Lua throw: set panic function as main priority

[israpps]

Makes panic function trigger first (if programmer provided one) instead of testing first on longjmp() Wich has issues on PS2SDK lately.

Impact on behaviour of already existing apps: Only if they already use a panic func.

Example of panic function (adapted to enceladus and tweaked to avoid the longjmp crash)

#define TPRINTF(arg, x...) \
    printf(arg, ##x); \
    scr_printf(arg, ##x)


int test_error(lua_State * L) {
    scr_clear();
    init_scr();
    scr_clear();
    scr_clear();
    scr_setCursor(0);
    int n = lua_gettop(L);
    int i;
    scr_printf("\t");
    TPRINTF("LUA ERR.\n");

    if (n == 0) {
        scr_printf("\t");
        TPRINTF("Stack is empty!!!!\n");
    }

    for (i = 1; i <= n; i++) {
        scr_printf("\t");
        TPRINTF("%i: ", i);
        switch(lua_type(L, i)) {
        case LUA_TNONE:
            TPRINTF("Invalid");
            break;
        case LUA_TNIL:
            TPRINTF("(Nil)");
            break;
        case LUA_TNUMBER:
            TPRINTF("(Number) %f", lua_tonumber(L, i));
            break;
        case LUA_TBOOLEAN:
            TPRINTF("(Bool)   %s", (lua_toboolean(L, i) ? "true" : "false"));
            break;
        case LUA_TSTRING:
            TPRINTF("%s", lua_tostring(L, i));
            break;
        case LUA_TTABLE:
            TPRINTF("(Table)");
            break;
        case LUA_TFUNCTION:
            TPRINTF("(Function)");
            break;
        default:
            TPRINTF("Unknown");
        }
    TPRINTF("\n");
    }
    SleepThread(); //else, crash still ocurs
    return 0;
}

[fjtrujy]

@israpps please take a look to check if this is still failing, if not close the PR

[israpps]

@israpps please take a look to check if this is still failing, if not close the PR

made some tests and it seems to be fixed.

[israpps]

thinking about this a bit, this feature is actually helpful, since the lua stack has more information when the panic function is called compared to the information obtained by checking it after the lua_pcall finished

and since the panic feature has to be explicitly coded by the developer, we dont have to worry about it changing existing projects behavior

[fjtrujy]

thinking about this a bit, this feature is actually helpful, since the lua stack has more information when the panic function is called compared to the information obtained by checking it after the lua_pcall finished

and since the panic feature has to be explicitly coded by the developer, we dont have to worry about it changing existing projects behavior

But this is a “common behavior to all platforms ” if you think it still makes sense my recommendation is to create PR to original LUA repo

[israpps]

thinking about this a bit, this feature is actually helpful, since the lua stack has more information when the panic function is called compared to the information obtained by checking it after the lua_pcall finished

and since the panic feature has to be explicitly coded by the developer, we dont have to worry about it changing existing projects behavior

But this is a “common behavior to all platforms ” if you think it still makes sense my recommendation is to create PR to original LUA repo

The original purpose of the panic function (if I understood correctly) is to provide a "if everything else fails" error handler. Not sure they'll like it.

[fjtrujy]

So then I would be against apply this change, the more “mainstream” we are, the better