astutils.cpp: added more bailouts in isMutableExpression()

[firewave]

No description provided.

[firewave]

@pfultz2 Please have a look. I obviously have no idea what I am doing.

The printing of the tokens we pass to isAliasOf() in isExpressionChangedAt() shows something like this:

{ 0 877 0
{ 0 877 1
std 0 877 0
std 0 877 1
:: 882 877 0
:: 882 877 1
strncmp 0 877 0
strncmp 0 877 1

The namespace prefix does not seem like any expression that might have any impact.

[pfultz2]

You shouldn't bail on ::, this will miss variable changes. Instead you should check astOperand2()(or perhaps both) when its a :::

if (Token::Match(tok, "::"))
return isMutableExpression(tok->astOperand2());

[firewave]

Given the example above in case of std :: strncmp that will pass any of that to this function and my understanding would be that it makes no sense for std :: in that case. Same for a type like std :: string which I also see.

The case you refer to would be a scoped variable. I wonder if all parts would also be passed as an expression that should be changed.

I have the feeling that namespaces are not handled properly here. I guess it requires some tests to be written.

[firewave]

You shouldn't bail on ::, this will miss variable changes.

It is called on all tokens (at least in terms of find aliases) so I think it would only make sense for the actual variable one to proceed on but not the scoping stuff or type. And neither for any function calls.

[pfultz2]

Please have a look. I obviously have no idea what I am doing.

I dont really have a context of what you are trying to do here, but it is definitely problematic.

[firewave]

I dont really have a context of what you are trying to do here, but it is definitely problematic.

I want to reduce the amount of calls to isAliasOf() since most of the time during analysis is spent there.

[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

[firewave]

Closing in favor of #8145. This was mostly about starting a discussion.