Re: C++ PATCH: Fix PR 862

[Jason Merrill <jason_merrill at redhat dot com>]

>>>>> "Mark" == Mark Mitchell <mark@codesourcery.com> writes:

>>>>> "Jason" == Jason Merrill <jason_merrill@redhat.com> writes:
Jason> It has certainly worked (on some targets) in the recent
Jason> past; it may need some tweaking to work with the new ABI,
Jason> but I would not be surprised if it worked now.  It doesn't
Jason> make sense to me to disable a useful (and actively used, if
Jason> mostly on embedded targets) flag because it might have a
Jason> couple of bugs.

> I agree -- but I had a totally different understanding of the
> situation.  The last time I talked to RTH about it, I thought it was
> totally broken w.r.t. the new ABI, and that it wasn't easy to fix!

I would doubt that, but in any case I don't think the specific degree of
brokenness is the issue; I don't see the problem with having flags for
functionality that is still under development.  -fnew-abi comes to mind.
Or, more similarly, -fssa.

> If you're correct that it mostly works, then I totall agree that we
> should reenable the option (even on the branch) -- if someone provides
> some invoke.texi documentation.

I agree that this flag should be documented, but I'm not sure all flags
should be; is it actually useful to say "here's this flag, don't use it"?

Jason