This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch for stricter implicit conversions between vectors
On 11/14/06, Mark Mitchell <mark@codesourcery.com> wrote:
Ian Lance Taylor wrote:
> Andrew Pinski <pinskia@gmail.com> writes:
>
>>> We do have such flags. For example, -fpermissive and
>>> -ffriend-injection.
>> Except those are only for C++ and this case is a bit special.
>
> I disagree. I appreciate your message, but it doesn't add any new
> facts, and doesn't change my feelings on this matter. I won't bother
> to reply in detail, since it would add nothing to the conversation.
I agree with you that the flag is consistent with our precedent, and
that we should keep it.
Now Sony/IBM, and others have compilers that reject this code even
with a 4.0.2 based compiler.
So again I don't think an option is useful for this case, it just
allows for sloppy code.
Which just happened to newlib beause this patch has not been applied yet.
See http://sourceware.org/ml/newlib/2006/msg00929.html .
I know for a fact I am going to remove this option from the PS3
toolchain once we go to 4.3 because it just lets people write sloppy
code.
We have a regression from 3.4.x about accepting invalid code. Can any
time you have some C++ code that is recent regression, can you add a
flag, if we do that, then we end up with 4 flags currently. Do we
really want to support all those 4 flags for at least 3 releases, I
don't think so. At least one of them looks like it should have a flag
if we go by the above cretia, PR 20721 and another for C, PR 21419.
If this was not a regression, I would not be questioning a need for a
flag but this is a regression from a recent release that people
actually still use.
Thanks,
Andrew Pinski