GCC 14.2(.1) compilation issues/errors (appstub.mod, io.mod ...) because of now required casts

[GWRon]

So I tried to compile on ArchLinux running a GCC 14.2.
(required an install of webkit2gtk to get MaxIDE running first).

Compiling a simple sample for now:

SuperStrict
Framework Brl.StandardIO

print "Hello World"

first error popping up is about some implicit declaration (because it does not find the function)

/home/ronny/Downloads/BlitzMax/mod/brl.mod/appstub.mod/appstub.linux.c: In Funktion »bmx_process_vm_readv«:
/home/ronny/Downloads/BlitzMax/mod/brl.mod/appstub.mod/appstub.linux.c:38:25: Fehler: Implizite Deklaration der Funktion »process_vm_readv«; meinten Sie »bmx_process_vm_readv«? [-Wimplicit-function-declaration]
   38 |         size_t result = process_vm_readv(pid, &local, 1, &remote, 1, 0);
      |                         ^~~~~~~~~~~~~~~~
      |                         bmx_process_vm_readv

I simply prepended #define _GNU_SOURCE to "brl.mod/appstub.mod/appstub.linux.c" to fix that.

next error is something which thrown a lot (I for now stopped trying to "fix" things as I am not sure if it is even correct what I do - or if it just removes the warning but breaks the code :D)

Sample error:

/home/ronny/Downloads/BlitzMax/mod/brl.mod/appstub.mod/debugger.stdio.glue.c: In Funktion »bmx_debugger_DebugDeclNext«:
/home/ronny/Downloads/BlitzMax/mod/brl.mod/appstub.mod/debugger.stdio.glue.c:33:31: Fehler: Rückgabe von »char *« aus einer Funktion mit inkompatiblem Rückgabetyp »struct BBDebugDecl *« [-Wincompatible-pointer-types]
   33 |         return ((char *)decl) + sizeof(struct BBDebugDecl);
      |                ~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~

Which I fixes with:

struct BBDebugDecl * bmx_debugger_DebugDeclNext( struct BBDebugDecl * decl ) {
    //OLD: return ((char *)decl) + sizeof(struct BBDebugDecl);
    return (struct BBDebugDecl *)(((char *)decl) + sizeof(struct BBDebugDecl));
}

As said similar needs to be done in other files (eg blitz_app.c):

	//OLD: bbGCStackTop=ebp+28;
	bbGCStackTop = (void **)(ebp + 28);

It also affects stuff like physfs (io.mod) - but I think there the error can be fixed differently - there the first element of a char array pointer was tried to be assigned to a char pointer. Without the ampersand C should (according to the internet :D) already return the first element (so a char pointer here):

int bmx_PHYSFS_setRoot(BBString * archive, BBString * subdir) {
	char abuf[1024];
	size_t len = 1024;
	bbStringToUTF8StringBuffer(archive, abuf, &len);

	char sbuf[1024];
	size_t slen = 1024;

	char * sd = 0;
	if (subdir != &bbEmptyString) {
		bbStringToUTF8StringBuffer(subdir, sbuf, &slen);
		//OLD: sd = &sbuf;
		//not needed to &... because name of arrays already point to first element
		sd = sbuf;
	}

	return PHYSFS_setRoot(abuf, sbuf);
}

Is it worth to try "fixing it on my own" or is it not correct what I am doing here?
Maybe just adding the compiler flag to ignore these things should do (plus the #define-line ...)

[GWRon]

adding "-Wno-incompatible-pointer-types" removed a lot of the clutter and made it compile".
For now I simply added it into make.bmk (after the disable warnings line):

	# disable warnings ?
	if %CC_WARNINGS% == "" then
		opts = opts .. " -w"
	end
	#only usable for C/ObjC but not C++.
	if ext == "c" then 
	  opts = opts .. " -Wno-incompatible-pointer-types -Wno-implicit-function-declaration"
	end

As it is only available after GCC 5 I am not sure if we need to take care of it with some GCC-Version check ...
Also it cannot be used for C++ - compilations (spits out some warnings there). I also disabled the implicit function declaration warning which skips requiring the change (gnu source...)

[GWRon]

The bbThreadRegister-call n "sdl.mod/SDL/src/thread/SDL_thread.c SDL_Runthread" also does some conversion it does not like.

	/* register with BlitzMax */
	BBThread * bbThread = bbThreadRegister(SDL_ThreadID());

I supressed it with -Wno-int-conversion (in the same line)
albeit I am feeling it should be handled there "directly" (as it is a modification of some 3rd party file anyway).

[GWRon]

the gnu_source-part is now added via 0ed5db0

debugger-issue resolved and
thread-casting resolved by adding a typedef/custom "type" in 120dc1f and bmx-ng/sdl.mod@347575c

[GWRon]

To "revive" this issue:

if we simply did ./bmk makemods -ra lot of more issues will pop up ... warnings first of course.

A lot of them can be fixed more easily ... others are just "tedious":

    Field drawColor:SDLColor = New SDLColor(255, 255, 255, 255)
    Field clsColor:SDLColor = New SDLColor(0, 0, 0, 255)

-> there it blames that 255 and 0 are integers, a simple solution would be to :Byte all these values (but then it looks no longer "clean" ... if people are required to do similar "base modules" like max2d... they will surely find BlitzMax less "clean" (from the looks).

So solution here would be to use overloads (int,int,int,int).

Similar happens to some GL-Calls (in mojo.mod and sdl.gl-stuff) ... where we use False and True. While there is GL_TRUE and GL_FALSE ... but they are mistakingly defined as "int" (indirectly, they are simply: const GL_FALSE=0) ... original definition is unsigned char (so byte).

This aside you then land in errors like this:

 15%] Compiling:archive_read_support_filter_grzip.c
/home/ronny/Downloads/BlitzMax/mod/archive.mod/core.mod/glue.c: In Funktion »bmx_libarchive_seek_cb«:
/home/ronny/Downloads/BlitzMax/mod/archive.mod/core.mod/glue.c:46:82: Fehler: Übergabe des Arguments 4 von »archive_core_TArchiveCallbackData__seek« von inkompatiblem Zeigertyp [-Wincompatible-pointer-types]
   46 |         archive_core_TArchiveCallbackData__seek((BBObject*)data, offset, whence, &ret);
      |                                                                                  ^~~~
      |                                                                                  |
      |                                                                                  la_int64_t * {alias long int *}
/home/ronny/Downloads/BlitzMax/mod/archive.mod/core.mod/glue.c:31:101: Anmerkung: »BBInt64 *« {alias »long long int *«} erwartet, aber Argument hat Typ »la_int64_t *« {alias »long int *«}
   31 | void archive_core_TArchiveCallbackData__seek(BBObject * data, BBInt64 offset, int whence, BBInt64 * count);
      |                                                                                           ~~~~~~~~~~^~~~~

which is just telling us: _la_int64_t is just a long, wile it wants a long long but libarchive only uses long long (int) in specific situations:

/* Get a suitable 64-bit integer type. */
#if !defined(__LA_INT64_T_DEFINED)
# if ARCHIVE_VERSION_NUMBER < 4000000
#define __LA_INT64_T la_int64_t
# endif
#define __LA_INT64_T_DEFINED
# if defined(_WIN32) && !defined(__CYGWIN__) && !defined(__WATCOMC__)
typedef __int64 la_int64_t;
# else
#include <unistd.h>
#  if defined(_SCO_DS) || defined(__osf__)
typedef long long la_int64_t;
#  else
typedef int64_t la_int64_t;
#  endif
# endif
#endif

Further fixes regarding implicit function definitions etc are already tackled and discussed in the discord-servers issue-channel. Here and there it misses some #include (and also there I am not sure if my fixes are done in a way Brucey would like to see it in the sources).

I am open to assist removing warnings (and compiler issues) but need feedback from @woollybah first ... am not keen to do things and then the PR gets not merged (mostly because I did it ... "incorrect" ... but also because of lacking sparetime of Brucey regarding checking PRs / merging)

We could even collaborate by creating a branch (fix_gcc_issues) in the modules, commit to there (or do Pull Requests to there) and once all is done it is merged in a single squash commit.