Duplicate defines in library config

[Frank Winklmeier]

Using cppcheck with multiple library configs that define the same macro results in some surprising results, in terms of which one takes precedence.

Consider the two following libraries:

> cat lib1.cfg 
<?xml version="1.0"?>
<def format="2">
  <define name="slots" value=""/>
</def>

> cat lib2.cfg
<?xml version="1.0"?>
<def format="2">
  <define name="slots" value="foo"/>
</def>

and execute:

> cat test.cxx 
a = slots;
> cppcheck -E --library=lib1.cfg --library=lib2.cfg test.cxx
a = ;

No matter the order of --library the result is always the same. That's because all library definitions are stored as <name> <value> string in a std::set. So by alphabetical ordering the empty define in the above example always takes precedence.

From a user-perspective it would make more sense to either:
* throw an error if the same macro is defined with different values (currently that's only the case for identical name/value pairs), or
* define a clear precedence depending on the order of the libraries specified on the command line.

Personally, I would probably favor the second option. That would also allow the user to re-define some of the built-in defines, which is how I came across this.

-Frank

P.S.: If there is a consensus I might be able to provide a PR...

[Oliver Stöneberg]

I thought we bail out on duplicated defines nowadays...

[CHR]

That's only for duplicates within the same .cfg IIRC.

[Frank Winklmeier]

Maybe that was the intention, but it's not what the code does. You only get an error if the name and value of the define are identical.

[André Coquim]

Hi, not about this topic, but somehow related with the problem mentioned here. 5 days ago I opened a similar situation related with the loading of the macros in std.cfg and their colosion, but I had got no answer, could you please provide an answer there?
https://sourceforge.net/p/cppcheck/discussion/general/thread/89612cfdd4/
Thanks