[zscript] [clang] Build failures

[kevans91]

Hi,

Hitting the following build failure on Clang:

---

[ 95%] Building CXX object src/CMakeFiles/zdoom.dir/scripting/vm/vmbuilder.cpp.o
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp14: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp14: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;
^~~~~
/usr/local/jenkins/workspace/ZDoom/src/scripting/codegeneration/codegen.cpp10: error: cannot initialize return object of type 'FxExpression *' with an rvalue of type 'bool'
return false;

Seems to have started here: https://github.com/rheit/zdoom/commit/a ... ba0aae86a7

[Graf Zahl]

Urgh. This is one of the more annoying traits of Visual Studio that it silently converts any zero constant to a null pointer...

[kevans91]

Eww, that's particularly ill-behaved. =( This is [fixed], by the way. =)

[Graf Zahl]

Actually, it's not. Assigning 0 to a pointer is perfectly legal in C++, NULL is nothing different. Actually it's Clang which is, strictly spoken, wrong here, it's understandable why but it's still adding semantics to something that cannot be made part of the standard.

[kevans91]

Actually, it's not. Assigning 0 to a pointer is perfectly legal in C++, NULL is nothing different. Actually it's Clang which is, strictly spoken, wrong here, it's understandable why but it's still adding semantics to something that cannot be made part of the standard.

Yeah, I suppose it just feels wrong to make the logical leap from false -> 0 -> pointer value. On the other hand, I also feel dirty about NULL = (void*)0, so fair enough. =)

This re-surfaced again, by the way, in https://github.com/rheit/zdoom/commit/3 ... 0bca7368e4

---

[ 60%] Building CXX object src/CMakeFiles/zdoom.dir/dobjgc.cpp.o
/usr/local/jenkins/workspace/ZDoom/src/d_dehacked.cpp48: error: invalid operands to binary expression ('int' and 'nullptr_t')
for(unsigned j = 0; AmmoPerAttacks[j].func != NULL; j++)
~~~~~~~~~~~~~~~~~~~~~~ ^ ~~~~
1 error generated.
[ 61%] Building CXX object src/CMakeFiles/zdoom.dir/dobjtype.cpp.o