Re: [commit] Do not rely on FIELD_LOC_KIND_BITPOS being zero

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> I do not like this apprach as it misleads the reader that
Jan> FIELD_BITPOS is newly set here.  This is why I did not want to
Jan> implement such change and just fixed the existing cases.

Jan> FIELD_BITPOS is being only added to, which the original code said.
Jan> Now the code is not so clear.

We could add a comment to clarify it.

This technique is already used elsewhere, so the reader presumably will
be familiar with the idiom.

Jan> That it gets caught by review is true but particularly with GDB
Jan> many patches are even never submitted.

Let's pretend that patches that aren't submitted just don't exist.
That's what I do :)

The reason I like the rvalue accessor approach is that it helps prevent
future bugs of this sort.  No approach is perfect, especially because C
has relatively weak access protection mechanisms, but I still see this
as an improvement overall.

I agree that this is yet another thing that C++ does better.
But that is being discussed under a different thread and, until some
decision is reached, I think we should continue to deal with the patches
we see as they are, and not link them to C++.  After all, even if C++ is
approved the migration will not be instantaneous.

Tom