[HEAD] (ARM) GCC 4.6.2: invalid casts in v_text.cpp

[Csonicgo]

https://github.com/rheit/zdoom/blob/6d4 ... t.cpp#L159
https://github.com/rheit/zdoom/blob/6d4 ... t.cpp#L153

These two lines in this block of code:

Code: Select all

		// We don't handle these. :(
		case DTA_DestWidth:
		case DTA_DestHeight:
			*(DWORD *)tags = TAG_IGNORE;   <--
			data = va_arg (tags, DWORD);
			break;

		// Translation is specified explicitly by the text.
		case DTA_Translation:
			*(DWORD *)tags = TAG_IGNORE;  <--
			ptrval = va_arg (tags, void*);
			break;

may cause an error on compilation on non x86-64 platforms, because type va_list cannot be cast to DWORD* (aka an unsigned int*) .

Solution: Change these offending lines to:

Code: Select all

assert(false && "unsupported");

[Graf Zahl]

Solution: Change these offending lines to:

Code: Select all

assert(false && "unsupported");

What does that solve? It only breaks the code.

[Blzut3]

The offending lines are disabling invalid input before passing the tags off to character drawing. Unless I'm missing something (my test case for the change suggested was ECWolf's Android port, I imagine Beloko did the same for GZDoom's Android port), using asserts here is harmless since you should never input those values when drawing text.

[Csonicgo]

Solution: Change these offending lines to:

Code: Select all

assert(false && "unsupported");

What does that solve? It only breaks the code.

Is there a better solution to this? This is the only showstopper in the code that prevents ARM compilation as far as I can tell.

[Graf Zahl]

That code is still needed so just silencing the error and ignoring the problem is not going to help matters.

[Blzut3]

That code is still needed so just silencing the error and ignoring the problem is not going to help matters.

Mind to give me a line #? I don't see any calls to DrawText that use these tags, which makes sense since they appear to be no-ops.

[Graf Zahl]

You are correct. These are never used by any DrawText call. I just replaced the offending code with an 'assert(false)'. Using these should be forbidden as they don't work anyway.

[Edward-san]

but in a release build, the 'assert' is simplified ... what about a 'I_(Fatal?)Error' call?

[Graf Zahl]

This should be caught during development - and any new addition here will inevitably trigger the assert - and immediately be removed.