ static bool Platform_OpenInShellFn_DefaultImpl(ImGuiContext*, const char* path) { return (INT_PTR)::ShellExecuteA(NULL, "open", path, NULL, NULL, SW_SHOWDEFAULT) > 32; } +#elif defined(__AROS__) patching Dear ImGui is problematic, because I usually just copy over the source files of new upstream versions.. Could you try to get this upstream so next time I update ImGui it doesn't break AROS? — Reply to this email directly, view it on GitHub <#690 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACWCNPFPLMF7T5JACNYMEQT3Y3PLDAVCNFSM6AAAAACDQC7VSWVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTGNRTGM3TKOBSGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[DanielGibson]

Any update on ImGui?

Also, I just merged something that creates a merge conflict (I added functions for getting the the time and sleeping at higher resolutions), could you rebase to current master?

Thanks :)

[Kalamatee]

Hi Daniel, I have rebased and reverted the imgui changes/adapted the build, so they are not necessary. I can confirm it successfully builds.

[DanielGibson]

Awesome, thank you!

[DanielGibson]

BTW, is it correct that AROS only has millisecond-precision in clock_gettime()?
If it provided something with higher resolution the game could run smoother

[Kalamatee]

I believe so - but would need to look into it. We did have initial support for HPET, but I'm not sure how complete it is, or if it exposes higher precision.

[DanielGibson]

Note that dhewm3 doesn't necessarily need a high precision/high frequency scheduler, just a precise time source (where querying it has relatively low overhead).

In the Windows implementation it does the last millisecond or so of waiting with a busy loop only slowed down by CPU PAUSE instructions:

dhewm3/neo/sys/win32/win_main.cpp

Lines 1121 to 1145 in 47ae272

	void Sys_SleepUntilPrecise( double targetTimeMS ) {
	double msec = targetTimeMS - Sys_MillisecondsPrecise();
	if ( msec < 0.01 ) // don't bother for less than 10usec
	return;
	if ( msec > 2.0 ) {
	// Note: Theoretically one could use SetWaitableTimer() and WaitForSingleObject()
	// for higher precision, but last time I tested (on Win10),
	// in practice that also only had millisecond-precision
	dword sleepMS = msec - 1.0; // wait for last MS or so in busy(-ish) loop below
	Sleep( sleepMS );
	}
	// wait for the remaining time with a busy loop, as that has higher precision
	do {
	#ifdef D3_CpuPause
	for ( size_t i=0; i < pauseLoopsPer5usec; ++i ) {
	// call it 4 times per loop, so the ratio between pause and loop-instructions is better
	D3_CpuPause(); D3_CpuPause(); D3_CpuPause(); D3_CpuPause();
	}
	#endif
	msec = targetTimeMS - Sys_MillisecondsPrecise();
	} while ( msec >= 0.01 );
	}

On Linux it does the same for the last 100 microseconds