Re: C++ PATCH: Add -fabi-version

[Mark Mitchell <mark at codesourcery dot com>]

> Shouldn't this issue a warning even if abi_version_at_least (2)
> returns fails?  E.g. something like
>
> 		warning ("ABI change: versions 3.2 and earlier of GCC "
> 			 "used a different ABI for `%T'", t);
Keeping track of this for more than two versions of the ABI will drive
you crazy, I think.

On the other hand, if you want -Wabi to do this, or you want -Wabi=n,
go ahead. :-)

> It would be helpful to document what the default is.
OK, I'll add that.

> With your current code, I think this test will fail if compiled with
> -fabi-version=1, since the warning won't be issued.  That might make
No, -fabi-version=1 is the default; my patch did not change the default
ABI.

> Furthermore, shouldn't the 3.2 ABI be 102, not 1, to match
> __GXX_ABI_VERSION?
Gentlemen, you can make this as complicated as you want.  If you want
to key it off dates, or __GXX_ABI_VERSION, that's fine.  In my judgement,
you're more likely to make it harder to get the actual code right, but
that's up to you.

> Also, wouldn't it be better to change that 0 to INT_MAX when
> initializing it, so the test could be simplified?
I don't know; maybe.

But surely you don't want me to write:

 -fabi-version=2147483647

to turn on the closest approximating ABI.

I'll be out of town much of this week, and I don't have strong feelings
about this stuff.  I think it's the stuff of which flamewars are made.

I'll happily go along with whatever people want; make changes as you
see fit.  Just don't change the ABI itself. :-)

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