Re: ABI Issues: C++ PATCH: PR 12053

[Mike Stump <mrs at apple dot com>]

On Thursday, September 4, 2003, at 10:34 AM, Mark Mitchell wrote:
> > At Apple, we are using the abi from 3.3 in our about to be shipped
>
> It seems that sentiment is against including this patch in 3.3.2, so I
> will back it out on the branch.

Thanks.

> That still leaves the question that nobody has answered -- do we want
> this patch on the mainline?

I answered that by not answering it.  If the libstdc++ folks won't lay 
down $500 on a bet that they in fact are binary compatible for heavy 
users of their library (3.3.x to 3.4), then... we can have 3.4 follow 
the abi standard, even if it isn't compatible with 3.3.  For 3.3 to 
3.3.2, I'm hoping that libstdc++ _is_ compatible (as compatible as can 
be), and we are betting on it.

> If this change had occurred in the -fabi-version=0 mode, our C++ ABI
> Testsuite would have caught it, but we do not have tests for the
> -fabi-version=1 mode.  I'm worried that if we promise to support every
> variant of the ABI forevermore, we're going to find that we fail to do
> that, and therefore disappoint people.

Welcome to a consequence of -fabi and trying to have binary 
compatibility!  :-)

Odd, you only test what people don't use, and don't test what people do 
use.  Maybe that is backward.  Imagined compatibility isn't more 
important than real compatibility.