SvFromTclObj(): unreachable code uses invalid `internalRep`

[chrstphrchvz]

In this code which appeared in 323f8e1:

tcl.pm/Tcl.xs

Lines 566 to 579 in 4067e75

	else if (objPtr->typePtr == tclBooleanTypePtr) {
	/*
	* Booleans can originate as words (yes/true/...), so if there is a
	* string rep, use it instead. We could check if the first byte
	* isdigit(). No need to check utf-8 as the all valid boolean words
	* are ascii-7.
	*/
	if (objPtr->typePtr == NULL) {
	sv = newSVsv(boolSV(objPtr->internalRep.longValue != 0));
	} else {
	str = Tcl_GetStringFromObj(objPtr, &len);
	sv = newSVpvn(str, len);
	}
	}

The statement

sv = newSVsv(boolSV(objPtr->internalRep.longValue != 0));

is unreachable (assuming tclBooleanTypePtr != NULL). It also does not make sense for the statement to be using objPtr->internalRep because it is invalid when objPtr->typePtr == NULL.

I am still trying to understand how this code is intended to work. I may propose changes to it anyway in #42 due to various other issues in SvFromTclObj() which I encounter in Tcl 9.0.

[chrstphrchvz]

I am not sure I will ever figure out what the original intention was…

At this point, I would be inclined to only ever return 0 or 1 even for “string” booleans. I don’t see how returning them as strings in Perl is useful, as I imagine doing so would only needlessly complicate things for users; why make the user check for "true"/"false"/"yes"/"no"/"on"/"off" when Tcl_GetBooleanFromObj() can do so for them?

Compare to how Tkinter converts any value Tcl says is already boolean (including string booleans) to a Python boolean—False or True.