ValueFlow: don't reset the struct size if there is a unknown member + empty structs have a size of 1

[autoantwort]

Fixes https://trac.cppcheck.net/ticket/13696 + #6257 (comment)
One could argue that the size of a member of an unknown type should be 1 instead of 0 because the member will be at least 1 byte large (if not empty and annotated with [[no_unique_address]])

[chrchr-github]

Thanks for your contribution. There are situations where the exact size of a type is required, and others (e.g. the passedByValue check) where a lower bound is sufficient. We should probably differentiate this by a flag passed to getSizeOf().

[firewave]

We also need to add a test for the referenced ticket.

[autoantwort]

We also need to add a test for the referenced ticket.

Can you point me to the place where I should insert it (in which of the test cases)?

[chrchr-github]

We also need to add a test for the referenced ticket.

Can you point me to the place where I should insert it (in which of the test cases)?

There is a iterateByValue() test in testother.cpp.

[autoantwort]

We should probably differentiate this by a flag passed to getSizeOf().

Done now. I was not sure what the right flag should be in every context.

We also need to add a test for the referenced ticket.

Done :)

[autoantwort]

There are situations where the exact size of a type is required, and others (e.g. the passedByValue check) where a lower bound is sufficient.

Btw the current implementation already returned the lower bound if inheritance is used.

[autoantwort]

@firewave Can you approve the workflow runs?