Uninitialized Return Value

[mikep]

Shouldn't the following code at least issue a warning message about an uninitialized return value:

Code: Select all

Private Function GetValue(ByVal c1 as Integer) as Integer	
	If c1 <> 0 Then
		GetValue = c1
	End If
End Function

[dkinzer]

Shouldn't the following code at least issue a warning message about an uninitialized return value

Yes, it should and it does here unless the compiler option --warn=no-undefined-var is specified. Check your .pjt file and the IDE user options file to see if that specific warning (or warnings in general) is being suppressed.

Alternately, you can try this code which turns on the undefined variable warning for the code within.

Code: Select all

#pragma warning(push;undefined-var:on)
Private Function GetValue(ByVal c1 as Integer) as Integer
   If c1 <> 0 Then
      GetValue = c1
   End If
End Function
#pragma warning&#40;pop&#41;

The resulting message should be something like this:

Code: Select all

foo.bas:12: Warning(2): "GetValue", not all paths set the return value

Perhaps the undefined return value warning should be distinguished from the more general undefined variable warning.

[mikep]

Shouldn't the following code at least issue a warning message about an uninitialized return value

Yes, it should and it does here unless the compiler option --warn=no-undefined-var is specified.

Or the routine is not referenced by other code. I suppose that this type of warning is issued by the code generation backend. No generation for dead code means no warning.

Problem solved for me but it might be worth investigating if warnings can be issued for this type of problem. A warning list of dead routines might be even better.

[dkinzer]

Or the routine is not referenced by other code.

The warning is currently generated during the code generation phase and it uses information generated during the optimization phase. Since the optimization phase is only performed for procedures that appear in the call graph (i.e. those that are invoked), moving the warning to the optimization phase wouldn't help.

A warning list of dead routines might be even better.

A warning option for unused procedures could be added. It would be best, I believe, if it were not on by default as is the case for the warning for unused parameters.