ClangImport: improved location parsing / better handling for "prologue" TypedefDecls

[firewave]

No description provided.

[firewave]

@danmar @chrchr-github
In the future it would be great if could also add the source which generated the AST as a comment. This would make it easier to understand what it represents. Also the AST is constantly evolving so it might be possible we will no longer be able to reproduce the same output with newer versions.

This also makes it harder to convert the tests to a different AST format (i.e. -ast-dump=json / https://trac.cppcheck.net/ticket/13398).

[firewave]

I just wanted to get rid of the setLang() hack so I could hopefully proceed with other related refactorings but as usual there was ripple effect which required more changes (for the better IMO).

[chrchr-github]

What is the status of clangimport anyway? The one time I checked, it didn't seem to work very well.

[firewave]

What is the status of clangimport anyway? The one time I checked, it didn't seem to work very well.

Still experimental and that basically reflects the reality of the implementation. There was talk of dropping it completely at some point.

[danmar]

What is the status of clangimport anyway? The one time I checked, it didn't seem to work very well.

Still experimental and that basically reflects the reality of the implementation. There was talk of dropping it completely at some point.

In my opinion it's still experimental and it should be rewritten to use the json output from clang instead. I have no plans to do it anytime soon though.

I think I might have suggested that this could be dropped but I don't think so anymore.

We have discussed this feature in Cppcheck Solutions company at some point and it's a bit interesting for us. Maybe someday we will decide that it's needed and ensure it's implemented.

[danmar]

Also the AST is constantly evolving so it might be possible we will no longer be able to reproduce the same output with newer versions.

yeah I fear that our clangimport gets more and more broken for every clang version. :-(

[firewave]

yeah I fear that our clangimport gets more and more broken for every clang version. :-(

Not essentially. The test coverage keeps improving slightly and there's the (in the CI still disabled) TEST_CPPCHECK_INJECT_CLANG which will easily shake out the most obvious issues. Something to look into if all known problems have been fixed.