build wwlib on linux by rm5248 · Pull Request #145 · w3dhub/OpenW3D

rm5248 · 2026-06-19T22:40:51Z

Build wwlib on Linux.

Mostly Claude generated, but I did clean it up slightly to better align with OpenW3D.

madebr · 2026-06-20T06:23:35Z

+
+/*
+** Linux implementation of RegistryClass backed by an INI file.
+** Persists settings to $HOME/.config/renegade/registry.ini.
+*/
+
+#if !defined(_WIN32)
+


wwlib/openw3d.h writes its settings to openw3d.ini.
Perhaps the OpenW3D:: functions should be moved into RegistryClass and RegistryClass should write its settings to openw3d.ini in an appropriate location instead?

Also, I think we want to remove the dependency on the registry on Windows as well.

OmniBlade

Few things I'd personally prefer to be done in alternate ways. Not sure if others agree.

OmniBlade · 2026-06-20T10:07:32Z

 #endif
+
+#ifndef OutputDebugStringA
+#define OutputDebugStringA(s) ((void)(s))


I'd be more inclined to remove these and replace with either WWDEBUG_SAY or pure printf. Generally I want to remove always.h and refactor into separate headers for the things its needed for.

reasonable to ask, but this is currently used many places in the code. I think that would make sense as a different PR?

OmniBlade · 2026-06-20T10:08:00Z

+
+// Windows BOOL type
+#ifndef BOOL
+typedef int BOOL;


Just replace use of BOOL with int and be done with it, as winapi is removed this shouldn't be needed.

similar to the OutputDebugString, a cursory glance indicates that there are 400+ files with a BOOL declared in them at some point, so removing it at the moment doesn't seem feasible.

OmniBlade · 2026-06-20T10:09:26Z

+
+// Windows file operation equivalents
+#ifndef DeleteFileA
+#define DeleteFileA(f) (remove(f) == 0)


Just replace calls to DeleteFile and MoveFile with calls to remove and rename which are supported by posix and msvc.

OmniBlade · 2026-06-20T10:10:10Z

+#endif
+
+// _splitpath: split a path into drive/dir/fname/ext components
+static inline void _splitpath(const char *path, char *drive, char *dir, char *fname, char *ext)


I'd rather this function be moved to its own header along with the constants it uses and only included where needed.

This can probably be replaced with std::filesystem functions. It's currently used in a lot of places, so standardizing it does make sense(but probably as a separate PR)

OmniBlade · 2026-06-20T10:14:54Z

+#if defined(_WIN32)
 	CRITICAL_SECTION Bar;
+#else
+	std::mutex Bar;


std::mutex is supported everywhere as our base C++ version is 20, maybe just convert to use that exclusively? Also, where is the corresponding cpp implementation?

We can probably remove this class, but this at least gets it into a state where it can compile. The cpp implementation was not changed, but I will go update that.

actually upon further investigation the h/cpp files are not used in the build at all.

Oh yeah, the same class with a slightly different implementation is implemented in mutex.h/cpp, can probably just remove this entirely then?

OmniBlade · 2026-06-20T10:15:31Z

 	//
-	::DeleteFileA (MixFilename);
-	::MoveFileA (full_path, MixFilename);
+	DeleteFileA (MixFilename);


See previous comment on dealing with these.

OmniBlade · 2026-06-20T10:16:17Z

 /*
 **
 */
+#if defined(_WIN32)


Probably needs a std::filesystem, SDL3 or dirent alternative implementation here though that can be TODO.

OmniBlade · 2026-06-20T10:17:06Z



 #include "refcount.h"
+#if defined(_WIN32)


Is the windows header needed at all if its not needed on posix?

OmniBlade · 2026-06-20T10:20:32Z

+
+#if !defined(_WIN32)
+
+#include "registry.h"


Ideally we want to replace the use of this class entirely throughout the codebase, so making a temp implementation seems counter productive. Maybe just stub it out instead if the intention is to just get wwlib compiling on linux to make further development easier.

OmniBlade · 2026-06-20T10:23:07Z

    add_compile_definitions(strnicmp=strncasecmp)
    add_compile_definitions(wcsnicmp=wcsncasecmp)
    add_compile_definitions(wcsicmp=wcscasecmp)
+    add_compile_definitions(DebugBreak=__builtin_trap)


Not a fan of these defines being added like this, would be better to add a few headers to wrap this kind of thing instead. Like how debug break is handled here: https://github.com/TheAssemblyArmada/Thyme/blob/ccef1e11c1355c6db577a057c06e7d790f1a0333/deps/baseconfig/src/intrinsics.h#L141

Long-term std::breakpoint would make more sense, but it's only used in 2 places at the moment so I'll simply remove it instead.

caseychaos1212 · 2026-06-23T17:49:38Z

CODEX REVIEW

Findings

P1: Windows webbrowser builds will lose RegisterCOMServer
[Code/wwlib/CMakeLists.txt#L53](

OpenW3D/Code/wwlib/CMakeLists.txt

Line 53 in 4ada1c6

win.cpp

) removes WWCOMUtil.cpp from wwlib, but [WebBrowser.cpp#L100](

OpenW3D/Code/Commando/WebBrowser.cpp

Line 100 in 4ada1c6

bool success = RegisterCOMServer(dllPath);

) still calls RegisterCOMServer, whose only definition is in [WWCOMUtil.cpp#L179](

OpenW3D/Code/wwlib/WWCOMUtil.cpp

Line 179 in 4ada1c6

bool RegisterCOMServer(const char* dllName)

). Since WebBrowser.cpp is still in COMMANDO_SRC, Windows builds with webbrowser enabled should hit an unresolved external. Keep WWCOMUtil.cpp in the Windows source list or move it to the consumer target.
P1: printf(buffer) introduces a format-string bug
[Code/wwlib/ini.cpp#L2363](

OpenW3D/Code/wwlib/ini.cpp

Line 2363 in 4ada1c6

printf(buffer);

) passes INI-derived text as the format string. buffer includes entry, section, and Filename, so % sequences in those values can crash or corrupt memory. Use printf("%s", buffer) or fputs(buffer, stdout).

P2: Linux registry export/import does not round-trip
[registry_linux.cpp#L223-L232](

OpenW3D/Code/wwlib/registry_linux.cpp

Lines 223 to 232 in 4ada1c6

    
           void RegistryClass::Save_Registry_Tree(char *path, INIClass *ini) 
        
           { 
        
           	const char *section = path; 
        
           	int index = 0; 
        
           	const char *entry = NULL; 
        
           	char buf[1024]; 
        
           	while ((entry = Registry_INI().Get_Entry(section, index++)) != NULL) { 
        
           		Registry_INI().Get_String(section, entry, "", buf, sizeof(buf)); 
        
           		ini->Put_String(section, entry, buf); 
        
           	}

) saves raw entry names, but Load_Registry only imports BIN_, DWORD_, and STRING_ entries. It also only saves/deletes the exact section, unlike the Windows implementation which recurses subkeys and prefixes typed values. Any Save then Load flow, such as slave registry cloning, will silently drop settings on Linux.

Validation

I fetched PR #145 locally and reviewed the full diff. git diff --check is clean. I did not run a Linux build from this Windows workspace.

madebr · 2026-07-13T22:43:44Z

+static void Ensure_Registry_Dir()
+{
+	const char *home = getenv("HOME");
+	if (!home) return;
+	char dir[512];
+	snprintf(dir, sizeof(dir), "%s/.config", home);
+	mkdir(dir, 0755);
+	snprintf(dir, sizeof(dir), "%s/.config/renegade", home);
+	mkdir(dir, 0755);
+}
+


This should use the openw3d.ini path as established by #48 and #104.

madebr · 2026-07-13T22:47:48Z

+	static INIClass ini;
+	static bool loaded = false;
+	if (!loaded) {
+		loaded = true;
+        RawFileClass file(OpenW3D::Get_Config_File_Path());
+		if (file.Is_Available()) {
+			ini.Load(file);
+		}
+	}
+	return ini;


This INIClass object will clash with the INIClass object created by the OpenW3D:: functions.

This sequence of actions will cause data loss:

global INIClass object is initialized (through RegistryClass), loads data from disk and stays alive

OpenW3D:: function reads ini

OpenW3D:: function writes ini

RegistryClass method is called: the ini is is not read again

I think #48 and #104 should be partially reverted such that:

RegistryClass is only served by the ini: the win32 code can be removed. We don't want to access the registry

the openw3d.ini path is calculated as master currently does

rm5248 added 2 commits June 18, 2026 21:32

build wwlib on linux

83fba9a

clean up claude work

14a250b

rm5248 requested review from OmniBlade and caseychaos1212 June 19, 2026 22:40

remove files that have whitespace-only changes

11319a5

madebr reviewed Jun 20, 2026

View reviewed changes

OmniBlade requested changes Jun 20, 2026

View reviewed changes

rm5248 added 2 commits June 20, 2026 20:18

update with PR comments

36cfa24

add cstdio to includes

4ada1c6

fix warning

0f8cf5c

This was referenced Jul 12, 2026

Remove OutputDebugStringA functions #149

Closed

Remove Windows registry #150

Open

madebr reviewed Jul 13, 2026

View reviewed changes

OmniBlade mentioned this pull request Aug 5, 2026

feat: Wrapper for __debugbreak equivalents on other platforms. #152

Merged

Uh oh!

Conversation

rm5248 commented Jun 19, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

OmniBlade left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

caseychaos1212 commented Jun 23, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants