Re: Your aliasing patch

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Richard" == Richard Kenner <kenner@vlsi1.ultra.nyu.edu> writes:

    Richard> I did, but I get so many failures due to some
    Richard> misconfiguration that I can't always tell which are new.
    Richard> I should try to figure out why I get the errors, but I
    Richard> suspect it's some library that I don't understand since I
    Richard> don't know C++.

I don't think that's a very good excuse.

If nothing else, you could always run the tests with and without your
change.  That would at least show you what had changes.  For
compile-only tests, like the ones that failed with your change, this
would work.  Note that I'm not advocating this as suggested practice;
I'm simply pointing out that there's an easy procedure that would have
got the problem with your patch.

Most of us are not experts on all of the languages that GCC handles,
let alone all of the targets it supports.  Yet, we have a mutual
obligation to figure out enough to ensure that the work we do doesn't
adversely affect other folks -- even in areas in which we are not
experts.  Running the tests is the *minimum* we should do to check our
changes.  We should really try to understand all the affects the
changes might have, and make sure that they're OK first.  Once we
think they are, then we should run the tests to confirm our analysis.

I'm sure that you can figure out what's going on.  Or, that someone
will help you if you ask.  I know there are other GCC developers that
use Alpha platforms for their development efforts.

    Richard>     In addition, I'm not sure what you're using this bit
    Richard> for on a FIELD_DECL.

    Richard> It means that it is permissible to take the address of
    Richard> that field individually.

Fair enough.  Please add documentation for this.  Thanks!

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com