Honor library ignorefunction() settings when reporting unused functions.

[rgetz]

This change updates CheckUnusedFunctions to skip functions explicitly marked as ignored by the library configuration, both during per-file checks and whole-program analysis. This prevents false positives for known framework hooks, callbacks, or externally-used symbols that should be intentionally excluded from unused-function diagnostics.

I have found this very useful when looking at libraries, and creating a self.cfg file which looks like:

   <function name="iio_context_get_pdata">
     <ignorefunction>true</ignorefunction>
   </function>

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[chrchr-github]

Currently, the attribute is used in CheckClass to allow certain virtual functions to be called in constructors. I doubt that it is much used, but I'm not sure if we should repurpose/dual-purpose this feature. ignorefunction is also quite unspecific, and there is no documentation.
A test should be added.

[rgetz]

Thanks for pointing this out.

You’re right that ignorefunction is not documented in the reference PDF. However, it is a first-class element in the cfg schema itself, cppcheck/cfg/cppcheck-cfg.rng defines

<element name="ignorefunction">
    <ref name="DATA-BOOL"/>
</element>

which implies this attribute is intended to be a general library configuration signal rather than something check-specific or accidental. (happy to be corrected in my understanding if I'm wrong).

Given that:

• There is no documented restriction that ignorefunction is exclusive to CheckClass
• The name itself is generic (unlike e.g. constructor-specific attributes)
• The schema allows it for any function symbol

…using it to suppress unused-function diagnostics seems consistent with the schema-level intent: “this function should be ignored by certain analyses.” Today, unused-function reporting ignores that signal, which leads to false positives even when the user has explicitly annotated the function in the library config...

That said, I agree on two points:

• Documentation gap – the reference PDF should either document ignorefunction and describe its semantics, or explicitly scope it if it is meant to be limited.
• Tests – I will add a regression test covering unused-function suppression via ignorefunction (both per-file and whole-program modes). (based on feedback, if this is the right direction or not).

If folks prefer to keep ignorefunction narrowly scoped to class analysis, I’m open to introducing a more specific cfg attribute for unused-function suppression instead. But given the existing schema and naming, honoring it in CheckUnusedFunctions feels like the least surprising behavior for users today...

Just let me know.

[chrchr-github]

@danmar Do you know anything about the ignorefunction attribute?

[danmar]

I don't remember specifics about ignorefunction. But I would like that the configuration is more specific. adding a checker id is acceptable i.e. <ignorefunction check="unusedFunction">dostuff</ignorefunction>. or do you have some alternative idea?

I don't know if that is OK for you but can't some "maybe_unused" attribute be used on functions?
https://en.cppreference.com/w/cpp/language/attributes/maybe_unused.html
Does Cppcheck handle such attribute?

[rgetz]

yeah, on further reflection, I don't want the function to be ignored (all tests should run), I just didn't want it to show up in the unusedFunction list - and at the time, this was the easiest method. I agree - the semantics of ignore is not natural for that, so this patch shouldn't be merged...

I think what would be more natural is something that maybe reuses <suppress >bar </suppress > but for functions (not types):

<?xml version="1.0"?>
<def>
  <function name="do_something">
    <arg nr="1">
      <minsize type="strlen" arg="2"/>
    </arg>
    <arg nr="2"/>
    <suppress >unusedFunction </suppress >
    <suppress id="unusedFunction"/>
  </function >
</def>

where the suppress is similar to the inline comment cppcheck-suppress-[begin|end] (it's just done in the cfg file, not the source).

Let me know the preferred direction, and I will see what I can do...

[danmar]

I think such <suppress> element sounds good. It's more clear to me what that does. I don't have a strong opinion if the id is written in a attribute or as the text but well it will be a bit more explicit as a attribute with a name.

[rgetz]

Ok - thanks - will close this and replace with suggestions...